Controversy due to AI data center buildouts generally centers around their massive power usage and draining of local water reserves, but noise is a third torment to nearby residents, and one that is arguably much harder to correct. Residents of Sturtevant, slightly south of Milwaukee, Wisconsin, filed a class-action lawsuit against Microsoft due to the excessive noise produced by the company's Fairwater facility. CEO Satya Nadella described the project as "the world's most powerful AI data center," projected to generate 865,000 tokens per second and have a final bill of $7.3 billion.

The Sturtevant residents live just 1.5 miles (2.4 km) from the facility. The lawsuit was filed by three citizens and represents the households within this distance, reportedly amounting to over 1,000 homes, including areas in Mount Pleasant. The filing describes the noise situation as "not only excessive, but consistent and pervasive," and claims Microsoft did not "implement adequate acoustic barriers, shields, or walls that absorb, mitigate, and/or prevent the escape of noise, thereby resulting in the offsite emission of excessive noise beyond its property." One resident claims he had to change his shift at work to be able to sleep at all.

Nearby residents also complain about excessive dust and traffic stemming from the construction work, as well as light pollution. One resident says he often can't see his house coming into town, while another claims that before the data center arrived, the sky was dark and full of stars, now mostly gone due to the bright lights.

To some credit, Microsoft appears to have been trying to improve the situation, judging by a fairly detailed blog post on its community website. On June 18, the company said its engineers applied several measures that "fully resolved the issue," and that it would apply additional mitigations over the following months, including "additional sound reduction components."

The backstory in the post says the firm was aware of the issue back in April, and "[it] did not expect the tonal quality of the sound to travel as far as it has," attributing the decibels to cooling fans operating at too-high speeds, now purportedly corrected. In an earlier project update about the data center, Microsoft says it would have street sweepers working 10 hours a day and limit construction to hours between 6am and 10pm.

The lawsuit was filed on July 1, indicating that either the issue isn't resolved, or that the particularly short distance between Fairwater 1 and Sturtevant might make for a conundrum that's exceedingly tricky to solve. Some Mount Pleasant residents even live across the street from the campus, too. Before the Fairwater data center arrived, the land was already zoned for heavy industrial use back in 2017 to Foxconn, a status that carried over to Microsoft upon purchase. Notably, Wisconsin's "direct legislation" apparently does not allow amending or repealing existing ordinances, and is only available to cities, not towns.

The Cheyenne Board of Public Utilities has stopped accepting industrial wastewater from data center fill-and-flush and closed-loop cooling operations after tracing a rare bacterium in the city's reclaimed water to Goat Systems LLC, the entity Meta uses to build its Cheyenne campus. In a notice reported by Cowboy State Daily, the Board said Goat Systems was in significant noncompliance for discharging water carrying Cupriavidus gilardii, a metal-resistant bacterium that interfered with two water reclamation plants and pushed the reuse system offline for months of cleanup. The Board revoked the contractor's fill-and-flush discharge privileges on March 24, and a wider suspension now covers every data center connected to city services.

Fill-and-flush is a commissioning step in which crews fill a cooling loop's piping with water, flush it to clear debris before the system is run, and then send the used water to drain. Goat Systems routed that flush water, which contained Cupriavidus gilardii, into Cheyenne's sanitary sewer, Frank Strong, the Board's engineering and water resource division manager, told the Wyoming Tribune Eagle. Strong said the fill water had been purchased from the Board itself and that the origin of the bacterium remains unknown, but said that lab staff caught it in February during routine fecal-bacteria sampling. "This isn't something we normally test for," Strong told the paper.

Microsoft and Nvidia market sealed liquid loops as a near-zero-water alternative to evaporative cooling, an approach that is spreading quickly as AI data centers expand into more communities. Microsoft describes cooling systems that are filled once during construction and then recirculate the same water, while Nvidia's Rubin platform runs a coolant that is 75% water and 25% propylene glycol. That one-time fill, however, is the step that produces a discharge, and the flush water leaves the site before the loop is sealed.

Strong went on to add that the Board's concern extends past the finding of the bacterium, because closed-loop systems can carry glycol and other chemicals that municipal treatment plants aren't built to process. Cheyenne sprays its reclaimed water on parks, golf courses, and other green spaces, and the Board worried the bacterium could become an aerosol hazard during irrigation. Cupriavidus gilardii isn't a regulated contaminant, yet the discharge disrupted treatment sufficiently to trigger pass-through and interference findings under the Cheyenne City Code and federal pretreatment rules.

Meta said that it's supporting its general contractor, Fortis, which stopped discharging and began hauling wastewater offsite, and that independent testing found no trace of the substance. Testing at the Dry Creek and Crow Creek facilities cleared in late June, and the reuse system is back online. Cheyenne City Councilman Pete Laybourn called the disclosure "a very, very unpleasant surprise." The Board hasn't said how the suspension affects other Cheyenne data centers still under construction.

Ampera, a nuclear tech startup, just revealed its first full-scale, 3D-printed nuclear reactor module that it says could be used to power AI data centers. According to the company’s press release, this module is the “world’s first subcritical, solid-state, factory-built thorium nuclear reactor.” It’s also the nuclear part of the company’s Integrated Energy Architecture, which comes with a Waste Heat Recovery module that could also be paired with either this 3D-printed nuclear reactor module or a Conventional-Fueled Power Generation module.

"This next-generation nuclear core and pressure vessel sets the foundation for factory-built, mass-produced nuclear energy," said Ampera CEO and founder Brian Matthews. "The advanced technology and additive manufacturing used demonstrate a clear commercial path for new nuclear technology coming to market in an accelerated manner."

The company’s description for its groundbreaking product is quite a mouthful, but it shows how advanced this nuclear reactor is. Subcritical means that its fuel does not reach criticality on its own, requiring an external source to jump-start the fission process. This is inherently safer because it reduces the chance of a runaway reaction that could lead to a meltdown. It also features a “solid-state” core, meaning it does not have any moving parts, making maintenance and operation simpler.

Factory-built suggests that the nuclear module could be manufactured relatively quickly in a production line. This would make it easier to scale for mass production and transport via shipping containers, thus cutting production costs to make it more affordable. Lastly, it uses thorium as its primary fuel source. Thorium is more abundant than uranium, so it’s easier and cheaper to acquire. It’s also not inherently fissile, requiring an external neutron source to produce protactinium-233, which then decays into U-233. While it could still be used for nuclear weapons, it’s relatively harder to do so, which is why the World Nuclear Association says that it “is sometimes promoted as having excellent non-proliferation credentials.”

While the startup unveiled the finished nuclear module, it didn’t mention turning it on and producing electricity with the system yet. This is in contrast to Valar Atomics, which turned on its Ward 250 nuclear microreactor recently and produced electricity to power an Nvidia RTX Spark desktop. Its CEO, Isiah Taylor, also announced that it’s partnering with Nvidia to power a 30MW data center using its technology.

Ampera is also targeting a 30MWe output with its complete system using a proprietary closed-loop, supercritical CO2 Brayton-cycle turbine to produce energy. One advantage of its modular approach to developing its power solution is that customers can purchase them as necessary. It can match the turbine module with its Waste Heat Recovery module to supplement power production in data centers, or it could attach a Conventional-Fueled Power Generation unit to the entire unit as a site’s primary power source. Once the nuclear module becomes available, operators could then easily swap that out to reduce their carbon footprint.

The company did not give a timeline for when it expects to start producing power with their solution, but it seems that they’re confident that they could deploy it soon in various applications. "Our reactors are built for the markets that need power the most: AI data centers, defense, industrial and maritime," Matthews said. "We expect to be the first company to industrialize factory-built nuclear power with near-term deployment timelines."

Data center developer QTS has officially terminated plans to build a portion of the planned 2,100-acre “Digital Gateway” data center campus in Virginia, following years of legal battles with local residents and historic preservation groups. According to a Bloomberg report, the Blackstone-owned company formally submitted a written filing to the Virginia Supreme Court on July 2, explicitly stating it was withdrawing its last remaining appeal “after careful consideration”.

QTS’s withdrawal marked the official end of the project, as other stakeholders had earlier pulled out due to prolonged litigation. In a major win for opponents of the data center, the proposed land — situated at the edge of the Manassas National Battlefield Park, a historic Civil War battlefield — will now remain under its original rural zoning restrictions.

Digital Gateway was a planned massive 22-million-square-foot, gigawatt-scale data center complex in Prince William County, Virginia, that would have been the world's largest. QTS was responsible for 800+ of the 2,100 acres, while a second developer, Compass Datacenters, controlled roughly 800-1,000 acres. The rest of the designated 2,100-acre "Digital Gateway" zone consisted of local roadways, environmental buffer zones, and parcels belonging to individual homeowners who had agreed to sell.

Initially approved by the Prince William Board of County Supervisors to capitalize on the soaring demand for AI and cloud computing, the project immediately faced fierce opposition from residents, despite the promise of tens of billions of dollars in capital investment and substantial local tax revenue. Opponents argued that the facility's location adjacent to the Manassas National Battlefield Park posed severe threats to the area's environmental, historical, and residential landscape, a common refrain in growing anti-data-center sentiment.

Several lawsuits followed, with the one that appears to have eventually collapsed the massive project hinging on an ironically small detail. The landmark mega project — roughly double New York’s Central Park’s size with city-sized power needs — collapsed in a domino effect after Virginia courts voided the county’s initial rezoning approval due to improper public notices. The newspaper notices publicizing the hearing at which the project was approved weren’t separated by at least 6 days, as mandated by state and local codes at that time, thereby invalidating the hearing and the resulting approval.

Virginia courts upheld the invalidity of the zoning approvals in March this year. Following the ruling, Prince William County gave up first, refusing to spend additional public funds to defend the project in court. Compass Data Center, a co-developer of the project, withdrew from the project the next month, with the company president stating, “While we still believe this project offered significant benefits for the region and our neighbors, recent legal actions and compounding regulatory hurdles have effectively closed a viable path forward.”

Only QTS remained, filing an appeal with the Virginia Supreme Court. It has now withdrawn that appeal, leaving the entire Digital Gateway project completely dead. The company says it will now “work on a responsible and orderly wind-down of development efforts”.

The “Digital gateway” case is the latest win for anti-data-center proponents and the latest blow for those on the other side of the fence. We recently covered a report that more than 75 data center build-outs worth $130 billion have been successfully blocked in the first Q1 2026, although several others have moved forward. While data centers are needed to power AI’s exponential growth, their development often raises valid concerns about water use, noise pollution, and spikes in electricity bills for nearby residential areas. There has been an uptick in protests against data centers and a corresponding rise in companies pursuing innovative solutions to the problem. SpaceX is currently building an 11-million-square-foot Gigasat factory to manufacture orbital data centers, while several companies are deploying offshore data centers. On the other hand, there are reports that China may be artificially stoking opposition to data centers in the U.S., even as Trump's administration is pushing for AI development.

Meta is building a cloud business to sell excess AI computing capacity, according to a Bloomberg report that cites people familiar with the matter. The company is reportedly weighing two service models: selling developers access to AI models hosted on its own infrastructure, including its closed-weight Muse Spark model, in an arrangement similar to Amazon Web Services' Bedrock, or selling raw computing capacity in the same way neocloud providers such as CoreWeave do.

The initiative is reportedly dubbed Meta Compute and is led by head of infrastructure Santosh Janardhan, Meta Superintelligence Labs leader Daniel Gross, and president Dina Powell McCormick. Either path would put Meta in direct competition with AWS, Google Cloud, and Microsoft Azure. Judging by Wednesday's trading, though, the market thinks the hyperscalers aren’t the ones with the most to lose.

Meta has been one of the neocloud sector's most important customers. The company expanded its cloud computing agreement with CoreWeave to $21 billion in April and has signed contracts worth up to $27 billion with Nebius, roughly $48 billion committed to renting other companies’ GPUs because its own buildout couldn’t keep pace with demand.

Investors repriced that relationship within hours of the Bloomberg report. Meta shares rose more than 10%, the stock’s biggest single-day gain in over five months, after a year in which it had fallen nearly 15% and lagged the S&P 500. CoreWeave fell 10.8%, and Nebius dropped 12.4%.

“The impact of adding Meta's capacity to the market is more likely to be on neoclouds than the big hyperscalers. Those companies like CoreWeave and Nebius rely on Meta for their growth, and Meta may not need them anymore,” Gil Luria, managing director at D.A. Davidson, told Reuters.

The report didn’t come out of nowhere. At Meta's shareholder meeting in May, CEO Mark Zuckerberg said entering cloud computing was “definitely on the table,” adding that companies were approaching Meta “almost every week” to buy access to its AI models or spare computing power.

Meta raised its full-year 2026 capex forecast to between $125 billion and $145 billion in April, citing higher component pricing and competition for land, power, and construction labor. The same week, Zuckerberg told employees that roughly 8,000 planned layoffs were a direct consequence of the company's infrastructure budget.

That budget buys a heterogeneous fleet. Meta signed a 6 GW, $100 billion agreement with AMD in February, holds GPU deals with AMD and Nvidia worth roughly $110 billion combined, has announced four generations of its MTIA inference silicon, and struck a multi-billion-dollar Graviton deal with Amazon to cover general-purpose CPU shortfalls. Its Prometheus and Hyperion campuses are designed to scale to 1GW and up to 5GW, respectively.

Capacity at that scale arrives in large, indivisible increments timed to demand projections, which is how a company that was rationed on Gemini access by Google and paid neoclouds tens of billions for GPU time can simultaneously find itself with surplus compute worth selling.

We’ve seen a similar story with SpaceX. After xAI’s infrastructure, the company leased the entire capacity of the Colossus 1 data center in Memphis, more than 300 MW, to Anthropic for around $1.25 billion per month through May 2029, and subsequently agreed to rent capacity to Google for roughly $920 million per month. Bloomberg Intelligence estimates the arrangements could generate more than $50 billion by 2028.

The boom in data center construction meant that millions of dollars’ worth of servers, cabling, and other parts and components are on the road at any given time. Because of this, cargo thieves and hijackers have started targeting these high-value items. Business Insider reports that a trailer containing about $300,000 worth of copper wire spool, which can be used for powering data center servers, was recovered last week by the Cook County Sheriff’s Office in a truck yard near Chicago. The truck yard owner also said that the same driver dropped off another stolen trailer a week earlier, this time containing about a million dollars’ worth of data center equipment.

Both trailers have been reported stolen — the former taken from Pine Hill, Alabama, and the latter from Jacksonville, Florida. The original locations where these trailers were stolen show the nationwide scale of these theft rings, which had previously mostly targeted retail goods. The trailer that contained the copper wire spools also had its tags replaced with one from Indiana, according to Freight Pulse, probably in an attempt to obfuscate its source. It’s unclear how the data center equipment cargo was discovered, but the trailer carrying the copper wire was located through its GPS tracker.

Data center equipment, especially servers, seems like a lucrative target, especially given how expensive these items are. However, they are also specialized machines that are often only bought by institutions and large companies, so thieves would likely have a hard time selling them on the black market. They also likely have serial numbers, which would make it easier for buyers to check the status of the equipment with the manufacturer. Given the cost required to invest in these things, buyers would most probably demand official receipts and warranties — something that thieves and fences won’t be able to provide.

However, consumer electronics are easier targets, and we’ve seen theft rings hit consumer electronics before. For example, about $1.4 million worth of Switch 2 consoles en route to a GameStop branch in Texas were stolen from the back of a semi-truck last year. Gaming GPUs are popular targets for theft rings, too, with a shipment of EVGA 30-series GPUs stolen from the back of the truck in 2021 appearing in Vietnam a few months later, being sold by a major retailer at a discount. Another set of MSI RTX 3090s was also lifted directly from the company’s China factory in 2020, with the 220 GPUs estimated to amount to $336,500.

There are probably more incidents of theft that have been reported to the police but not publicly known. And although the value of the stolen goods could be staggering for the average person, companies and freighters often have insurance coverage, so they don’t have to worry about paying out of pocket for the entire loss.

The Chinese government quietly approved the Space Computing Industry Innovation Center in early June, which aims to bring together rocket and satellite manufacturers, semiconductor fabs, and AI tech companies to build a space computing network. According to the Beijing government, this aims to “connect the entire industrial chain of space computing and boost the development of the satellite Internet of Things (IoT) sector. Research firm SemiAnalysis said on X that China made this move a week before Elon Musk announced his AI1 satellite, which will run AI workloads while orbiting in space.

The center is set to be officially launched later this month, and it will focus on six major research areas: highly reliable, heat-resistant space-native computing chips, high-performance hyper-interconnected space computing payloads, space computing satellite platforms and standard systems, space-based large models under constrained power conditions, integrated space-ground cloud-based measurement and control networking, and space computing power service-oriented and tokenized operations. These are designed to build an orbiting AI data center that will not rely on Earth-bound energy sources and will avoid the bottlenecks that many ground-based data center developments face today.

While Beijing made this quiet announcement earlier than Musk’s AI1 reveal, we should note that Elon already had technical details available during the interview. In fact, the world’s first trillionaire has been talking about compute in space since November last year and filed for a one-million-satellite Orbital Data Center System with the FCC in February 2026. Jeff Bezos is also getting into the game with the 51,600-satellite Project Sunrise set to operate in a sun-synchronous orbit.

What makes China’s announcement different, though, is that it’s making multiple companies work together to build a system. On the other hand, SpaceX and Blue Origin are going at it alone — the two companies are competitors, and it doesn’t look like they’re cooperating with each other to develop the technologies required for space-based AI compute. It even seems that the former is intent on vertical integration, with its new 11-million-square-foot (around 190 to 200 football fields) Gigasat factory and Musk’s TeraFab megaproject.

