IBM CEO warns that ongoing trillion-dollar AI data center buildout is unsustainable — says there is 'no way' that infrastructure costs can turn a profit
Krishna’s cost model challenges the economics behind multi-gigawatt AI campuses.
IBM CEO Arvind Krishna used an appearance on The Verge’s Decoder podcast to question whether the capital spending now underway in pursuit of AGI can ever pay for itself. Krishna said today’s figures for constructing and populating large AI data centers place the industry on a trajectory where roughly $8 trillion of cumulative commitments would require around $800 billion of annual profit simply to service the cost of capital.
The claim was tied directly to assumptions about current hardware, its depreciation, and energy, rather than any solid long-term forecasts, but it comes at a time when we’ve seen several companies one-upping one another with unprecedented, multi-year infrastructure projects.
Krishna estimated that filling a one-gigawatt AI facility with compute hardware requires around $80 billion. The issue is that deployments of this scale are moving from the drawing board and into practical planning stages, with leading AI companies proposing deployments with tens of gigawatts — and in some cases, beyond 100 gigawatts — each. Krishna said that, taken together, public and private announcements point to roughly one hundred gigawatts of currently planned capacity dedicated to AGI-class workloads.
At $80 billion per gigawatt, the total reaches $8 trillion. He tied those figures to the five-year refresh cycles common across accelerator fleets, arguing that the need to replace most of the hardware inside those data centers within that window creates a compounding effect on long-term capex requirements. He also placed the likelihood that current LLM-centric architectures reach AGI at between zero and 1% without new forms of knowledge integration.
Krishna pointed to depreciation as