We’re unsure which technique will be more effective in the long run: having one or two companies pour massive resources into this megaproject (with its success or failure being solely borne by those firms) or making several smaller companies and institutions work together to build a Space Intelligent Computing Research Institute, with the output presumably available for use by Chinese firms. But one thing is certain: Beijing is taking space-based data centers seriously enough to pour resources into them — a significant move for a nation with ample excess electricity and available infrastructure to build power-hungry data centers.

Amazon employees Darius Irani, Patrick Schloesser, and Liesl Wigand, who are members of Amazon Employees for Climate Justice (AECJ), say that they’re currently being investigated by their employer for breaking company policy after they testified in favor of the Seattle data center moratorium. According to Bloomberg, the three were called to separate Zoom meetings with an HR staff member after the city passed its one-year AI data center ban, where they were told that the company was investigating concerns about their public comments. Staff say they were intimidated during calls, are being monitored at work, and may face possible termination.

The company said that even though the three were free to discuss their working environment, they’re not allowed to speak as its representatives. “As we looked more closely at how these employees represented themselves, and how their comments were received by others, it became clear that they may have been speaking in their capacity as Amazonians and not as private citizens,” Amazon spokesperson Margaret Callahan told the publication. “We believe it’s important to apply our policies consistently so, just as we would with anyone else, we’re investigating whether there was a violation of our policies and may or may not take action based on what we find.”

Legal representatives of the three personnel said in a letter to the Seattle Office for Civil Rights (SOCR) that the investigations could lead to discipline, with one of them even saying that they could possibly be terminated. The lawyers argued that they were investigated by the company because of what they said when the Seattle City Council heard public comments on the measure, and said that Irani, Schloesser, and Wigand did not use company time when they made their comments, nor did they make any mention of their employer or shared propriety information.

The SOCR is a local government agency that enforces laws against employment discrimination, among other things, and the three affected employees are asking the agency to look into the situation. They alleged that they were intimidated during the Zoom meetings and complained about being monitored while they were at work. This also isn’t the first time that AECJ-affiliated workers faced an issue with Amazon. The company reportedly fired two people in 2020 who were part of the group’s leadership for putting a spotlight on Amazon warehouse workers’ safety (or lack thereof) during the pandemic.

Amazon says it is just investigating the situation and hasn’t made a decision yet. However, news like this will not do any good for the company and industry in general, especially as they gather bad press because of their AI data center projects. While the company has been trying to put its good side forward, like comparing its water consumption with other hyperscalers to highlight its high water efficiency, it still cannot shake off the perception that it could cause higher electricity costs, reduced water pressure, and noise pollution, among others.

The U.S. Federal Energy Regulatory Commission (FERC) is taking steps to ensure that AI data center operators are connected to the grid as quickly as possible. According to Bloomberg, the energy regulator is about to release an order asking grids to prove that they’re expediting AI data center connections, especially projects that “bring their own power, or curtail demand during times of high stress.”

“We promised some fireworks last month and we’re going to light the fuse today,” FERC Chairperson Laura Swett said in a meeting, Bloomberg reports. Commissioner David Rosner also said that any studies to enact these changes that grids will do must be completed in 90 days. These moves are in line with President Donald Trump’s “AI Action Plan” policy roadmap that he launched last year, which seeks to make it easier to build AI infrastructure.

Data centers need massive amounts of electricity — something that the U.S. grid is poorly-equipped to handle. Because of this, grid operators invested heavily in upgrading their infrastructure and then passed on their costs. PJM Interconnection, which is the largest power grid operator in the U.S., raised its power costs by 75.5%, with this increase largely blamed on AI data centers. The State of Maryland also filed a complaint with the FERC, as PJM Interconnection planned to charge it $2 billion for infrastructure upgrades needed for projects that do not directly benefit the state..

But despite FERC’s condition that data centers should bring their own power or reduce demand as needed for them to qualify for fast-track processing, this will still likely be a cause for concern for many Americans who are increasingly resisting data center developments in their backyards. After all, it’s not just power consumption that many communities are concerned about. Many data center developments are also being built in drought zones, where water shortages are already an ongoing concern. Some people are also concerned about the noise pollution that these facilities bring to otherwise quiet rural areas.

AI hyperscalers are keen to get their data center projects off the ground, especially as the demand for compute is still growing. The White House also wants these developments to go full steam ahead, especially as it deems that this infrastructure is crucial for the U.S. to win the “AI race” against China. But as long as citizens feel that these projects will threaten their quality of life, data center developers will only continue to see increasing opposition.

The city of McMinnville in Warren County, Tennessee has just passed an 18-month data center moratorium, allowing the city to determine the “electrical grid capacity, water and stormwater impacts, environmental and public health concerns, noise, and community fit” of these projects before even giving them the chance to apply for a permit. The Tennessee Lookout also listed several other bans that have recently been enacted or are under deliberation, like the larger Warren County, which is planning to vote on a temporary ban over its entire jurisdiction on the 22nd.

Aside from this area, other rural counties have passed moratoriums or are in the process of debating one. Coffee County, which sits adjacent to Warren, passed a data center ban at the same as McMinnville, while Knox County will vote on the issue on the same day as Warren County makes its decision. Even Nashville, the capital of the State of Tennessee, has passed a moratorium bill on its first reading with only a lone dissenter among the 40-member Metropolitan Council.

Tennessee is currently home to 63 data centers, according to the Data Center Map, with Elon Musk’s massive Colossus and Colossus 2 data centers located in Memphis. However, it seems that Tennesseans are pushing back on these developments, as several counties are considering and passing temporary bans. This is especially true for rural areas, where many data center developers take interest because of cheaper land, less regulatory friction, and weaker resistance.

“I think that they’re aiming these at … these smaller places in Tennessee because our land is so cheap, our laws are so lenient, the income isn’t taxed like it is anywhere else, and generally they’re just not going to get a lot of pushback, because a lot of people don’t know what’s going on, and they don’t know anything about this,” Kai Sage, a resident of McMinnville told the publication. “Luckily, a lot of people around here were informed, so as soon as we alerted people about this, people quickly reacted.”

These bans are not permanent, though. Instead, they’re designed to give local governments time to assess how it would go about these developments, helping prevent potential disasters that data center projects can bring to the local community. “A moratorium is not a permanent ban. It is a responsible time-out,” McMinnville City Administration Nolan Ming posted on social media. “It gives us time to study the issues carefully, update our zoning and land use regulations, and make sure any future decisions are based on solid information, not rushed approvals.”

No U.S. state has enacted a data center ban yet. Maine’s legislature passed a moratorium in April that would have paused all developments until October 2027, but Governor Janet Mills vetoed it at the last minute because it would affect one well-supported project in Franklin County. Still, the number of bans and moratoriums across the country have been steadily increasing, with 69 jurisdictions enacting one as of May 2026. Even Seattle — home to AI tech giants Microsoft and Amazon — has passed a one-year pause as its studies the community impact of data centers.

Most of these are just temporary delays, as governments try to come to terms with the massive amounts of resources that these projects demand. However, this is also throwing a wrench into AI hyperscalers’ plans to build more compute, whose demand is only expected to go up as AI becomes more advanced and gets more widespread adoption.

Data center water use is one of the hot topics right now in the U.S. and is one of the primary reasons why such projects are being blocked across the nation. However, Amazon claimed that it’s improving its water efficiency and that it uses the least amount of water on a per kWh basis among the AI tech giants, including Microsoft, Google, and Meta. Aside from that, it also said that despite using 2.5 billion gallons of water across the world in its data centers, it said that that amount is still a fraction of what other industries in the U.S. are using annually.

Amazon pointed at EPA data from 2017 that said Americans use 9 billion gallons of water daily just for landscape irrigation. This amounts to nearly 3.3 trillion gallons of water every year just for watering plants and gardens and doesn’t even include the irrigation needed for food production. This makes its 2.5-billion-gallon water consumption a literal drop in the bucket, amounting to just 0.075% of water used for keeping gardens green and thriving.

These gardens, however, have a direct benefit in that they could potentially keep ambient temperatures lower (versus an all-cement environment) and add to the aesthetics of the immediate area — the only question is how many people these developments benefit. Moreover, many new data centers in the U.S. are reportedly being built in areas suffering from water issues. While the overall data center water consumption might seem small if you look at the larger picture, we still cannot turn a blind eye to its potential effects on the local community.

For example, a Meta data center has allegedly caused a Georgia town’s deep-well water source to muddy — a sign that the level of the local water table is hitting low levels, allowing mud to be siphoned by the residents’ water pumps. Another site reportedly sucked 29 million gallons of water in 15 months, which caused low water pressures for residents. Reports like these make it harder for other data center projects to secure permits from local governments to start construction, especially as communities fear that these issues could happen in their area, too.

Amazon boasts water efficiency improvements

As the pushback against data centers’ egregious resource consumption has become front-and-center in many permitting fights, Amazon is showing that it’s taking steps to reduce its water consumption and claims that it’s on track to being “water positive” by 2030. It achieved this by implementing innovations in its data centers to reduce water use without reducing performance.

The company says that it mostly uses air cooling for its data centers, which uses up a lot less electricity compared to water cooling. But when ambient temperatures rise, it switches to Direct Evaporative Cooling, with the company spraying water on an absorbent medium, which the company describes as “a sophisticated, giant sponge,” and then runs the hot air through it to reduce temperatures by five to 10 degrees. It says that this is more power efficient than using chillers, reducing power use by about 20% to 25% during the hottest time of the day when power usage by other users is at its highest, too.

It also said that it raised the temperature thresholds in its data centers. While this meant that its servers ran hotter compared to previous years, it didn’t impact on the longevity of their hardware and they were still able to deliver the same amount of computing power as before. Through trial and error, the company discovered that it could keep running its servers using air cooling only, with its water-cooling systems only kicking in once the ambient temperature reaches 85 degrees F (or more than 29 degrees C). This resulted in 50% reduction in water use, says Amazon water specialist Beau Schilz.

Efforts like this are essential for data center operations to become sustainable. After all, despite the pushback against new data center projects, we cannot forever put a moratorium on them as AI’s demand for compute is increasing. But until AI hyperscalers can earn back the trust of the people and prove that they can build infrastructure without increasing costs for everyone else and hoarding the resources that communities need for their survival, then they can only expect opposition to continue.

Data center projects are being delayed nationwide as many opposition groups have successfully blocked their progress. According to Data Center Watch, a boutique research firm tracking data center development, at least 75 data center projects worth around $130 billion have been blocked or delayed across the nation in the first quarter of 2026 alone. By comparison, this is the same amount that was blocked throughout 2025.

This metric reflects the growing opposition to data centers across the United States. In late 2025, an Ipsos survey showed that nearly half of Americans didn’t want a new data center project near their neighborhood. However, this number jumped to 70% a few months later, indicating increasing resistance to data centers. Many of the concerns raised by residents include massive electricity price hikes, egregious water consumption, and potential noise pollution, among others. However, there have also been reports that Chinese users have been amplifying these issues while posing as Americans.

These concerns have resulted in numerous moratoriums passed in different jurisdictions. At least 69 local government units have enacted bans as of May 2026, with more joining the list since then. Seattle, the hometown of AI tech giants Microsoft and Amazon, is the biggest city to have passed a one-year pause on data center projects so far, affecting five proposed projects. It has even gotten to the point where Maine almost passed a statewide ban on large new data center projects until October 2027, with only the governor vetoing the bill because it affected one particular data center project.

Despite President Donald Trump’s push for AI development in the U.S., pushback against data centers has come from both sides of the aisle. While a few politicians have been pushing for data center development despite community opposition, many more have been listening to their constituents and passing regulations to halt or delay projects as they study how these projects could impact the community and mitigate any negative effects.

The number of projects being blocked across the nation is concerning for both the federal government, which is competing against China to gain the edge in AI technology, and AI tech companies like Anthropic, which is struggling to gain access to more compute. But if these developments damage the environment and the lifestyle of the people living around them, then it seems that Americans would rather do without them.

Data centers consume significant resources, often to the detriment of local communities. But while water use has been on the lips of those fighting back against data center deployments, the sheer scale of their impact wasn't entirely clear until recently.

Many new U.S. data centers are being built on drought-prone land, and their cooling systems, which seem to consume gargantuan quantities of water, only represent a small portion of their total water usage by 2050. So, what makes up this overall figure for water usage? Cooling, of course, plays a part, but ongoing energy demands and chip fabrication also make up a larger part of the story.

They're consuming how much?!

A recent Guardian report on water consumption cited figures from Xylem, a water technology company. That does raise some questions about the accuracy of the report, and led us to dig further into the numbers from other sources. Although they're not quite the same as Xylem's original reporting, they do paint a similar picture.

Quantifying the sheer amount of water data centers use isn't easy. Much of it depends on the specific hardware being used within the facilities, how they're cooled, and what the local grid infrastructure is like. In a recent MostPolicyInitiative report, it highlights that U.S. data centers in 2023 alone consumed 17.4 billion gallons of water. And that's before all the recent gigawatt+ scale data center projects kicked off. By 2028, direct consumption could increase by as much as 73 billion gallons as some of these new facilities come online.

But the keyword, there, is direct. The UNU Environmental Cost of AI's Energy Usage report from earlier this year highlighted how global data center electricity consumption required just under a trillion gallons of water in 2025. AI workloads account for around 20% of that, or 200 billion gallons. That equates to around 300,000 Olympic swimming pools. That share is projected to rise to 40%, or by 400 billion gallons, to a total of 600 billion gallons of water by 2030, giving it a global electricity demand in excess of the entire country of Nigeria and using enough water to supply 500 million people in Sub-Saharan Africa.

However, with 200 billion gallons of water (and counting) dedicated to AI workloads, how does that compare to other industries?

That's still a footnote compared to the or 26.4 trillion gallons used by U.S. agriculture in 2024 (as per USDA), but it is closing in on international oil refining numbers. OPEC produced roughly 86 million barrels of oil a day in 2024, and with a rough conversion of 0.4 barrels of water per barrel of crude, that's in the region of 550 billion gallons a year.

There are still industries using much more water than AI, and much more even than all the data centers combined. However, water use is growing dramatically, and it's not really coming from the increased cooling demands. It's coming from power demands and hardware manufacturing.

As Americans' feelings towards data centers sour and, in turn, increase their opposition to their development, addressing the growing water needs of these new deployments may be key for developers who want to see their projects reach completion. For residents facing water shortages, droughts, and water contamination, the danger is more existential.

Cooling innovation

For data center developers, the main way they've looked to address growing water demand from their facilities is to improve cooling. Moving away from evaporative to closed-loop, direct-to-chip cooling can have a huge effect on the water used in the cooling process.

That's a good thing. Indeed, Microsoft's Satya Nadella claimed that the company's newest AI data centers have cooling systems so efficient that they "can operate effectively with zero water consumption." He also compared these mega data centers to single restaurants in terms of annual water consumption.

For home PC enthusiasts, Nadella is talking about using the data center equivalent of an AIO watercooler, rather than just letting the hot water evaporate to cool it down. It's easily more efficient in terms of water usage, which is great from that direct-use perspective.

Closed-loop cooling is absolutely an important innovation in data center development, and combined with more exotic ideas like fanless liquid coolers and immersion cooling (perhaps even undersea and in-orbit deployments) could see data center cooling use very little water in the future.

But it's not the cooling that's the problem: it's the power being generated to run them. And the one downside to closed-loop cooling systems is it uses more power than evaporative cooling systems. As those power demands rise, so does the indirect water usage of these facilities.

Power is the issue

The vast majority of indirect water use by data centers by 2050 will be down to power generation, the Xylem study suggests, and if recent generations of GPU development are anything to go by, those power demands are going to be enormous.

Nvidia's Ampere generation A100 enterprise GPU had a TDP of 300-400W. An H200 of the Hopper generation has a TDP up to 700W. A Blackwell GB200 GPU can pull as much as 1,200W. The next generation Vera Rubin? That's now up to 2,300W per chip.

Once you start scaling these GPUs up to their full racks, the power consumption is extreme. Where traditional data center server racks consumed between 10 and 15 KW, the latest GB300 NVL72 designs could consume upwards of 150KW a piece. Vera Rubin might be more energy efficient, but its racks could consume upwards of 230KW each.

Data centers just weren't traditionally designed with this kind of density of energy demands in mind. Powering them alone will be an enormous challenge and require enormous quantities of water to do it.

However, other industries also consume far more energy than all data centers, and thus, consume more indirect water usage from power generation. For instance, steel and iron manufacturing, chemicals and petrochemicals, cement and glass manufacturing, and many other industries each use multiple times more power than all data centers combined.

Renewables, water recovery, and nuclear

For hyperscalers, the near-term solution to data center power problems has been to use (occasionally law-breaking) mobile methane jet turbines. These aren't too heavy on water consumption, but have their own environmentally damaging effects with heavy carbon emissions.

A longer-term solution to this issue, and hopefully the extreme indirect water demands of these facilities, will be a combination of renewables, nuclear power, and water recovery.

The Switch Tahoe Reno exascale data center shows how it can be done. It's a 650MW facility built in 2017 that uses 100% renewable solar energy. In Portugal, the SINES DC Start Campus is a 1.2 Gigawatt facility that's partially online and uses 100% renewables, as well as using seawater cooling to offset its water usage. Although many of the newer data center projects are far larger, as the price of solar deployment continues to plummet, it's certainly possible that data centers powered by renewables can be effective and profitable, without consuming such vast quantities of water.

Another option is nuclear energy. New, smaller reactor designs are making it possible to get these facilities online faster and with a more modular design. There's even the possibility of repurposing old aircraft carriers and submarines with onboard nuclear reactors to power data center facilities.

Data center developers know this is at least one path for the future. That's why they're rushing to secure access to key materials like Uranium.

The future of data center power, especially AI data centers, and the water that they require, is almost certainly some mix of renewables and near-site nuclear power.

As the mobile methane turbines popping up at data centers have shown, the developers will often just use what they can get their hands on. Perhaps alongside moratoriums and pauses in construction, data center protestors could make sure that if data centers are built in their local area, the developers should also be required to invest in water infrastructure to offset their ever-growing demands, alongside more renewable energy solutions.

An ongoing fight over a proposed data center sited just 50 yards from Nashville Zoo has escalated further, with the zoo’s land use attorney filing a zoning appeal to overturn permits already approved for developer DC BLOX. An opposition petition has already passed 331,000 signatures, with Grammy-winning country star Brad Paisley calling the project "an absolute nightmare scenario" in a video posted to Instagram over the weekend.

The pressure now shifts to the Metro Planning Commission, which is set to hold a public hearing Thursday, June 11th, on legislation that would ban data centers larger than 500,000 square feet across Davidson County and keep smaller ones at least half a mile from homes, schools, and zoos. The 69,220-square-foot facility at 648 Grassmere Park would draw an estimated 50 MW from the local grid, and the zoo argues its noise and light could disturb vulnerable species, including clouded leopards it is working to conserve.

Bill BL2026-1391, filed by District 20 Councilmember Rollin Horton and passed after a first reading earlier this month, will create Nashville's first zoning rules for data centers, according to the local news outlet WKRN. Facilities over 500,000 square feet would be prohibited outright, while those between 100,000 and 500,000 square feet would need Board of Zoning Appeals approval following a public hearing. All data centers would have to run closed-loop cooling systems that don’t return water to the public supply, and smaller facilities would face buffer zones of 100 to 500 feet from homes, daycares, churches, parks, and other data centers.

It's been reported that the bill will also look to restrict the use of backup generators to emergencies and testing, and require developers to prove Nashville Electric Service has capacity before approval. Governor Bill Lee signed a separate state law last week requiring data centers to cover the cost of grid infrastructure upgrades their facilities need.

The DC BLOX project itself falls well under the size that the bill seeks to ban, and the company already claims closed-loop or waterless cooling for the site, but it’s the half-mile zoo buffer that’d be the blocker should it come to pass. The zoo told NBC News that land use attorney Bill Herbert, a former Metro Codes Director, filed the appeal to overturn DC BLOX's approved permits.

"As we move forward, we have taken the next step in our fight against the proposed data center," the zoo said in a statement to NBC News, confirming that land use attorney Bill Herbert, a former Metro Codes Director, filed the zoning appeal to overturn DC BLOX's approved permits. A company spokesperson told the outlet the facility “would not be an AI factory placing a burden on local resources” and sits on land that previously hosted a data center, though he couldn't say what the new facility would be used for or whether AI companies would be customers.

The standoff in Nashville fits a pattern of opposition to new data centers we’re seeing nationwide. A Gallup survey from last month found that 70% of Americans oppose data centers built near their homes, and Seattle is set to pass a one-year moratorium on new AI data center construction. If Horton's bill passes, Nashville would have some of the strictest siting rules of any major U.S. city, written in direct response to a facility too small for most AI workloads.

Samsung Heavy Industries signed a three-way agreement with Greek shipowner Capital Clean Energy Carriers and classification society Lloyd's Register at the Posidonia shipping exhibition in Athens earlier this month to commercialize a 50MW floating data center, according to Chosun Biz. The deal followed the signing of a joint development agreement between the shipbuilder and Supermicro to verify whether AI servers can run reliably on water, with the design pairing seawater cooling with LNG-fueled fuel cells for onboard power. Under the proposed business model, shipowners will buy the platforms and lease capacity to operators on long-term contracts, much like how they charter tankers.

The memorandum of understanding splits the work three ways: Samsung Heavy handles technology and construction, Capital leads project sourcing and investment, and Lloyd's Register covers regulation and certification. Samsung Heavy signed a second memorandum with consulting arm Lloyd's Register Advisory covering feasibility studies and an assessment of the North American market.

Receiving approval in principle from the American Bureau of Shipping and Lloyd's Register in April, the 50 MW project can draw external power through subsea cables when moored in ports or coastal waters, or generate its own using solid oxide fuel cells running on LNG, sidestepping long grid connection queues that have stalled land-based projects in the U.S. and Europe. Jerry Kalogiratos, CEO of Capital Clean Energy Carriers, said in Lloyd's Register's announcement that floating data centers offer "a scalable and flexible solution, with the unique advantage of mobility."

Meanwhile, the agreement with Supermicro is designed to address the question of whether precision AI hardware can tolerate environmental factors like vibration, tilt, salt, and humidity over a multi-year service life. Samsung Heavy will develop offshore positioning control and technologies to seal out salinity and moisture, while Supermicro validates server operating conditions in river and marine environments. The shipbuilder is leaning on its floating LNG production facility experience to integrate power, cooling, networking, and safety systems into a single hull.

Besides Samsung Heavy, Japan’s MOL is also building a 73 MW floating data center with Karpowership for a 2027 deployment, China's 24MW subsea facility off Shanghai entered full operation last month, and Nautilus Data Technologies operates a much smaller 6.5MW barge at the Port of Stockton, California. Samsung Heavy also has a potential anchor tenant, having signed a letter of intent with OpenAI in October alongside Samsung's 900,000-wafer-per-month Stargate memory commitment, covering joint development of floating data centers. To date, however, there’s no signed customer contract for a named deployment.

China is drafting a plan to spend roughly 2 trillion yuan ($295 billion) over five years on a nationwide web of AI data centers. The goal is for at least 80% of the underlying technology, AI chips included, to be sourced from domestic suppliers such as Huawei, according to a Bloomberg report citing people familiar with the discussions.

The National Development and Reform Commission is responsible for the blueprint of this network, while state carriers China Mobile and China Telecom will operate most of the facilities and link them up to a single computing grid by 2028. The build-out of this grid leans heavily on sovereign debt and ultra-long special government bonds. Folding in power grid upgrades could push the total capital requirement beyond 5 trillion yuan, those sources told Bloomberg.

Funding the build-out is easy, though; filling them with domestic accelerators is a different story. The 80% domestic sourcing requirement effectively locks out Nvidia and AMD accelerators, so China will be capped by whatever amount of chips SMIC can physically produce. The foundry’s most advanced stable node remains its N+2 process, which is roughly equivalent to 7nm and is currently running above 93% utilization, leaving little headroom as every government-certified Chinese chipmaker competes for the same wafer slots.

Another major chokepoint is high-bandwidth memory. Highly limited domestic HBM supply constrains how many Ascend-class accelerators Huawei can assemble. Huawei alone shipped around 812,000 chips last year and projects some $12 billion in processor revenue for 2026, a pace that its own supply chain has struggled to sustain. It’s estimated that China’s domestic suppliers will cover only around 76% of all Chinese AI chip demand by 2030, even as that market grows toward $67 billion.

Beijing has massively tightened its restrictions on foreign silicon in a series of new controls. Last August, Beijing introduced a requirement that data centers source at least 50% of chips locally, and by November, state-funded projects were barred from foreign accelerators entirely, with builds less than 30% complete reportedly told to strip out Nvidia, AMD, and Intel parts.

China's own industry has questioned whether domestic hardware can keep pace. SMIC co-CEO Zhao Haijun has cautioned that the rush to add capacity risks leaving data centers idle, comparing it to building highways ahead of the traffic. Chinese chip executives have separately conceded the country trails the leading edge in AI data center silicon by five to 10 years. When DeepSeek was steered toward Huawei hardware for model training, it eventually reverted to Nvidia hardware, lending credence to the idea that domestic parts still struggle with the heaviest training workloads, even where they suffice for inference.

Elon Musk has laid out the first detailed design of SpaceX's AI1 satellite in a 30-minute video posted to the company's X account, the opening generation of an orbital craft SpaceX wants to build by the million to run AI workloads off Earth's power grid. Carrying a 150 kW peak compute payload across a 70-meter deployed wingspan, the spacecraft uses an interchangeable hardware design that lets different chipmakers supply the processors. The timing of this announcement is no accident, coming just three days before SpaceX’s IPO, which is set to price on June 11th and trade on June 12th at a target valuation near $1.75 trillion.

In his announcement, Musk pegged the satellite’s compute payload at roughly the draw of a single Nvidia GB300 rack, which pulls around 140 kW on the ground. One AI1, in those terms, is about one rack in orbit. In terms of its overall specs, SpaceX disclosed an average compute payload of 120 kW, a peak of 150 kW, and a density of 70 kW per ton, with the craft operating at roughly 600 km.

A satellite with these specs comes with some serious space requirements, and its 70-meter deployed wingspan edges past the 68.4-meter span of a Boeing 747-8. As for the interchangeable compute, that leaves the platform open to whichever vendor ships the most competitive AI silicon, rather than locking it to a single supplier.

This interchangeability is no doubt important to Musk, not least because SpaceX can’t yet guarantee its own supply of chips. The company is currently building Terafab, a chip fab that’s running as a joint venture with Tesla, while its S-1 IPO filing warns it can’t currently secure enough chips.

That aside, the elephant in the room is cooling: a rack on Earth sheds heat into moving air and circulating water, neither of which exists in a vacuum, where the only viable route is radiating it away as infrared. AI1 features up to 110 m² of deployable liquid radiators, as well as redundant pumping loops and integrated micrometeroid shielding. By comparison, the International Space Station’s ETACS rejects roughly 70 kW of heat — around half of what’s needed to cool a 140 kW GB300 rack — across 422 m² of radiator at a cost of up to $500 million, according to SemiAnalysis.

Musk has previously waved off potential thermal critiques, telling SpaceNews back in March that it's "safe to say SpaceX knows how to do heat rejection in space" and pointing to the company's fleet of more than 10,000 Starlink satellites.

SpaceX filed with the FCC in January to launch up to a million orbital data center satellites and has already signed compute deals, including a $920 million-per-month agreement with Google. The model has prominent skeptics: OpenAI's Sam Altman called orbital data centers "ridiculous" earlier this year.

SpaceX has announced a new 11-million-square-foot Gigasat factory in Bastrop, Texas, dedicated to building the infrastructure needed to achieve the company's orbital data center goal. In an internally conducted interview posted on X on June 8, SpaceX CEO Elon Musk confirmed that the facility is expected to begin producing complete "AI satellites" by 2027. The company is targeting 1 gigawatt (GW) of orbital AI compute capacity by the end of next year, with plans to scale that figure by an order of magnitude annually thereafter.

During the interview, Musk unveiled the build and specs of the proposed AI1 satellite, the company’s take on orbital data centers. The satellite will span roughly 70 meters (229.66 feet), with a massive solar array — generating power at a density of 250 W/m² — making up the bulk of the structure. AI1 will also feature vertically oriented, double-sided radiators for cooling, with the 150-kilowatt (kW) peak compute payload positioned right in the middle of the structure.

The massive 11-million-square-foot Gigasat facility — more than ten times larger than Starfactory, SpaceX's current largest spacecraft manufacturing complex — will vertically integrate much of the AI1 satellite supply chain on a single campus. The facility will manufacture solar ingots and wafers, solar cells, printed circuit boards (PCBs), silicon-based electronic components, user terminals, gateways, and the AI1 satellites themselves. The 1,000-acre site will also include dedicated satellite development and testing facilities, warehousing and logistics infrastructure, and a large-scale AI satellite production line.

According to Musk, the solar manufacturing facilities are already under construction, while the AI satellite production building is about to break ground. SpaceX expects to be producing a "reasonable volume" of these orbital data centers by the end of 2027. While each satellite will carry 150 kW of compute power, the company aims to achieve 1 GW per year of space AI compute in that same time. That would mean launching on the order of more than 6,000 AI1 satellites in a single year. For context, Starlink has about 10,500 active satellites as of June 2026.

Musk hopes to scale to 100 GW per year by 2030 and even has eyes on Terrawatt-level computing, completely solar-powered in space. "This is what we are going to try to do and think we probably can do, which is to get to roughly an annualized rate of a gigawatt per year by the end of next year at Space AI compute. And then aspirationally, scale that by an order of magnitude per year. So in two and a half years, hitting an annualized rate of 10 gigawatts a year at Space. And three and a half years, maybe 100 gigawatts," he said, while also expressing a desire to one day scale to a terawatt per year, depending on progress in chip making.

The largest AI data center anyone has actually announced is Meta's Hyperion in Louisiana, designed to scale up to 5 GW and house roughly 2 million GPUs at full buildout, at over $100 billion — and even that only reaches its first 2 GW phase by 2030. xAI's own Colossus 2 in Memphis just expanded to nearly 2 GW of capacity, with 555,000 GPUs, for about $18 billion, making it the world's largest single-site AI installation. So 100 GW/year is approximately 20 Hyperions or 50 Colossus-2s per year.

Achieving this will take an unprecedented volume of chips, and Musk's answer is Terafab — a SpaceX/Tesla/xAI venture in Austin aiming to fab 1 terawatt of compute a year, roughly 100 to 200 million advanced chips, in a 100-million-square-foot plant. However, the project itself is being met with widespread skepticism. For starters, none of the three companies has ever made a chip, yet they're starting at the bleeding-edge 2nm node. Our in-depth Terafab industry analysis explores various other reasons why this skepticism is valid.

Multi-Gigawatt ambitions aside, Gigastat is a significant step towards orbital data center goals that are increasingly seen as a potential viable solution to the extreme power consumption of on-ground data centers. Not unexpectedly, SpaceX is leading the race. Manufacturing solar arrays, wiring, and satellite bodies at volume is largely conventional work — much of it built on technology SpaceX already produces for its Starlink V3 satellites, just at far greater scale — which is part of why a 2027 production start is credible.

About two-thirds of the 809 data centers planned across the U.S. are slated for land that has been in drought over the past year, an analysis from The Guardian found this week. The research claims that 517 data centers are set to be built in areas classified as drought-stricken in the last year, according to the National Oceanic and Atmospheric Administration's (NOAA) National Integrated Drought Information System.

According to the report, data center cooling accounts for only about 4% of the additional water AI will demand by 2050, according to a January report from Xylem and Global Water Intelligence, while power generation takes roughly 54% and semiconductor fabrication about 42%. A UN University report published last week reached a similar split, putting 2025 data center electricity use at about 448 TWh.

The chips inside the racks, and the power to run them, account for the rest. A modern logic fab consumes between 2 million and 10 million gallons a day, and chipmaking can’t use ordinary water. Producing 1,000 gallons of the ultrapure water required by fabs takes roughly 1,400 to 1,600 gallons of municipal supply, so the input is lossy before a single wafer is etched.

TSMC's three Phoenix fabs, for example, are projected to draw a combined 16.4 million gallons a day once complete, in the fourth-driest state in the country. The company offsets much of that through on-site reclamation, rated at 85%, climbing toward 90%, but it’s still drawing water from an area where it’s already in short supply.

Data center operators routinely note that the sector uses a fraction of what agriculture does, but that’s only accounting for one of three legs. Only counting cooling excludes the fab and generation demand created by the same infrastructure, which is where most of the growth capacity is concentrated. A data center and the fab supplying its accelerators can sit in the same drought area and pull from the same groundwater, but only one of them will end up showing in the cooling figures.

A potential fix will be sealed, direct-to-chip liquid cooling. Nvidia rates its GB200 NVL72 system at up to 300 times the water efficiency of air cooling, but that figure covers the cooling loop alone. Those racks pull 120 kW to 140 kW each, and the Vera Rubin platform, arriving later this year, pushes a single rack toward 600 kW. More power per rack means more generation, and thermal and gas-fired generation is itself water-intensive, so rack-level water savings reappear at the power plant.

Meta's proposed Hyperion data center in Louisiana is one example. Closed-loop cooling here will be paired with the output of roughly 10 gas-fired plants, which consume water of their own to generate that electricity. So, on-site cooling water usage falls while the combined draw of cooling and generation rises, and as rack power scales toward Rubin's 600 kW, that combined figure will climb faster still.

Several states have started legislating around cooling, including California, Michigan, and Iowa, which are weighing mandatory water-use reporting. South Carolina and Kansas may require closed-loop systems, and New York lawmakers have floated an outright data center moratorium. Each measure targets the visible 4%, and leaves the fabrication and generation demand that drives most of AI's water growth untouched.

Back in 1999, 87 acres of land in Taylor, Texas, was donated (nominal fee $10) to the city by a farmer, with a condition in the deed that it would be used for community parkland. In 2025, the land was sold for $10M to a data center developer, who has won several legal battles against the nearby residents who are trying to stop the massive construction project, reports 404 Media. Now, the disgruntled locals are planning to take their case to an appeals court.

To introduce this case, let’s go back to 1999, when the $10 deed was inked. For some background, 404 Media talked to long-time local Pamela Griffin, who used to play on the farmland, and watched her children grow up and enjoy the same freedom. Griffin recalled that old farmer Mr. Bland used to talk to her father from time to time. According to her, Bland once said to her dad, “I see the kids don’t really have nowhere to play.” He continued, “I’m thinking about giving this land for parkland because these kids need somewhere to play.” The original July 1999 deed has since been unearthed, and the farmer did indeed follow through with his words. Now, let’s make the following chain of events simple using a bullet point timeline:

• Pre 1999 – a farmer’s promise to his neighbors,
• July 7, 1999 – Bland granted the land to the Texas Parks and Recreation Foundation, a public trust, for $10 on the condition it be used as a park,
• 2003 - Texas Parks and Recreation Foundation granted the land to another non-profit called the Williamson County Park Foundation,
• 2003, one month later, Williamson County Park Foundation gave the land to the City of Taylor,
• 2008 - the city of Taylor sold the land to the Taylor Economic Development Corporation (TEDC) for $15,000,
• 2025 – TEDC sold the land to data center developers Blueprint for $10 million.

This is quite a tale, and there appears to be a lot at stake for parties on either side of the dispute. In summary, locals face a multitude of undesirable side effects coming from having a data center in their backyard. There’s also the principle of the original deed being ignored, a big deal in Texas. Meanwhile, the council asserts that stopping this kind of development (in the city zone it is situated) is beyond its scope, and the millions in tax raised will benefit residents.

Griffin only found out about the planned 135,000-square-foot data center on the land when local organizers called around the neighborhood to raise awareness in 2025. She didn’t even know what a data center was at the time. Looking it up with her family, the idea of such a construction on your doorstep didn’t appeal.

Locals like Griffin became concerned about impacts to air, water, electricity, and noise. In correspondence with the City Council, they have been assured that work will be done to minimize health risks to people living nearby. Mitigations like a barrier wall, landscaping, closed-loop water cooling, and the developers building their own power substation were mentioned.

While folks might not want to live near a data center for the aforementioned reasons. It may also negatively impact the resale price of nearby homes. However, the City Council says that the expected extra $30M in tax revenue over the coming decade will be a positive for the area. With $20M earmarked for the school district.

Meanwhile, the council has also painted its situation as one where it is powerless to resist the data center development. This was made clear in a council website FAQ. 404 Media investigations revealed that this lack of power to change the course of the development is likely because of the property’s existing Employment Center zoning. The City can only regulate form, not function, it seems. However, the dev hasn’t yet secured the City’s approval for planning and building permits.

With the lack of satisfactory response from the council, Griffin and her family decided to hire a lawyer. There then followed a series of legal disputes, which, so far, have favored Blueprint. Nevertheless, Griffin and family members are filing an appeal with the Third Court of Appeals in Austin, Texas.

Land deeds are powerful documents in Texas. Activists opposed to the data center unearthed the original July 7, 1999, deed, which is shared on the 404 Media site. Indeed, we can see the legal document does stipulate that the 87.97 acres of land “be held in trust for future use as parkland.”

Central to Griffin’s determination to continue is the notion that “I’m not fighting just because of a data center. I’m fighting because this land was deeded for parkland.” Texas deeds need to be upheld, and the community should have a park, according to the resident.

SpaceX just announced that it closed a multi-year deal to provide compute capacity to Google. The agreement, which is worth $920 million per month, will begin in October 2026 and is expected to continue until June 2029. Reuters said that the transaction includes 110,000 Nvidia GPUs, plus CPUs, memory, and all other components needed for AI processing.

It appears that Elon Musk's company will not deliver the entire 110,000-strong GPU compute capacity in one go — Google will pay a reduced monthly fee as the company brings more server racks online through September 30, 2027. If SpaceX cannot hit the 110,000-GPU target on that date plus a one-month grace period, then Google can cancel the agreement or settle for the lower number of available GPUs “with a corresponding pro-rata reduction in the monthly fees.” It also gave the two parties the option to cancel the deal altogether after December 31, 2027, provided that they give a 90-day notice to the other.

This is the second major deal that SpaceX announced in months, as Anthropic secured the entire computing power of SpaceX’s Colossus 1 data center in early May. This was a surprising move, especially as Colossus 1 is one of the company’s most hyped assets, which Elon Musk launched in just 19 days. It turns out that launching it at such speed meant that it has a mix of H100, H200, and GB200 GPUs, which is resulting in efficiencies for training AI LLMs as the faster GB200 GPUs end up waiting for the older, slower GPUs before it can complete each computational step. Anthropic is instead using it for inferencing, especially as it is struggling to keep up with the demands of its growing user base.

The combined annual value of just these two deals is already worth more than SpaceX’s entire revenue for 2025. Reuters estimated that they would bring in more than $25 billion annually to the company, compared to the less than $20 billion that it made from Starlink, launch services, and AI revenue.

These massive deals, worth more than $70 billion in total, will lift SpaceX as it targets a $1.75 trillion IPO on June 12, 2026. While it started out as a space exploration company and is known for commercially launching satellites at a fraction of the cost compared to NASA and providing relatively affordable and stable satellite internet, it’s actively expanding towards orbital data centers. SpaceX acquired xAI earlier this year to help achieve that dream and has even filed some documents at the FCC detailing its plans. Google is also reportedly in talks with the company for a slice of the orbital data center pie.

Two city council committees just passed a data center moratorium in Seattle, making it the biggest jurisdiction in the U.S. to enforce a ban on these projects. According to The Guardian, the moratorium and an accompanying resolution must still go through a full council vote on Tuesday, June 9, 2026, but this is just a formality given the massive support that the data center ban has from the general public.

Several other towns, cities, counties, and areas in the United States have passed data center bans, especially as projects like these have increasingly become unpopular among Americans. Still, many didn’t expect tech-industry heavy Seattle to follow them in their footsteps.

Seattle mayor Katie Wilson first heard about the plan to build five large data centers near the city in April, after the Seattle Times broke the story. “That was the first that I, as the mayor, had heard about this,” Wilson said. “Both I and many of the councilmembers were happy to move toward a moratorium, especially knowing that there was really strong public support out there for that course of action.”

The city intends to spend the one-year hold on data center development to study its impact on the community surrounding it and what the local government should do to ensure that its constituents benefit from the project. This includes putting up pollution standards, determining energy connection requirements, setting contract terms, setting labor standards, and other regulations that will apply to data centers. Wilson also mentioned that Seattle will ascertain if these projects are a “good use of urban land” during this period.

Seattle’s residents supported the temporary moratorium, with more than 50 residents showing their support for it when the city hall opened public comments on the measure. Aside from increasing electricity bills and impacts on the water supply, some of the concerns that were raised at this time included the climate impact of data centers, especially behind-the-meter projects that are often powered by jet engines running on fossil fuels, noise pollution, and the conversion of arable land into sprawling campuses.

The city council’s decision to pass the one-year ban is in stark contrast to the actions of some politicians in other states that support data center developments despite opposition. For example, one Indiana mayor was caught on camera insulting his constituents who were against a proposed data center project in their area, while a Utah senator slapped the phone out of the hand of a reporter who was covering a story about the cases of harassment against his business because of his support for a 40,000-acre site in Box Elder County.

Meta has moved from building traditional structures for its data centers to putting up tents across the U.S. and sticking AI servers inside them. Michael Thomas, founder of market intelligence and data center tracking firm Cleanview Energy, said on X that the AI tech firm has already built or is in the process of constructing three data centers that use the strategy.

One site, located in New Albany, Ohio, already has five buildings that took approximately two to three years to complete. The company then started putting up five tents, with an area of around 125,000 square feet each, in the area. City permits seen by Cleanview Energy say that the construction started in April 2026, while recent satellite images show that the structures have already been completed.

Meta CEO Mark Zuckerberg first announced the strategy of pitching tents and filling them with AI servers last year. It seems that he wanted the infrastructure to come online quickly while demand for compute is increasing exponentially. It’s said that Meta is inspired by Elon Musk’s feat with xAI, which built a 100,000-strong AI data center in just 19 days in 2024 — something which usually takes four years, according to Nvidia CEO Jensen Huang. The technique is apparently quite effective, and it’s now being applied to two other sites, including one in Tennessee.

Putting AI servers inside tents, officially called “rapid deployment structures,” is one of the more unique approaches to the AI build-out, Thomas said. They’re certainly not as sturdy as physical buildings made from steel and concrete, with one commenter comparing it to the “classic $10k racing bike with a $9 lock” situation. Nevertheless, the company has probably weighed the pros and cons of such a setup and has decided that it was worth taking the risk to gain an advantage in the AI infrastructure race.

Another factor that allowed Meta to bring its data centers online at a much faster pace is its use of “behind-the-meter” power, in which the company installed its own turbines to produce power on-site rather than relying on grid power. This is similar to what Musk did with his Memphis Supercluster, which he initially powered with portable power generators. However, Meta’s turbines would be a permanent feature on the Ohio site, as it’s designed to run independently of the power grid.

Thomas said that there’s about 2GW of capacity available from behind-the-meter data centers, with an extra 1GW expected to go online this year for a total of 3GW. If on-going projects stay on schedule, the Cleanview report says that the total capacity could hit 13GW by the end of 2027 — about the same amount of power generated by 13 nuclear power plants and more than enough to power NYC.

This combination of seemingly temporary structures and jet engines strapped to the ground is certainly an interesting combination. As one reporter said, “It’s like a scene out of the movie Mad Max.”

Scott Ferguson (R), the mayor of Shelbyville, Indiana, was caught on camera insulting community members who were opposed to a proposed data center project near the city. The mayor's office has been forced to issue a statement after he was secretly recorded claiming he had only seen 'No AI data center' signs in "sh***y," unkempt houses.

Shelbyville is a city of 20,000 people located around 30 miles southwest of Indianapolis International Airport and is the county seat of Shelby County. Fox 59 reports that its city council has advanced a billion-dollar data center proposal in the area, amid increasing resistance from the public.

“I’ve seen a lot of these all over town, but I only see ‘em in sh***y houses with sh***y,” Ferguson was recorded commenting on "No AI data center" protest signs. Another person cut him off and corrected him, saying, “You see them in working class houses.” The mayor then continued, saying, “most of them are rentals.” Another off-camera speaker said, “It doesn’t matter whether they’re rentals, they’re still human beings.” To which he answered, “Yeah, I know, but they’re unkempt properties.” It’s not clear if Scott knew he was being recorded.

These comments have understandably caused a lot of controversy, and the video has blown up online. One commenter said, “Tell us you aren’t getting paid under the table without telling us,” while another added, “Wow… So, we know who the expendable people are for him.” Another person said, “Disgusting behavior. Who insults their constituents?”

In response to this, the mayor’s office released a statement saying, “The Mayor regrets that his choice of words may have caused offense. His comments were intended to reference property maintenance and not the character, value, or importance of any resident, homeowner, or renter in our community.”

Ferguson’s offhand remarks are adding a lot of fuel to the fire against data centers. The project that the Shelbyville city council approved is a 429-acre site that will play host to 11 data center structures. Fox59 reports that more than 2,000 people signed a petition against the project, but were ultimately ignored by the council. Despite the city council giving it a go signal, the data center developer must still go through several other hoops before it can begin construction.

The majority of Americans are now against having a data center constructed in their community, especially as several issues of increased utility prices and poor water quality have been linked to recent projects.

It has gotten so bad that many town halls and meetings have become contentious, and a few politicians who supported data centers are becoming quite defensive. For example, one farmer was arrested for going a few seconds over the allotted speaking time and was charged with trespassing in court. A Utah State Senator also slapped the phone out of the hand of a reporter who was covering the public harassment against his employees for his support for another project in the state. Most worryingly, one Indianapolis politician had their home shot at by an unknown assailant who left a “No Data Center” sign after the incident.

With the debates around data centers becoming more passionate than ever, some U.S. authorities are now worried that “anti-tech extremists” will take advantage of the situation and execute acts of violence against AI data centers. On the other hand, critics say that this could lead to mass surveillance and even the criminalization of peaceful opposition to data center projects.

SpaceX has won a 100% property tax abatement from Grimes County, Texas, for its proposed, massive $55 billion Terafab semiconductor chip manufacturing facility. According to a local news outlet, KBTX, Grimes County commissioners voted 4-1 on Wednesday, June 4th, to approve both a reinvestment zone designation and the tax abatement, which will fully exempt SpaceX from property taxes tied to the project in exchange for a regular payment instead.

In exchange for the tax exemption, SpaceX will instead remit a lump sum of $10 million to the county, followed by $20 million annually for 35 years. The vote — centered on a three-part proposal for the SpaceX Terafab facility, consisting of an obligation framework for jobs and infrastructure, a Tax Increment Reinvestment Zone designation, and a 35-year PILOT tax abatement agreement — followed a hearing that drew in more than 100 residents to the Grimes County courthouse.

Proponents of the proposals, including local officials, argued that the project promises massive economic investment, job creation, and national security benefits. Conversely, residents opposed to the project cited a lack of transparency in the approval process, potential strain on water and power infrastructure, and the destruction of rural identity. Commissioner David Tullos, who cast the lone “No” in the vote, questioned SpaceX’s absence in earlier meetings, saying that the hearing was the first time it had sent a representative since the project was announced. Tullos also questioned the company about the size of the proposed reinvestment zone and the company's plans for portions of the land included within it.

Several residents reportedly expressed deep concern about environmental impacts and land loss amid tears. In response, John Federspiel, senior director of Starlink Product Engineering at SpaceX, said: "We recognize that large projects bring legitimate questions about infrastructure and environmental stewardship. Our company is committed to proactively ‌addressing those ⁠concerns and taking care of them responsibly.”

Despite an apparent majority opposition from residents who urged the commissioners to delay the vote on the tax abatement — which public records show SpaceX applied for just last month — all proposals passed. A number of local residents expressed disappointment in the outcome, with many insisting that a 100% tax rebate — for a company about to launch the largest IPO in history with an expected $1.75 Trillion valuation — was too far.

Defending the arrangement in an X post on Wednesday, SpaceX owner and world's richest man, Elon Musk, said that the exemption was necessary to remain competitive with other chip manufacturers. “The reason SpaceX asked for this, which is standard practice for massive capital investments, is because Terafab will have a large number of extremely expensive machines for making chips. Property tax on these crazy money machines would put us at a serious competitive disadvantage relative to other chip fabs in the world,” Musk said.

He also implied that the county stands to significantly benefit from the project. “Something that perhaps isn’t clear is that, if this location works out (other locations are still in the running), SpaceX will still be paying an annual amount that increases tax revenue for Grimes County by ~25% and will be by far the biggest source of revenue for the county,” Musk argued.

SpaceX is advancing the Terafab project as part of its plans to expand beyond aerospace into computing infrastructure and domestic chip production, a strategic pivot investors see as vital for future growth. The company already owns Colossus 1, a massive data center it rented out to Anthropic, and is currently building Colossus 2.

Microsoft CEO Satya Nadella has claimed that the company's newest generation of AI data centers uses so little water that its annual consumption is roughly equivalent to that of a single restaurant. Speaking at Microsoft Build 2026 on June 2, Nadella made the claim while outlining Microsoft's "Community-First AI Infrastructure" strategy aimed to address growing concerns about the impact of data centers on local communities.

"The cooling loop is filled once, and the data center can operate effectively with zero water consumption," Nadella said during his keynote, while describing how new architectures are redefining data center water consumption. "The daily water usage over the course of an entire year is roughly equivalent to what a single restaurant would use."

While Nadella did not specify the size or type of restaurant used for comparison, the claim is striking given the scale of modern hyperscale AI facilities. Industry estimates suggest that restaurants can consume anywhere from several hundred thousand to more than a million gallons of water annually, while 100MW+ hyperscale data centers have historically required tens to hundreds of millions of gallons of water annually for cooling, depending on climate, workload, and cooling design.

Nadella's statement — not the first time Microsoft has used a restaurant comparison — is based on Microsoft's new closed-loop liquid-cooling architecture. Traditional data centers often rely on evaporative cooling systems that continuously consume water to remove heat generated by servers and networking equipment. In Microsoft’s new approach, over 90% of the facility's cooling relies on a closed-loop liquid cooling system that is filled during construction and then continuously recirculates the same water rather than constantly consuming a fresh supply. The remaining portion of the cooling system primarily relies on outside air and only uses additional water in the hottest conditions.

The system works by circulating cooled water through heat exchangers connected to the AI hardware. As the water absorbs heat, it is routed to a massive chiller plant where large cooling fans dissipate the heat before the water is recirculated back through the facility. Because the water remains in the cooling loop rather than evaporating and being discarded, ongoing water consumption is dramatically reduced. This can translate to savings of multiple billions of gallons of water across Microsoft’s data centers.

The current reality is much more subdued. The new cooling design and its water savings are currently implemented only at Microsoft's Fairwater AI data center campus in Mount Pleasant, Wisconsin. On the other hand, the company has stated that the new cooling approach will become the standard design for future AI-focused data centers. Microsoft has confirmed that multiple identical Fairwater facilities are already under construction elsewhere in the United States as part of a broader expansion of its AI infrastructure footprint. The rollout forms part of Microsoft's longer-term goal of becoming water positive by 2030, meaning the company aims to replenish more water than it consumes globally, as part of its community-first strategy.

While the new cooling design will substantially reduce water use in future facilities, critics note that Microsoft already operates a vast global network of data centers. Azure now spans more than 500 facilities across 80 regions worldwide, many of which were built before Fairwater's cooling architecture was introduced.

Microsoft has not announced a comprehensive retrofit program that would convert existing data centers to the new closed-loop design. As a result, the "restaurant-level" water footprint applies primarily to newly built facilities that use the Fairwater blueprint, rather than to Microsoft's entire global infrastructure estate.

The debate reflects a broader challenge facing the AI industry. Companies, including Microsoft and Google, are racing to build ever-larger AI clusters while facing mounting public pressure to reduce their environmental impact, as communities grow increasingly disapproving of large-scale AI infrastructure projects. Google recently announced a series of water stewardship commitments of its own, including a pledge to replenish more water than its data centers consume by 2030.

U.S. Senator Bernie Sanders (I-Vt.) has announced that he will introduce a bill that would force the largest American AI firms to hand over half of their stock to the public. The senator proposes a 50% tax on the stock of these firms, which, it seems, will be held by the public through a sovereign wealth fund. It would also mean that the government would have a direct ownership stake in these companies, allowing the people to have a say in how these companies would run.

“The foundation of AI is our collective human intelligence. Our books, songs, artwork, journalism, computer code, scientific research, videos, conversations, images, and ideas spanning generations. As Sam Altman himself acknowledged, AI models were trained on our ‘collective experience, knowledge, and learning of humanity’,” said Sanders. He also added, “Since AI is built on the collective knowledge of humanity, the wealth it generated must benefit humanity. Not just Elon Musk, Jeff Bezos, Mark Zuckerberg, Larry Ellison, and other billionaires, or the venture capitalists and Wall Street firms who see AI as the next great wealth extracting machine.”

According to the senator, the biggest AI companies themselves have started this idea. OpenAI has already proposed a public wealth fund so that the public would benefit from AI-driven economic growth, while Elon Musk himself posted on X that “Universal HIGH INCOME via checks issued by the Federal government is the best way to deal with unemployment caused by AI.”

He then compared his AI sovereign wealth fund proposal to Norway’s Government Pension Fund Global and the Alaska Permanent Fund. Both are sovereign wealth funds whose proceeds came from their nations’ oil wealth — the former is now worth US$2.3 trillion and is used to fund the European country’s extensive public welfare system, while the latter has been paying out $1,000 to $2,000 to every Alaskan citizen annually since 1980.

Sanders’ proposal plans to combine the benefits that both Alaska and Norway deliver, while simultaneously forcing AI companies to hand over 50% of their ownership to the public through the government. This might seem a radical solution, but the senator is no stranger to extremes. He has previously called for a halt to all AI data center construction, saying that such a move “will give democracy a chance to catch up.” While a federal moratorium still hasn’t been approved, more and more jurisdictions across the U.S. are approving data center bans as communities are pushing back against their development in their backyards.

SoftBank Group will commit as much as €75 billion ($87 billion) to build five gigawatts of AI data center capacity across France, with a first €45 billion ($52 billion) phase delivering 3.1GW in the northern Hauts-de-France region by 2031. The Japanese group called it its largest AI infrastructure investment in Europe to date, with state-owned utility company EDF and France's Schneider Electric named as partners. Phase one covers three sites: Dunkirk's Loon-Plage, Bosquel, and Bouchain, and Masayoshi Son and President Emmanuel Macron are set to formalize the plan on Monday at the Choose France summit.

What France offers is electricity: the country draws roughly 70% of its power from nuclear reactors run by EDF, is the world's largest net electricity exporter, and posts industrial power prices well under half the UK's. Son told the publication La Tribune du Dimanche that France being a producer and exporter of energy was "absolutely decisive" for AI infrastructure spending. EDF is contributing to the Bouchain site, a former power plant, which it will hand over for conversion.

There’s a stark contrast between this and SoftBank’s American plans, where, to feed its planned 10GW data center in Ohio, the company has to build its own energy generation infrastructure in the form of a natural gas plant that’ll cost roughly $33 billion to build and generate roughly 9.2GW. France lets it skip that step and plug into an existing low-carbon fleet, sidestepping the strained grids and rising local opposition slowing U.S. buildouts.

With Schneider Electric, SoftBank plans to build an industrial cluster at the Port of Dunkirk, built around two plants, one run by SoftBank to make enclosures and one run by Schneider to integrate data center power modules

That pulls SoftBank's automation work into prefabricated electrical kit, extending the company’s vertical integration — which already spans Arm, robotics, and its own water-based data center batteries — even further. EDF chairman Bernard Fontana said in the announcement that the Bouchain project showed France's ability to host large-scale digital infrastructure on "competitive, sovereign and low-carbon electricity."

The French investment adds to a global spending run that includes the Ohio project and more than $30 billion sunk into OpenAI for an 11% stake. SoftBank carries over $130 billion in debt and took a $40 billion bridge loan in March to fund its latest OpenAI investment. The €75 billion figure is also a ceiling; only the €45 billion first phase and its 3.1GW are firm commitments, with the remaining sites described as plans for later.

Amazon has unveiled a new data center networking architecture that it claims delivers up to 33% higher throughput while reducing network power consumption by 40%, marking what the company describes as the first large-scale deployment of a flat network based on random graph theory, Wired reports. The company revealed that it had been quietly deploying the design in its data centers since last year, confirming that it has already become the default data center network for most AWS workloads.

The architecture, called Resilient Network Graphs (RNG), replaces the hierarchical networking structures that have dominated cloud data centers for decades with a flatter, quasi-random architecture designed to move data more efficiently between servers. Amazon says the design uses 69% fewer networking devices than traditional architectures and can reduce infrastructure costs by up to 45%, potentially translating into billions of dollars in savings across its global cloud footprint.

The company first deployed RNG in a Dublin data center in 2024 before expanding the architecture into facilities in Germany and Spain. AWS says the design is now being rolled out across most newly built data centers and will form the foundation of future deployments.

Data center and AI capability conversations typically focus on processing and storage. However, networking is one of the most important aspects of capability. Every request sent from your device to a cloud application, database, AI model, or storage service depends on data moving rapidly across thousands, or even hundreds of thousands, of servers. As facilities continue to grow, ensuring information reaches the right destination quickly and efficiently has become an increasingly difficult engineering challenge.

For decades, most large-scale data centers have relied on a hierarchical networking structure known as a "fat-tree" topology. In this design, data travels up and down layers of switches and routers arranged in a tree-like hierarchy. While the approach is proven and reliable, it can create bottlenecks and strain available bandwidth. Traffic can become concentrated at specific points in the hierarchy, even when capacity is unused elsewhere in the network. The approach also requires a large number of expensive networking devices.

Researchers have long theorized that a flatter network based on random graph principles could solve many of those issues. Instead of forcing traffic through predefined layers, routers are connected in a highly distributed mesh-like arrangement that creates numerous possible paths between endpoints. In theory, this will improve resilience, increase utilization of available bandwidth, and reduce the likelihood of congestion.

The challenge was making such a network practical. Although random graph networks have been subject to academic research for more than a decade, major obstacles have prevented their successful deployment at hyperscale. Routing traffic efficiently through such a network is considerably more complicated than routing traffic through a predictable hierarchy, and physically connecting millions of fiber-optic links without creating an operational nightmare presents an entirely different challenge.

AWS says it overcame those obstacles with two key innovations: software and custom hardware. The first is a custom routing protocol called Spraypoint, which distributes traffic across large numbers of available paths rather than relying primarily on the shortest path. The second is a passive optical device called ShuffleBox, designed to organize and standardize the immense amount of cabling required to build the network at scale.

According to Amazon, the resulting architecture not only moves data roughly one-third faster than conventional designs but also requires significantly fewer switches and routers. The company says those reductions lower both capital expenditures and operating costs while simultaneously reducing power consumption.

AI models are growing larger and more capable, and so are their user bases. To sustain this growth, hyperscalers have to continue optimizing the entire technology stack. Faster chips remain crucial, but so are advances in cooling, power delivery, storage systems, optical interconnects, and networking. AWS believes RNG is one such advantage. The company rightly considers the achievement of deploying random graph theory at scale, “a breakthrough that will deliver greater reliability and performance for AWS customers, save billions of dollars in hardware, and lower CO2 emissions.”

The region of Lombardy, Italy, which has Milan as its capital, just approved a 100% tax on data center developments in rural areas and a 200% tax for those being built in agricultural and green zones. According to the Italian news outlet Il Sole 24 Ore, the law was designed to discourage hyperscalers from buying up rural tracts of land for projects like these, which Lombardy leaders say are often done “without clear timeframes and plans.”

“We cannot, in the light of these numbers, block the development of companies and employment, the race for artificial intelligence is already a fact,” Lombardy councilor Massimo Sertori told the publication. “We can, however, try to keep the phenomenon under control by avoiding excesses and the exaggerated exploitation of the territory.”

It seems that even data center developers in Europe have started gaining interest in rural land. Even though these are often less developed compared to urban and suburban areas, it also usually comes with fewer restrictions, are cheaper to build on, and might even have a faster permitting process. One Texas county noticed this and has taken steps to limit AI hyperscalers’ ambitions while the county government is studying their potential impact on rural land. On the other hand, Italian lawmakers are hoping that the law will push developers toward disused industrial zones. These areas are typically built for operations like these, and there are no additional burdens for AI data center developers who choose to build here instead.

Aside from land use, Lombardy officials also have concerns about energy use. Sertori said that data center applications have now reached up to 30GW throughout Italy, with more than half of them planned to be situated in the region. However, they should only authorize 2GB based on “real and concrete projects.” At the moment, Milan hosts 33 active data centers, with 10 more in the construction stage and 23 more applying for approvals. This makes Lombardy the region with the highest concentration of data centers in Italy.

The data center buildout needed to power the “AI revolution” is quickly becoming a hot topic in the United States. Many Americans are pushing back against these developments, with a recent survey saying that 70% do not want a data center in the vicinity of their homes. While it doesn’t seem that Italians are pushing back with the same ferocity, their authorities are still concerned that unchecked data center development would have a negative impact on the community. Opposing politicians even voiced their concerns that this move is not enough.

“In Lombardy, we have tried to fill the void of the Meloni government on industrial policies. The development of data centers is necessary, but it cannot be left to chance or to the market alone; a political vision is needed,” Democrat Mattero Piloni said. “However, this law lacks real and decisive soil protection, because neither the government nor the region have put in place stringent constraints. We will have to wait for a national law.”

Erin Brockovich, famed for her role in the landmark 1996 class-action lawsuit against Pacific Gas & Electric (PG&E) that resulted in a $333 million settlement, and the subsequent movie that bears her name, is on another crusade. This time, the environmental activist and consumer advocate is targeting data centers, launching the Brockovich AI Data Center Reporting website to compile complaints from affected residents and community members.

"The RACE to build AI infrastructures is unfolding town by town across America. In some places, data centers are welcomed. In others, they are delayed, contested or abandoned altogether,” Brockovich wrote on her website. “This MAP captures the real-world footprint of that race — revealing patterns of growth, conflict and uncertainty.” At the moment, the website lists 33 operational (built & running) AI data centers, 44 under-construction projects, and 27 proposed sites. It also shows 2,716 community-reported locations nationwide as of May 24, 2026.

There are several primary concerns against data centers, the biggest of which is energy consumption, which has been causing utility prices to skyrocket across large swaths of the United States. Water usage is also another issue that communities have been raising, with these massive-scale projects using a lot of water and catching the attention of lawmakers. There are also reports of noise pollution, with some members of the public complaining about the infrasound phenomenon often brought about by large-scale industrial operations. Brockovich also lists location risks, scalability, and e-waste as potential long-term problems that the industry must deal with in the future.

While the White House has been pushing for AI-friendly policies, the rest of the American people are increasingly growing skeptical about its development, particularly for data centers that are popping up in their backyards. It has gotten to the point that 70% of Americans oppose data center construction near their homes, a 23% increase from a survey conducted less than six months before. The pushback against these developments is gaining steam, too, with at least 69 jurisdictions passing moratoriums on data center construction to give local governments time to assess their impact.

It’s unclear what Brockovich intends to do with the data she’s gathering, but this is often one of the first steps needed for lawyers to see if a class-action lawsuit has merit. While there is no case at the moment, this website puts data center developers on notice that their actions are being publicly tracked, and that Erin Brockovich has her keen eye on them.

China’s “world's first” offshore wind-powered underwater data center (UDC) has begun operations, according to reports from Chinese media. The project, officially launched in June 2025 and completed in October 2025, hit full commercial operation last week, after successful initial trials in February.

Located off the coast of Shanghai's Lingang Special Area, the $226 million data center was built and is managed via a direct partnership between the Chinese government and HiCloud Technology (the primary private engineering contractor specialized in subsea data centers), along with state-backed telecom providers like China Telecom.

The 24 MW facility houses nearly 2,000 servers (including GPU clusters from China Telecom and LinkWise), and is expected to process artificial intelligence, big data annotation, and 5G infrastructure workloads. Unlike conventional land-based data centers that rely heavily on industrial chillers and large HVAC systems to remove waste heat, the Shanghai UDC uses the surrounding seawater as a massive passive heat sink. The servers are sealed inside pressure-resistant subsea modules deployed roughly 35 meters beneath the surface, where stable ocean temperatures continuously absorb heat generated by the computing hardware.

Cooling has become a major bottleneck for modern AI data centers, where dense GPU racks can consume hundreds of kilowatts, converting nearly all of that energy into heat. The underwater design uses surrounding seawater as a passive heat sink, sharply reducing cooling power requirements.

Chinese media reports claim the facility achieves a Power Usage Effectiveness (PUE) below 1.15, placing it among the most energy-efficient large-scale data centers in operation. Traditional enterprise data centers often operate closer to 1.5 or higher, meaning a significantly larger portion of their total electricity consumption goes toward cooling and supporting infrastructure rather than computation itself.

The project also reflects China’s broader push to integrate renewable energy directly into digital infrastructure. The underwater data center is connected to nearby offshore wind farms, allowing a substantial portion of its electricity demand to be supplied directly from renewable generation sources. As AI expansion drives explosive growth in electricity consumption worldwide, countries and hyperscalers are increasingly exploring unconventional infrastructure approaches to address both energy availability and thermal management constraints.

However, underwater data centers also introduce substantial engineering and operational challenges. Saltwater corrosion, long-term pressure sealing, subsea cable reliability, and maintenance accessibility remain major concerns. Replacing failed hardware is considerably more complex than in conventional facilities, where technicians can physically access racks within minutes. Operators therefore rely heavily on sealed modular designs, remote monitoring systems, and highly redundant infrastructure intended to minimize the need for physical intervention.

The Shanghai project follows earlier experimental efforts such as Microsoft’s Project Natick, which tested submerged data center capsules off the coasts of Scotland and California. Microsoft ultimately discontinued the program commercially, but the trials demonstrated that underwater deployments could achieve lower hardware failure rates.

Offshore-powered, ocean-cooled data center projects are continuing to emerge worldwide as AI infrastructure power and cooling demands continue to soar. Last month, we reported on a Peter Thiel-backed startup, Panthalassa, which is developing wave-powered floating data centers designed to operate far offshore using ocean water for passive cooling while drawing electricity from onboard renewable energy systems.

Commissioners of Hill County have voted 3-2 to issue a one-year moratorium on data center projects on rural land, with one member saying that the local government will take this time to study the effects of data centers on the community around it before developers can move forward. Although it is not the first local government to join the growing list of jurisdictions that have passed a data center moratorium, The Texas Tribune says that it would be the first one to enforce a temporary ban in Texas, which is known for being a data center-friendly state. The move comes after data center builders have begun to target rural areas to avoid the harsher regulations and oversight within cities.

AI hyperscalers are increasingly looking to unincorporated county land to reduce regulatory friction, allowing them to get their projects online much quicker. While they still have to go through county commissions and other authorities that work at the county level, they get to skip city-level rules and debates, which can get testy at times.

The decision was spurred by a proposed data center campus by Provident Data Centers, which is located outside of and just north of Hillsboro, a city of about 8,000 people. The 300-acre plot of land where the project is envisioned to be constructed sits in a rural area about 60 miles south of Dallas-Fort Worth International Airport. This unincorporated territory is managed directly by the county and is not part of Hillsboro or any other municipality.

However, it seems that some county officials are catching on with this pattern and are actively moving to block or at least delay these power-hungry projects. This is especially true as an increasing number of Americans are opposing the construction of these AI data centers in their neighborhood. “The data center folks have found a sweet spot in the state that has limited regulations, limited enforcement, limited code, and they’re coming faster than we can keep up with,” Hill County Commissioner Jim Holcomb told the publication. “I think it’s imperative … that we tap the brakes and we get our arms around what we’re faced with and do the research, do the studies.”

One of the biggest issues that communities have around data centers is the increased power rates caused by the power-hungry infrastructure. The U.S. power grid companies are upgrading their infrastructure to handle the increased loads that AI infrastructure demands, but these expenses are evenly distributed to all ratepayers. This meant that even residential users and small businesses are slapped with higher utility bills, with electricity costs across the U.S. rising by more than 30% since 2020. In fact, the state of Maryland complained to the Federal Energy Regulatory Commission (FERC) after PJM Connection, LLC., the grid operator for the state (and 12 others), slapped it with a $2 billion bill to be passed on to all consumers for its grid upgrade costs.

Some projects are bypassing the electricity grid problem by building their own power. One data center project in Utah plans to use this approach by running entirely off-grid using natural gas. However, residents are concerned about the potential air pollution that such an operation could bring, especially as the site has a 9GW capacity — more than twice the amount of power that the entire state needs. Some people are also complaining about the noise pollution that these sites can bring, as well as raising issues about inaudible but “felt” infrasound that is suspected of causing adverse health effects.

These issues are just some of the things that the county likely wants to review, so the temporary delay is a win for the people of Hill County. Nevertheless, it’s not without risk to the local government and its leaders. County Attorney David Holmes said that they could be sued if they pass the moratorium, telling the commission, “You’re damned if you [do] and damned if you don’t.” Furthermore, Sen. Paul Bettencourt (R.-Houston) said in a letter to Texas Attorney General Ken Paxton that counties do not have the right to pass development moratoriums and asked them to investigate Texas counties that have passed one.

Data center developers are rushing to get their projects online, especially while there’s high demand for compute and funding is readily available. However, shortages in power infrastructure have delayed or canceled half of the planned projects in the U.S. When combined with the pushback from citizens and lawmakers, it could mean that what used to take 19 days to build could now take several years.

Monitoring Analytics, the federally mandated independent watchdog that keeps an eye on the critical PJM Interconnection that distributes power around the U.S., said in a new report [PDF] that a massive 75.5% increase in power costs in the largest region of the U.S. has been directly caused by data centers, and it also blamed the regional market operator for failing to keep up with the rising demand. The price increases have been steep; wholesale electricity prices went up from $77.78 per MWh in the first quarter of 2025 to $136.53 per MWh in the same period of this year.

“The price impacts on customers have been very large and are not reversible,” the watchdog said in the report. “The price impacts will be even larger in the near term unless the issues associated with data center load are addressed in a timely manner, prior to the next BRA (base residual auction), scheduled for June 2026.” Monitoring Analytics says that PJM Interconnection is trying to rewrite the rules for the capacity market and bake in data center demands into its forecasts. The watchdog is critical of this proposal as it will raise the prices for all electricity consumers, putting an unnecessary burden on households and small businesses.

There is a possible solution to this problem: make data centers and other major consumers negotiate directly with power producers instead of mixing that demand with the BRA. This auction is where capacity is sold some three years ahead of when it’s needed, and including data center demand in the BRA will push prices up for everyone. In fact, the report says that without the AI infrastructure built out, the PJM Capacity Market would not have seen the high prices we’re experiencing now. So, by making data centers negotiate directly with power producers, this will ensure that the expanded capacity is solely shouldered by these large consumers and help keep utility bills stable for everyone else.

However, this is not within the interests of PJM Interconnection. After all, by keeping massive data center loads baked into the general capacity forecast, the power auction would result in higher prices as demand moves up, but supply stays at relatively the same level. This higher cost, in turn, would then be passed on to transmission operators and local utilities, eventually making its way into the bill of the individual consumer.

The increasing backlash against data center development, particularly its impact on electricity prices, has caught the attention of the federal government. In March of this year, President Donald Trump gathered some of the country’s biggest AI hyperscalers in the White House and made them promise that they would “pay their own way” when it comes to AI infrastructure costs. This is, in principle, what Monitoring Analytics is pushing for: have tech companies pay for their own power — both the electricity they consume, and the infrastructure needed for it.

Unfortunately, the “ratepayer protection pledge” is nothing but a promise, and it cannot force institutions like PJM Interconnection to not pass on the burden of cost to the average American unless Congress passes a federal law that forces the Federal Energy Regulatory Commission (FERC) to prevent cost-shifting.

Major Chinese optical fiber manufacturers have booked orders stretching into early 2027, as AI data center construction drives demand growth that the supply chain cannot match, according to a DigiTimes report. Hengtong and FiberHome told the publication their production lines are running at full capacity, with delivery cycles for some products extending from weeks to months.

The scale of demand from AI infrastructure dwarfs anything the fiber industry could ever have planned for. Data center fiber demand grew roughly 76% year-on-year in 2025, according to CRU data cited by various industry publications, and the segment is projected to account for 30% of total global fiber demand by 2027. In 2024, that figure was below 5%.

All this is because AI training and inference clusters require far denser interconnect fabrics than conventional cloud infrastructure. Rahul Puri, CEO of STL's Optical Networking Business, told Fierce Network in December that AI-focused data centers need approximately 36 times more fiber than traditional CPU server racks.

Unfortunately, manufacturing optical fiber preforms, the glass rods from which fiber is drawn, is a technically demanding process with high barriers to entry. New preform capacity typically takes 18 to 24 months to build, according to industry sources cited by DigiTimes, which limits near-term supply regardless of how quickly downstream cable production can scale.

Compounding the bottleneck, manufacturers have shifted production from standard G.652D fiber used in telecom networks to higher-margin G.657A fiber suited to AI data centers and drone applications. That reallocation has created secondary shortages in conventional telecom-grade fiber and contributed to broad price increases. Global fiber prices have risen roughly 70% from a 2021 trough of $3.70 to approximately $6.30 per fiber-kilometer.

Companies like Meta are responding by locking in multiyear commitments. It signed a $6 billion supply agreement with manufacturer Corning in January, which also disclosed two additional deals of similar scale with unnamed hyperscalers in its Q1 2026 earnings. Earlier this month, Nvidia invested $300 million in Corning to build three new fiber manufacturing plants in North Carolina and Texas.

North American demand growth is projected at 22% to 25% this year, while supply expansion trails at 12% to 19%, according to Rebio Group estimates. Data Center Dynamics reported that lead times have stretched to 20 weeks for large-volume buyers and up to a year for smaller purchasers.

Corning plans to expand its Hickory, North Carolina, plant under the Meta deal and will build three additional facilities under the Nvidia agreement. Still, new capacity from those projects isn’t expected to come online until 2027 or later.

A group of Pennsylvania residents spent a two-hour town hall lambasting Gov. Josh Shapiro over his support for data center projects, claiming residents' concerns are being ignored and the governor is losing his support base. According to Inside Climate News, 20 speakers expressed their frustration about the threats that these projects pose to the communities surrounding them. Some of the raised issues include increasing electricity prices, massive water consumption, and noise pollution, which are some of the effects that have been widely reported around the country.

Among them is Kelly Donia, who lives about 20 miles northeast of Philadelphia in East Whiteland Township and identifies as a registered Democrat. She’s a vocal critic of the governor, saying that he has lost her support because of his part in the data center boom. “He is losing his base,” Donia told Inside Climate. “I want him to hear this loud and freaking clear. I’m going to make it my job to make sure that man never gets elected again for any office.” Another said the issue was about "public trust and transparency," claiming that too many Americans "are finding out about these projects after decisions have been made." Jennifer Dusart of Mechanicsburg went on to say concerned residents "have been bulldozed over."

Shapiro does support data center developments within the state of Pennsylvania, but says he did not give them free rein to do anything they want. “If companies want the Commonwealth’s full support — including access to tax credits and faster permitting — they must meet strict expectations around transparency, environmental protection, and community impact,” spokesperson Rosie Lapowsky told the publication in a statement. “This is about setting a higher bar for projects, not lowering it, and ensuring development happens responsibly and in a way that benefits Pennsylvanians.”

The state government has taken concrete steps to insulate its citizens from increasing utility costs brought about by the massive infrastructure investments needed to deliver the power that data centers demand. The state’s Public Utility Commission (PUC) has released guidance on how utility providers should deal with high-demand users like data centers. From this, PECO, the electricity provider for Philadelphia and other parts of southeast Pennsylvania, said that it put agreements in place with data center providers that completely pass the burden of upgrading their high-voltage lines and other long-distance transmission infrastructure to these large customers. This ensures that smaller users do not have to pay for these upgrades through higher electricity bills.

Still, this isn’t enough for many Pennsylvanians. State Representative Jamie Walsh (R.-Luzerne County) said that one of the reasons the state has been experiencing an inflow of data center investments is because of a 2021 law that gave tax breaks to their developers. As a way to help slow down the breakneck speed of proposed data centers in the state, State Senator Katie Muth (D.-Philadelphia) said that they want to introduce a bill that will put a three-year moratorium on all projects in the state. This would be a major decision if passed into law, making Pennsylvania join the numerous other smaller jurisdictions that have already passed a temporary ban.

The tension within the state is mirrored all across the country as residents and communities push back against data center projects. Some AI hyperscalers are trying to reassure people that they will take steps so that they will have limited to no effects on the environment around these sites. However, this is too little, too late for many, as the negative effects of other projects like these have already been widely reported since last year.

In fact, many issues continue to surface, like this one data center in Fayette County, Georgia, which used up 29 million gallons of water in 15 months, resulting in low pressure for everyone else. We’ve also seen recent complaints about noise pollution, especially projects that were built in close proximity to homes and public infrastructure.

Shapiro's balancing act is trying to ensure that the state enjoys the huge boon that data centers could have on Pennsylvania’s balance sheet while ensuring that the interests of the people are still met. But as people become more aware of the risks that data centers pose to their way of life, many critics say that he’s not doing enough.

Last week, Anthropic announced that it had struck a deal with SpaceX to lease all of the latter's Colossus 1 data center, with over 220,000 GPUs and 300 megawatts of compute capacity. The deal immediately raises questions, foremost among them: why would Musk lease one of xAI’s most aggressively hyped AI assets to a direct rival? With SpaceX's IPO just around the corner, a related strategy appears to be at play, but it also turns out that the system's mixed architecture with different types of GPUs may be a key reason Musk has decided to lease the system.

Anthropic says the newly acquired capacity will primarily be used to ease long-standing usage bottlenecks across Claude’s paid ecosystem. According to the company, the additional compute will enable significantly higher Claude Code limits, the removal of peak-hour throttling for Pro and Max subscribers, and substantially increased API request limits for Claude Opus models used by developers and enterprise customers.

The seemingly unlikely partnership — a complete turnaround of Musk's earlier stance on Anthropic — also reveals Anthropic is straining under the Claude ecosystem’s compute demands. The company says it needs the entire 300 MW AI supercluster just to improve the experience of using Claude.

Anthropic appears to have hit the compute wall

The earliest signs that Anthropic was struggling to keep up with the computing demands of its growing user base were the increasingly aggressive usage limits placed across Claude’s services. Free users frequently complained about rapidly exhausting tokens — the units Claude assigns for processing tasks. However, the restrictions extended beyond the free tier. Paid Pro, Max, Team, and Enterprise users also regularly encountered message caps, peak-hour throttling, API rate limits, and strict time-based usage ceilings on Claude Code sessions, particularly during periods of heavy demand.

It was clear that Anthropic was running out of inference capacity. While training an AI model is an expensive, one-time computational undertaking, serving that model to millions of users simultaneously creates a continuous, round-the-clock demand for compute that scales directly with every new user and every new query. The apparent solution is to build more data centers, which Anthropic is apparently pursuing via massive gigawatt deals with Amazon, Google, Microsoft, and Nvidia.

However, modern hyperscale AI data centers can cost tens of billions of dollars and take years to build. Utilities are increasingly struggling to supply sufficient electricity for AI projects, while land, transformers, cooling infrastructure, and high-end GPUs themselves remain constrained. There is also growing sentiment against AI infrastructure from local communities. We recently reported that a U.S. senator got physical with a reporter after a confrontation on a data center issue.

Anthropic's compute capacity problem was immediate and urgent, but the solution was significantly long-term. If only there were a massive AI supercluster with hundreds of megawatts of compute power just sitting there. Turns out there was: SpaceXAI’s Colossus 1. Following the deal, Colossus 1’s entire computing power now belongs to Anthropic — for now.

Musk, xAI, SpaceX, and an upcoming IPO

When Musk unveiled Colossus, it was framed as one of the clearest signs that xAI intended to compete seriously with OpenAI, Anthropic, and Google at the AI frontier. The Memphis-based cluster became famous for how quickly it was assembled. Tens of thousands of Nvidia GPUs were reportedly brought online in record time, eventually scaling to over 220,000 accelerators. Musk repeatedly boasted about xAI’s future compute ambitions, including plans to expand toward million-GPU-class systems through Colossus 2.

So why does he seem to have wrapped the whole thing in a neat little bow and handed it over to Anthropic, xAI's rival? One possible answer is utilization. Reports suggest that Colossus 1 may have had more available capacity than Grok’s current user base required. However, according to a detailed report by Mirae Asset Securities — a major South Korean investment bank — the bigger utilization issue was architectural. Colossus 1 is a heterogeneous cluster, mixing roughly 150,000 H100s, 50,000 H200s, and 20,000 GB200s — three different generations of Nvidia silicon running under one roof. This was largely a byproduct of how fast xAI assembled the cluster, with different GPU generations coming online as supply allowed, rather than a deliberate design choice.

For AI training, the heterogeneous configuration creates a significant efficiency problem. Distributed training requires every GPU in the cluster to complete each computational step simultaneously before the system can advance. When the faster GB200 chips complete their work first, the entire cluster waits for the slower H100s to catch up — a well-known bottleneck known as the straggler effect. At 220,000 chips, this effect is exponential.

As a result of these issues, xAI's real-world GPU utilization reportedly sat at just 11% — meaning 89% of the cluster's theoretical computing power was going to waste. For context, Meta and Google typically operate at 40% or above.

AI GPUs are not static assets that quietly sit on shelves, gaining value over time. They depreciate rapidly, consume enormous amounts of electricity, and require expensive maintenance and cooling infrastructure. Unused GPUs are effectively burning money.

From that perspective, Anthropic may have arrived at exactly the right moment. The company had exploding demand and an urgent need for ready-made compute, while SpaceX/xAI had a gigantic, not-so-great first-generation AI cluster. For Anthropic, however, the same cluster looked quite different. The company needed compute power for Inference — running queries through an already-trained model, which does not require the tight synchronization that training workloads demand. So, what was a structural inefficiency for xAI's training workloads is a workable infrastructure for Anthropic's inference needs.

Multiple reports suggest xAI is now heavily focused on Colossus 2, a far larger next-generation cluster reportedly aimed at gigawatt-scale AI infrastructure. Unlike Colossus 1's chaotic mix of chip generations, Colossus 2 is built entirely on Nvidia's Blackwell architecture — a homogeneous cluster where every GPU is identical. In a uniform cluster, every chip completes each training step at roughly the same time, allowing GPU utilization to theoretically surpass the range in which Meta and Google currently operate. xAI can also properly optimize its software stack for a single hardware generation rather than trying to serve three simultaneously.

According to the Mirae Asset report, xAI has already moved its core training workloads entirely onto Colossus 2, effectively treating Colossus 1 as a retired first-generation asset. In other words, Colossus 1 may have transitioned from "cutting-edge frontier training weapon" into a monetizable first-generation compute asset, while Musk continues to build towards xAI’s “takeover” with Colossus 2.

Musk has long treated his companies less like isolated entities and more like interconnected pieces of a broader ecosystem. Tesla technologies appear across SpaceX projects. SpaceX infrastructure supports xAI ambitions. xAI products increasingly feed into Musk’s wider platform strategy.

The deal also hints at another possibility: Musk could be positioning SpaceX/xAI as more of an AI cloud infrastructure provider. That would not be entirely surprising. xAI has already launched Grok Business and enterprise-focused offerings featuring APIs, security controls, audit logging, and corporate integrations. This also aligns with Musk’s reported plans for broader structural changes at SpaceX and xAI ahead of the company's upcoming IPO.

Earlier this year, Musk publicly attacked Anthropic and Claude, calling the company “misanthropic and evil.” Yet this week, he claimed he approved the deal after speaking with Anthropic executives and determining that “no one set off my evil detector.”

Mirae Asset’s analysts attempted to estimate the value of the Anthropic deal, using estimated hourly lease rates for different Nvidia GPU types. The analysts projected that Colossus 1 could theoretically generate roughly $5–6 billion in annual revenue. That nearly perfectly offsets xAI's annualized net loss of approximately $6 billion as of Q1 2026, effectively pulling the company to breakeven in a single contract.

For Anthropic, the analysts applied CEO Dario Amodei's own publicly stated estimate that roughly half of all AI industry compute spending goes toward inference, and that inference compute converts to revenue at a 3x multiplier. On that basis, the $5 billion being directed toward inference capacity could generate approximately $15 billion in incremental ARR — a significant addition to Anthropic's already rapidly growing revenue base.

Stellar ambition

Another critical aspect of the announcement involved “orbital AI compute capacity” — basically, data centers in space. Granted, it does sound like science fiction marketing language. But it directly ties into a core problem both companies, alongside several other AI giants, are increasingly facing: AI infrastructure is beginning to outgrow terrestrial constraints. So when a joint announcement comes from the world's largest AI company and the company that built the world’s largest reusable rocket system and operates thousands of active satellites in orbit, you best believe we may soon have data centers floating around in space.

Despite Mirae Asset’s analysis, the factual financial details of the Colossus deal are not publicly available. However, Anthropic recently raised $30 billion in a Series G funding round, valuing the company at $380 billion. It would not be too wild a guess to say some of that cash may have gone into funding the Colossus agreement. Then again, the company said last month that its annualized revenue run rate had already surpassed $30 billion, highlighting the staggering scale at which Claude’s business is now operating.

xAI built Colossus 1 fast — too fast, it turned out. The resulting mixed GPU architecture created structural training inefficiencies that made the cluster hard to justify as a long-term platform. With Colossus 2 now operational and built properly on uniform Blackwell hardware, Colossus 1 became a first-generation asset in search of a better use.

Anthropic, with explosive demand and not enough compute, provided exactly that. The deal converts what was effectively a depreciating liability into roughly $6 billion in annual revenue — enough to bring xAI close to breakeven. For Anthropic, the same compute could unlock an estimated $15 billion in additional ARR. Both companies got what they needed, and Musk gets a compelling infrastructure story heading into a potential IPO.

A new Gallup survey just revealed that seven out of 10 Americans oppose data centers built near their homes and communities — a massive jump from the 47% that objected to these projects in late 2025. According to The Washington Post, opposition towards these projects have become so intense that many of the respondents would rather live near a nuclear power plant.

Pushback against data centers has been steadily increasing across the nation, with projects getting delayed by regulatory fights, public debates, and political revolts. Elected officials have to listen as more and more people are making noise against these projects, with 69 jurisdictions across the United States enacting moratoriums.

Data center developers who are going through the permitting process are also facing stiff resistance. Even though these projects may be supported by some officials, they’re still required to go through public hearings that are quickly becoming passionate battlegrounds for residents concerned about the negative effects of these data centers on their way of life. Even those data centers that have already received approval from various councils see their supporters resigning or quickly getting voted out and getting replaced by opponents.

Many developers argue that these projects will bring in positive cash flow and bring a net positive to their surrounding communities. However, reports of egregious electricity price hikes from last year are pushing public sentiment against these developments, especially as they coincided with the cost-of-living crisis that the U.S. is experiencing. The data center boom has pushed wholesale electricity prices by up to 267%, with households and small businesses facing increased utility costs because of it. Maryland even had to go to the Federal Energy Regulatory Commission (FERC) to complain about the $2 billion bill that PJM Interconnect is charging it for upgrading its infrastructure to support out-of-state data centers.

AI hyperscalers are trying to reduce regulatory friction by building on rural, unincorporated land instead of municipalities. This would mean that they do not have to go through city or town council approvals and avoid lengthy legal battles at the local level (although they still have to pass through county-level authorities).

The political fights for and against data centers are even turning physical as tempers flare up. One alarming incident had an Indiana politician’s home shot at 13 times, with the assailant leaving a “NO DATA CENTER” sign on their doorstep, while a sitting Utah senator slapped a reporter’s phone away from their hand while they were trying to cover the harassment against the business of said senator for supporting a proposed 9GW data center.

This survey is bad news for AI hyperscalers, as data center developers can only expect the permitting process to become harder and take a lot more time. The White House has responded to the increasing price hikes earlier this year by gathering some of the biggest AI tech companies and made them promise to “pay their own way” when it comes to developing the technology. It seems to be too little, too late, though, as this is just a statement and has no regulatory teeth to back it up.

Aside from electricity cost concerns, many people are also looking at data centers’ impact on air pollution, especially off-grid sites that run their own power, as well as their potential for abuse of water resources. Citizens are also complaining about noise pollution around these sites, with some even bringing up inaudible but “felt” infrasound, which allegedly causes adverse health effects. Unless data center developers address these issues head-on and prove that they will have a limited negative impact on the community, we can only expect public sentiment to continue going against these projects.

SoftBank has announced that it will begin manufacturing battery cells and energy storage systems at its facility in Sakai, Osaka, targeting gigawatt-hour-scale production by the fiscal year ending March 2028.

The Japanese company is partnering with two South Korean startups, Cosmos Lab and DeltaX, to produce zinc-halogen batteries that use water-based electrolytes instead of the flammable organic solvents found in lithium-ion cells. SoftBank expects the battery unit to generate more than ¥100 billion ($637 million) in annual revenue by fiscal 2030.

The Sakai site is the former Sharp LCD panel factory that SoftBank acquired for ¥100 billion ($676 million) in March last year. SoftBank is already converting the 440,000-square-meter complex into an AI data center with an initial capacity of around 150 megawatts, expandable to over 400 megawatts. Battery manufacturing will be co-located on the same campus alongside solar panel manufacturing and an AI hardware plant, creating what SoftBank calls its "GX Factory" for energy infrastructure and "AX Factory" for AI compute.

Zinc-halogen batteries have the potential to solve the thermal runaway risk associated with large lithium-ion installations, particularly when they’re sited inside or adjacent to server halls. SoftBank's zinc-halogen cells use pure water as their electrolyte, which the company said eliminates ignition risk, while the use of zinc and halogen compounds will reduce dependence on Chinese-controlled supply chains for lithium and cobalt.

While zinc-based batteries have shorter lifespans than Li-ion equivalents, Cosmos Lab, which developed the zinc-halogen cell technology, is working to address dendrite buildup on electrodes. DeltaX, meanwhile, will contribute containerized storage system design, with its Cell to Pack technology achieving 5.37 MWh in a standard 20-foot container.

The battery business extends a pattern of vertical integration that SoftBank has been building across AI infrastructure over the past 18 months. The company also controls Arm, the chip IP firm whose architecture underpins the majority of AI accelerators, and it spent $4 billion to acquire DigitalBridge for data center development capacity.

It’s also planning a 10-gigawatt data center complex in Ohio that would require its own $33 billion natural gas power plant, and, last month, SoftBank announced a new robotics unit designed to automate data center construction.

A $1 billion data center that Microsoft and Abu Dhabi-based AI firm G42 planned to build in Kenya has stalled after the Kenyan government failed to meet Microsoft's demand for guaranteed annual capacity payments, Bloomberg reported Sunday. Kenyan President William Ruto put the scale of the project's power requirements into clear terms at a recent state event in Nairobi, saying the country would need to "switch off half the country" to keep the facility running.

The project, announced in May 2024 during Ruto’s visit to Washington, was supposed to bring a geothermal-powered data center to the Olkaria region in Kenya's Rift Valley. G42 was to lead construction, with the facility running Microsoft Azure in a new East Africa cloud region. The first phase targeted 100 megawatts of capacity and was expected to be operational by this year, with a long-term goal of scaling to 1 gigawatt.

President Ruto isn’t exaggerating about shutting off half the country’s power. Kenya’s total installed electricity capacity sits between 3,000 and 3,200 megawatts, and peak demand reached a record 2,444 megawatts in January, according to data from KenGen, the country’s government-owned electricity producer.

The full 1 gigawatt build would therefore have consumed roughly a third of the country’s total capacity, and even the first 100 megawatts would have required a significant share of the Olkaria geothermal complex's output, which currently generates around 950MW across all its plants.

John Tanui, principal secretary at Kenya's Ministry of Information, told Bloomberg that the project hasn’t been withdrawn and that talks are continuing, adding that the “scale of the data center they [Microsoft] wanted to do still requires some structuring.” A separate 60-megawatt project with local developer EcoCloud is also still under discussion.

Kenya’s Microsoft campus was set to be the first facility that Microsoft and G42 built together after Microsoft invested $1.5 billion in G42 back in 2024. That deal followed G42's agreement to divest from Chinese holdings and strip Huawei equipment from its systems under pressure from Washington. Microsoft President Brad Smith joined G42's board as part of the arrangement and described the Kenya project at the time as the “single biggest step forward” for digital technology in the country's history.

Meanwhile, Huawei is expanding its presence in Kenya, having launched a new fiber broadband service with Safaricom, Kenya’s largest telecom operator, last week. Africa currently hosts roughly 1% of the world's data center capacity.

Microsoft is spending $190 billion on capex in 2026, and the company adds approximately 1 gigawatt of data center capacity every three months globally. But power constraints are proving to be a universal bottleneck: nearly half of planned U.S. data center builds this year have been delayed or canceled due to shortages of electrical infrastructure.

Developers are showing increased interest in unincorporated county land for putting up data centers, reducing regulatory friction, and allowing construction to proceed at a much quicker pace. According to a SemiAnalysis post on X, since these parcels of land are outside city limits, they do not have to go through city or town council approvals, rezoning votes, and land-use reviews. Many projects are experiencing delays because of pushback from the communities surrounding potential projects, with many jurisdictions approving moratoriums on data center construction, some of them permanent bans.

Because of this, AI hyperscalers are looking for regions that offer speedier permit approvals, and they apparently see this in the rural areas surrounding towns and cities. Because these are administered directly by the county, developers remove one regulatory layer from the many that they must go through to put up power-hungry infrastructure. Many town and city councils have been becoming increasingly hostile towards these projects, with the community pressuring boards to disapprove applications and even replace council members who have previously said yes to projects.

Building in areas further from population centers likely means that developers will have to spend more on infrastructure to connect their projects to the power grid and water supply. But it seems they consider this a worthwhile expense if they can get their approvals at a much faster rate and get their projects online much sooner.

We’ve seen this move in some high-profile data center projects recently: Utah approved a 9GW data center in the unincorporated land in Box Elder County, while Meta is building a 7GW data center in rural Northern Louisiana with its own natural gas power plants. One Kentucky farming family even received a $26 million offer for 600 acres of their land — seven times higher than the average land value in the area — although they declined it and said that they would rather “stay and hold and feed a nation.”

Another advantage of putting these data centers in rural areas is that it will reduce their impact on the surrounding community, especially as the larger parcels of land mean that their operations will impact fewer people. This is especially important as these sites could potentially generate a lot of noise pollution, while off-grid operations can increase air pollution in the area because of their natural gas turbines.

Still, that does not mean that these companies can have free rein and do whatever they want without considering their impact on the people in the area. These companies still must go through county commissions, water authorities, and planning boards before they can get the go-ahead signal to build their projects, shifting political battles from the town and city to the county level. We’ve seen this with the Utah data center project, where a state senator got into a physical altercation with a reporter who was covering the community backlash of his yes vote on his business.

Data center developers are in a race to get their projects up and running, especially as more people are realizing the negative effects of these data centers, especially when it comes to electricity costs and power quality. Because of this, 47% of Americans now oppose the construction of data centers in their neighborhoods. When paired with permitting delays, lack of power hardware, and the years that it takes to connect big sites to infrastructure, many AI hyperscalers are seemingly desperate to get their projects off the ground and running, while investors are pouring a lot of money into projects like these.

Fayette County, Georgia, was telling residents to stop watering their lawns to conserve water. The request came after homeowners in a nearby subdivision reported unusually low water pressure, but when the county investigated, it found the cause: a data center campus 20 miles south of Atlanta had been drawing roughly 29 million gallons through two water connections the county didn't know existed, Politico reported Saturday.

Quality Technology Services (QTS), the Blackstone-owned developer behind the 615-acre Fayetteville campus, owed $147,474 in retroactive charges for the unmetered consumption, but the county didn’t fine the company.

The QTS Fayetteville (“Project Excalibur”) campus is among the largest data center developments in the country, currently comprising 13 buildings totaling approximately 6.2 million square feet, with plans for up to 16 buildings at full buildout. QTS has said it expects to invest up to $1 billion in the project, which began construction in 2023 and isn’t expected to be completed for another three to five years.

QTS told Politico the 29 million gallons were consumed during temporary construction activities, including concrete work, dust control, and site preparation. The company markets a "closed-loop" cooling system for its data centers, which recirculates the same water rather than drawing from the municipal supply. Once operational, QTS said its facilities would only require water for domestic needs like bathrooms and kitchens.

However, the discrepancy between QTS’s stated and actual water usage remained undetected for months, with Politico reporting that the county’s water system director, Vanessa Tigert, attributed the oversight to a procedural error during the county's transition to a cloud-based metering system.

Tigert told Politico that her department has a single employee handling both inspections and plan reviews, saying, “... we don’t have enough staff. We can’t keep staff.” QTS and the county disagreed on how long the water went unmetered, withTigert estimating about four months and QTS saying 9 to 15 months. Despite the unauthorized connections, Fayette County opted not to fine the company. "They're our largest customer, and we have to be partners," Tigert said. "It's called customer service."

The incident came to light last week after a Fayette County resident obtained the utility's May 2025 letter to QTS through a public records request. Fayetteville had already moved to restrict data center growth before this, with the city council banning new data centers in every zoning district earlier this year, adopting Ordinance 26-0-12. A separate proposal from developer Crow Holdings was denied by the city's planning commission in January, and the company withdrew its appeal in March.

Georgia's Public Service Commission also froze Georgia Power's base rates through 2028, specifically to prevent data centers from shifting electricity costs to residential customers. The state is currently experiencing moderate to severe drought, and Gov. Brian Kemp declared a state of emergency last month over wildfires. Georgia hosts more than 200 data center facilities.

The QTS campus is projected to generate $150 million to $200 million annually in property tax revenue, according to the city. Fayetteville is one of at least 50 cities across the U.S. that currently have active bans on new data center construction, with four adopting permanent prohibitions, according to the U.S. Data Center Moratorium Tracker.

Data center projects have faced resistance from residents and communities over their impact on power prices, but another complaint is being raised more frequently — noise pollution. One form of sound pollution is called infrasound, which is inaudible to humans but can be felt, and some claim it causes headaches, insomnia, nausea, and anxiety. Then there's the normal garden-variety sound pollution. The Environmental and Energy Study Institute (EESI), a non-profit organization, said that high- and low-frequency sounds emitted by these industrial sites can be heard and felt for hundreds of feet in surrounding areas, with noise levels reaching as high as 96dB for 24 hours a day and seven days a week.

Infrasound is another complaint that researchers are studying. Heatmap Plus reports that this is the phenomenon of frequencies so low they’re inaudible to humans. Nevertheless, some people can feel it, and there have been claims linking them to various negative health effects such as headaches, insomnia, nausea, and anxiety. Infrasound and its effects need further study, but it’s one of the issues local governments have been raising as they place a moratorium on data center projects. If you expand the tweet below, the embedded video includes a recording of the sounds emanating from an AI data center.

Normal noise pollution remains an issue, and communities living near off-grid data centers that generate their own power have it the worst. These sites generate their own power, typically using natural-gas-powered turbines — essentially jet engines bolted to the floor and used to turn generators that produce electricity. Aside from pollution concerns, such as those raised by residents around Elon Musk’s Colossus Supercomputer, which used over 30 mobile gas turbines for power, these turbines can be as loud as a passenger jet, making the site sound as loud as an airport. What’s worse is that, unlike backup generators, which only operate occasionally, these machines run continuously, meaning nearby communities will lose the peace of the neighborhood as long as these data centers operate.

A single modern AI GPU is estimated to consume up to 3.7 MWh of power annually, and with each server rack containing multiple AI GPUs, that means that it consumes a lot of electricity just to keep it running. This also means they generate a lot of heat — so these systems must be cooled to operate efficiently. Cooling accounts for nearly 40% of data center power use, and it’s also another major source of noise pollution, especially if a site relies on air-cooling. Although a single fan might not make much noise, the scale at which they’re deployed is starting to have a noticeable impact on the sound level they generate.

The noise from industrial-grade backup generators compounds this. On-grid data centers require constant power levels to operate at 100%; that’s why most, if not all, of these projects have on-site diesel backup generators. If you have a small diesel generator at home, you know how much noise it makes just to deliver a small amount of electricity. Now, scale that up to data center levels, and you’re looking at multiple generators with massive outputs. The noise output from these machines can reach 105dB, louder than a snowmobile and as intense as a car horn at full tilt.

These are meant to run when there’s limited to no power supply, but that is not limited to blackouts. Electricity providers can also reduce the amount of power they deliver to a data center during peak hours, meaning the site’s backup power supply will have to take over to ensure that it operates at full capacity even with less power from the utility company. Aside from that, it must also be tested for up to 50 hours annually, meaning the community surrounding the data center can expect these noises to come on at any time.

The United States does not lack flat, open land away from population centers on which to build data centers. However, AI hyperscalers prefer to locate their campuses near existing infrastructure so they don’t have to spend massive amounts of time and resources building everything from scratch. A few data centers are being built on former industrial sites, like shuttered factories and abandoned paper mills, but there are not enough of these around for the number of projects being proposed and built. As the negative effects of building these sites too close to population centers are slowly being revealed, we expect opposition to these projects to keep increasing.

The Maryland Office of People’s Counsel (OPC), a state agency that represents its utility consumers, filed a complaint before the Federal Energy Regulatory Commission (FERC) regarding PJM Interconnection, LLC’s plans to charge it $2 billion of the $22 billion it spent to upgrade its grid to accommodate increasing demand from data centers. According to the OPC’s press release, this $2 billion bill will cost the state’s consumers an extra $1.6 billion in the next ten years alone — that means an extra $823 million for residential (approx. $345 per customer), $146 million for commercial (approx. $673 per customer), and $629 million for industrial customers (approx. $15,074 per customer).

“Without FERC action, Maryland customers face paying billions for transmission infrastructure that PJM is advancing to benefit data centers,” said Maryland People’s Counsel David S. Lapp. “PJM’s cost allocation rules are broken. Maryland customers have neither caused the need for these billions in new transmission projects nor will they meaningfully benefit from them.”

PJM Interconnection, LLC is the United States’ largest electricity transmission company, and covers 13 states plus Washington, D.C. This includes Delaware, Illinois, Indiana, Kentucky, Maryland, Michigan, New Jersey, North Carolina, Ohio, Pennsylvania, Tennessee, Virginia, and West Virginia, covering about 65 million people, or about 20% of the entire U.S. population. Some of these states, including Maryland, host a large number of data centers, so the firm needs to upgrade its infrastructure to meet projected demand from these power-hungry AI systems.

However, the state argues that PJM’s methodology for allocating costs places an unfair burden on Maryland ratepayers. The OPC argues that the forecasted growth for the state is nowhere near the figures for other states, including Virginia, Ohio, Pennsylvania, and Illinois, which play host to many more data center projects. If the electricity transmission firm keeps its current computation, it would mean that Maryland customers are subsidizing the upgrade costs for projects that do not directly benefit the state.

Maryland says that these infrastructure costs should be charged directly to the areas where they’re being constructed, or, as President Donald Trump made tech companies promise with the "ratepayer protection pledge," the companies themselves should be billed directly for these grid upgrades. The OPC said there is “extreme uncertainty” regarding load growth driven by data center demand, and that utility providers tend to benefit from these upgrades even if the demand never materializes. This is especially true because the costs of these investments are borne by existing utility customers, not by the data centers themselves, if the data centers do not follow the “ratepayer protection pledge.”

This is one of the biggest reasons why many AI hyperscalers are facing pushback from the communities where they intend to place their data centers. At the moment, around 69 jurisdictions have passed some sort of moratorium on projects like these, and a survey has shown that nearly half of Americans do not want a data center in their neighborhood. Debates around these projects are passionate, with a few cases turning violent and even resulting in shootings (thankfully, without any casualties), especially as many feel that the construction of these power-hungry assets is threatening their lifestyles and quality of life.

The number of bans being enacted against data centers is increasing across the U.S., with one tracker listing 14 new bans from March to April. According to the U.S. Data Center Moratorium Tracker, there are currently 50 active bans across different jurisdictions, with an addition of four local governments enacting a permanent ban in their area of responsibility. There are also three proposed bans, plus several more in various stages, including those in the process of creating a new ban, exploring the possibility of a ban, and some with expired bans.

Many AI hyperscalers in the U.S. are rushing to build data centers across the nation, especially as they rush to become the dominant force in AI. However, this unbridled investment in data centers is resulting in shortages of various resources. We’re currently in the midst of a massive memory and storage chip shortage, and we soon might see a CPU shortage, especially as AI inference workloads increase. However, the construction of AI data centers has a direct impact on the communities that surround them, especially when it comes to increased electricity costs and noise and air pollution.

Wholesale electricity prices have skyrocketed by up to 267% in the past five years as utility providers are forced to upgrade their infrastructure to handle the increased demand from data centers. However, the upgrade costs are being passed on equally to the data centers and the average consumer, resulting in higher utility bills for everyone. It has gotten to the point that President Donald Trump met with the biggest AI tech companies in the White House and made them promise to “pay their own way” with the “ratepayer protection pledge.”

While this move will theoretically reduce the burden on the average American, at least in electricity costs, nearly half the nation is against having a data center near their home. The debates around AI infrastructure have become heated, especially as residents are concerned about their impact on both the environment and their wallets. One incident had an Indiana politician’s home shot at by an unknown assailant who left a “NO DATA CENTERS” note at their doorstep. It also saw town council members who said yes to these projects resigning or being ousted en masse as community members revolt against this threat to their lifestyles.

The tracker shows how quickly moratoriums are popping up across smaller jurisdictions across the nation. The website only listed eight moratoriums as of May 2025, but just one year later, we now have a total of 78. This is going to be a problem for many AI hyperscalers, especially as they grapple with delays caused by hardware shortages around power infrastructure and more. And with investors spending billions (if not trillions) of dollars on the promise that this technology will someday change the world and make lots of money, these delays might spook them and cause funding that many AI startups rely on to dry up.

The Middle East has long been keen on becoming a data center hub: as early as 2017, the United Arab Emirates (UAE) launched an AI strategy that was designed to place it as a global leader in the space by the start of the next decade. It quickly showed how it wanted to do that by setting up G42 a year later to corral its cloud computing capabilities.

Qatar followed with its own national AI strategy in 2019, and Saudi Arabia did the same in 2020. All have thrown significant investment into their projects, which has in turn attracted global investment, which is also eager to take advantage of the region’s cheap energy costs and significant sovereign wealth.

Saudi Arabia and the UAE are seen as the third and fourth most attractive places to develop data centers, according to Adrian Cox, managing director and thematic strategist at Deutsche Bank Research, in an April note. They sit only behind Virginia and Texas.

The largest projects are becoming pieces of national infrastructure, requiring vast amounts of electricity, cooling capacity, fiber connectivity, and political certainty. That’s why the Gulf looked so attractive to many. In parts of Europe and the United States, data center developers are running into grid constraints, permitting delays, local opposition, and power bottlenecks. In the Gulf, by contrast, governments can work on energy policy, land allocation, planning permission, and sovereign capital with a single national strategy.

All that combined makes the region unusually well-suited to the industrial scale of AI buildout. Training and running frontier models requires dense clusters of specialized chips, which in turn require dependable power and cooling. For hyperscalers, the appeal is obvious: build where the state wants you, where capital is available, and where energy supply is less constrained than in many traditional data center hubs.

“The Middle East was a prime candidate for the expansion of data centers before the conflict given readily available supply of power, available capital for development, domestic regulatory push, and strong political ties to the US,” said Mayank Maheshwari, an equity analyst at Morgan Stanley. Big money projects were announced for the region, including the Stargate project for the Middle East, among others.

From boom to bust?

But that belief has been shaken by the disruption within the region after Israel and the United States launched their first attempt to decapitate the Iranian regime.

Iran’s Islamic Revolutionary Guard hit a number of Amazon Web Services data centers in the UAE and Bahrain with drones and missiles as part of its ongoing war against the United States and Israel. Alongside that, it has posted videos online threatening to strike the planned Stargate project on the outskirts of Abu Dhabi as payback for the United States’ involvement in the war. It all adds up to a messy, dangerous time to be operating in the space in the region. In early May, Amazon's Middle East data centers were damaged by an Iranian drone.

The whole selling point of data centers is certainty. Cloud contracts are built on promises about uptime, redundancy, and service-level agreements. AI infrastructure adds another layer of pressure because companies are building compute clusters that may be booked months in advance, paid for through long-term contracts, and integrated into the internal systems of major companies.

The question for customers is not just whether a facility can survive a strike — so far, they largely have, once restored. It’s whether customers are comfortable putting critical workloads in a region where geopolitical escalation can suddenly become an operational variable. For all the talk of sovereign AI and national compute strategies, the basic commercial promise of a data center is simple: it has to be there when you need it.

The demand for data centers in the region hasn’t gone anywhere — but those scoping out projects have. They’re eyeing up locations further eastward that are in less close proximity to an active warzone.

“We see Asia's AI data center expansion getting even stronger in the coming years, especially in Southeast Asia, Japan, and Australia as hyperscalers could divert projects from the Middle East towards Asia,” said Maheshwari.

Too much panic

Not everyone is so convinced, though. The recent instability in the Middle East hasn’t fundamentally changed the investment calculus for data centers, said Mark Whyte, global head of built environment and infrastructure at Control Risks, in an interview with Tom’s Hardware Premium.

“There has been direct impact on, I think at least one data center that has been hit by a drone, but overall, we see no slowdown in activity,” he said. “If anything, the volumes of work are only up.”

That rosy outlook isn’t echoed by everyone. “Whether or not the current ceasefire holds, it has threatened the region’s economy, supply lines and facilities, and could yet squeeze investment not only in the region but also abroad,” reckoned Cox.

Whyte doesn’t dispute the potential for foreign backers to have second thoughts — but said that wouldn’t necessarily have a massive impact on the broader direction of travel. “It may well have an impact on some of the external investment, but I wouldn't see that as being a long-term impact,” he said.

Even if projects continue, the price of building them may change, though. The biggest, most strategically important projects may still go ahead because they are backed by governments with long-term horizons and deep pockets. But marginal projects that rely on external debt, cautious institutional investors, or multinational customers that have other options could become harder to justify.

Keep calm and carry on

Part of the reason that work is continuing on projects in the Middle East is the calculus by those within the region that this is a short-term instability and that AI — and the need to power it using data centers — is so consequential that they are going full steam ahead regardless. “Countries like Saudi Arabia are looking well ahead of this, rather than reacting in a tactical way,” said Whyte.

That doesn’t mean they’re being naïve, though: drones, bombs, and missiles are flying, and that’s indubitable. As a result, protecting those data centers is all important. “From a risk and resilience perspective, I think you have to look at the threats and risks to data center networks as a military planner would,” said Whyte.

And for those who have been engaged in the idea of being central to the global data center sector for close to a decade now, there’s little reason to back off much.

Saudi Arabia is not treating the conflict as a reason to back away, Whyte argued, but as a risk to manage while it pursues a longer-term strategic goal. “The Saudis in particular, see themselves as a global powerhouse for this type of thing going forward, and how they're trying to position themselves to the future,” he said. “It may well have an impact on some of the external investment, but I wouldn't see that as being a long-term impact.”

Utah State Senator Jerry Stevenson (R.-Layton) got into a heated confrontation with a reporter from ABC4 who was covering the reported harassment of his business, which soon turned physical after he slapped the reporter's phone out of his hand.

Stevenson is one of the lawmakers on the state’s Military Installation Development Authority (MIDA) who approved the 9-GW data center that’s slated to generate and use more than twice the power that the entire state consumes. However, there is swift backlash against his vote, as there have been calls on social media to boycott his business, J&J Nursery and Garden Center, and ABC4 reports that some of its employees have reportedly been harassed by disgruntled community members.

ABC4 Reporter and Anchor Bayan Wang and his photographer were doing a story about the nursery and the threats they were receiving because of the decision of one of its owners. While they were at the business’s parking lot, a man was allegedly agitated and screaming at the photographer. Wang approached the two, and he started recording their interaction as he attempted to explain that they were covering the harassment of the business and its people. However, the man, later identified as Stevenson, approached Wang and slapped his phone out of his hand.

A witness who saw the interaction called the authorities, resulting in the filing of a police report. Once the situation had cooled down, law enforcement officers told Wang that the Senator was apologizing for the incident. However, the crew was also served with a trespassing notice, warning them that they are prohibited from going on the property or premises of the business for a year.

Senator Stevenson has a lot of power in the state of Utah as he sits across multiple boards and even chairs Utah’s Executive Appropriations Committee, the legislative body in charge of the state’s budget and funding. Aside from his membership on the MIDA board, he also sits on the boards of Point of the Mountain State Land Authority and the Utah Inland Port Authority.

Data centers are facing a lot of backlash from the communities surrounding these developments, and many state, city, and town council members are increasingly facing hostility for agreeing to these developments. One of the most alarming incidents is the shooting of the home of an Indiana politician because of their support for a data center project. Aside from this, more than half of the original town council in Archbald, Pennsylvania, have resigned due to community pressure, with some of them fearing for their lives, citing the Indiana incident.

The White House is pushing for more AI development as it races to achieve technological superiority. However, many Americans are now pushing back against this untethered development, especially as data centers’ insatiable power consumption is causing electricity rates to skyrocket. Even though the 9GW Utah data center will not be connected to the grid, many residents are still wary of the potential environmental impact of the site, as it will basically put up a natural gas power plant to power itself.

Even though the project is moving forward, MIDA is said to be planning more town halls in the near future. Furthermore, construction isn’t expected to start any time soon, as its developer is said to still be raising money for the massive data center campus.

After a developer sued a township, OpenAI and Oracle's massive new Stargate data center, which will consume 1.4 Gigawatts of electricity, is now underway despite a township's vote to reject the new facility, sparking backlash. As a result, at least 19 Michigan municipalities have enacted moratoriums on new data center development since a $16 billion Stargate facility for Oracle and OpenAI was pushed through in Saline Township over near-unanimous local opposition last year, according to Bridge Michigan. The backlash now covers county resolutions, bipartisan state legislation, and a regional water authority refusing to serve proposed facilities.

The Saline project advanced despite a 4-1 vote by the township board to reject Related Digital's rezoning request in September. The developer sued within two days, alleging exclusionary zoning, and the township settled within weeks. Residents secured roughly $14 million in community benefits, including funding for the local fire department, farmland preservation, and environmental restrictions. “I think the plan was to move as fast as possible—so by the time anyone challenged it, they could say it was too far along to stop,” said one resident speaking to Forbes recently.

Michigan has since become a focal point for data center opposition. At least 19 municipalities across the state had enacted moratoriums on new data center development as of early February. Washtenaw County commissioners passed a resolution in March supporting local moratoriums, and bipartisan state legislation proposed a one-year statewide pause, though both Governor Whitmer and House Speaker Matt Hall have opposed that measure.

Related Digital and Blackstone announced last month that they secured financing for the Saline campus, purpose-built for Oracle as part of the Stargate AI infrastructure project with OpenAI. Its price tag has more than doubled from the $7 billion figure cited when it was first announced late last year, though there has been no explanation for the jump. Michigan Governor Gretchen Whitmer has called the project the largest single investment in state history.

The campus consists of three single-story data center buildings delivering more than a gigawatt of compute, with DTE Energy supplying roughly 1.4 GW of electricity using existing grid resources supplemented by battery storage funded entirely by Oracle. Related Digital says the arrangement will generate $300 million in savings for existing DTE ratepayers by spreading fixed grid maintenance costs across a larger customer base.

The facility will use a closed-loop cooling system rather than evaporative cooling, and Related Digital says ongoing water consumption will be comparable to a standard office building. The developer also claims that 750 of the roughly 1,000-acre site will be preserved as open space, farmland, and wetlands.

The Michigan Public Service Commission conditionally approved DTE's special contracts to service the facility in December, but the Michigan attorney general's office has appealed that decision. Construction began in November, and Related Digital says the project is on schedule for delivery to Oracle.

In April, the Ypsilanti Community Utilities Authority approved a 12-month moratorium on supplying water to data centers, blocking service to facilities including a proposed University of Michigan and Los Alamos National Laboratory supercomputing center. The City of Saline, which was sued by Related Digital in October, enacted its own 12-month moratorium in January.

Several additional AI data center projects are in the pipeline across southeast Michigan. Anthropic is the intended end user of a proposed hyperscale facility in Lyon Township, while Google is evaluating a one-gigawatt campus in Van Buren Township near Detroit Metropolitan Airport. Developers have identified at least 16 potential data center sites across 10 Michigan counties, Fortune reported.

According to a new survey conducted by Ipsos at the end of last year, almost half of all queried Americans said they would oppose an AI data center being built near their community. This level of opposition is higher than that given to the creation of multi-apartment buildings, new apartment complexes, or mixed-use developments.

This survey appears to highlight the growing opposition to data center construction in America. Around half of all previously announced data center projects have been delayed or cancelled entirely. Often this was for financial or component supply issues — such as Chinese power transformer shortages — but growing opposition from local lawmakers and communities about their impact on water and air quality, and electricity prices, has also been a factor.

As resentment toward AI data centers appears to be on the rise, this problem for hyperscalers may only worsen as larger swathes of the American public oppose the creation of new “AI factories.”

Millennials are the least concerned

In the survey, 4,000 U.S. residents were polled about their feelings towards the construction of new AI data centers near their communities. It was designed to gauge their fears around local environmental and service disruption, as well as the broader case of AI displacing human workers and upending social institutions.

In total, 47% of respondents said they opposed the construction of new AI data centers in their neighborhood, with just 38% saying they supported it. That support was spread differently throughout various age ranges, however.

Of those questioned, 50% of Millennial age respondents said they either somewhat or strongly supported the creation of new AI data centers in their neighborhood or local area. This was closely followed by 48% of Gen Z respondents. There was a large drop off after that, with only 38% of Gen X saying they supported their creation. Baby boomers were the least enthusiastic, with just 22% claiming they felt the same.

In what is perhaps an example of the current U.S. administration’s influence and its entanglement with top tech firms, 49% of surveyed Republicans claimed they would support new data center creation in their local area. This stood in stark contrast to just 36% of Democrats. A causal trend could also be drawn from the fact that Republican voting states and counties tend to be more rural, with less economic activity. Data center projects do require construction, and there is the potential for local job creation.

When it came to homeowners and renters, surprisingly, it was the homeowners who were more likely to support it, with 39% versus 36% of renters claiming they either somewhat or strongly supported new data center development in their neighborhood.

Turning distrust into action

Although less than half of respondents in this survey offered strong opposition to data centers, the opposition they actually throw up to the construction of these new facilities is growing and having a serious effect.

City councils that back data center projects are being voted out, other city councils are putting moratoriums on data center construction, and instances of more extreme violence towards AI companies and their employees are becoming more common.

Although this latest survey does show that there is some support for data center creation, the fact that the opposition is in the majority suggests that any data center projects that haven’t already been delayed or postponed are likely to face increasingly terse pushback that could derail their eventual development entirely.

Considering this survey was from November 2025, too, there has been further evidence of pushback against hyperscalers in recent months.

Who benefits?

This latest survey was commissioned by real estate agent services company, Redfin, which highlighted how it had heard from frustrated homeowners about data center construction in their local areas. Citing one realtor operating in Prince George’s County, MD, they said there were concerns from residents that county officials were trading long-term community quality of life for projects that don’t directly benefit the people who live there.

That’s a core component of many people’s misgivings with AI in general. Executives aren’t increasing their returns because of it, and companies are finding it hasn’t boosted productivity much either. It’s also becoming ever more expensive to run. Although there are outliers, the companies appearing to benefit the most from AI are the companies developing it, although they aren’t making anything close to a profit, aside from the chipmaking industry itself.

Even the hyperscalers like Oracle, which have received hundreds of billions of dollars worth of compute orders since the large-scale AI buildout began in 2025, are heavily reliant on AI developers like OpenAI paying their bills. Considering OpenAI specifically is struggling to make the kind of money that would allow it to make good on those orders, the list of beneficiaries of new data center developments could be small.

The AI industry’s massive push to build data centers and deliver more compute has found another unsuspecting victim — home construction contractors. According to The Texas Tribune, data centers are hiring electricians en masse, resulting in a shortage of workers for housing projects. The Lone Star State is currently in a two-pronged boom: it’s adding more residents, with more than 2.6 million people migrating into the region from 2020, and it’s also playing host to hundreds of new data center projects. This means that a lot of construction projects are competing for specialized electrical workers from the same pool of about 71,000 electricians.

These specialists are required for every construction project — both homes and data centers — as electricity can be dangerous if not set up properly. Unfortunately, AI hyperscalers have deeper pockets than the typical contractor, meaning they can pay electricians much more. For example, Scotty Wristen, who owns WE Electric in Abilene, Texas, (the same site as the first Stargate data center), told the publication that he can only afford to pay his workers $20 an hour. By comparison, data centers are able to offer up to $35 an hour, plus over time and additional benefits.

The International Brotherhood of Electrical Workers told The Texas Tribune that, out of a data center’s entire construction budget, nearly 45% to 70% are earmarked for electrical work. This is because of the massive amounts of electricity that these sites consume. The megawatts and gigawatts that tech executives easily throw around will have to be taken from the grid and distributed throughout the various buildings, facilities, floors, rooms, and individual servers in a single data center campus, and it takes professional electricians with years of experience to do that safely and efficiently.

No one can blame the individual for finding better wages and gaining a more secure future for themselves and their families. But it’s also putting a strain on the timeline of other construction projects, which are now struggling to find workers to fill the gap. This, in turn, causes delays in the turnover of housing projects, with some contractors saying that it is now taking up to two months longer to finish a single structure.

Shortages like this could also lead to delays in data center projects. An analytics group said that 40% of AI data construction sites have possibly been delayed, despite many hyperscalers denying this. Aside from labor shortages, the build out is also facing constraints in the power supply chain, with half of planned projects affected. This massive push to build as many data centers as possible is just pushing available resources — in chips, energy supply, labor, and everything else related to building and running them — to the brink. However, hyperscalers do not want to stop pouring money into it, and so AI is now an arms race funded by very deep pockets.

Now, aside from AI competing against contractors, time is also taking a toll on the workforce. It’s estimated that 20,000 electricians across the country retire annually, accounting for about one in three workers between 50 and 70 years old. While the state has on-going training to bring more professionals into the field, it takes years of apprenticeship and experience before they can be licensed. Texas took steps to alleviate this by easing the licensing requirements for out-of-state practitioners from Iowa, Alabama, and Arkansas late last year, but it’s still too early to see if this will influence the current shortage of workers.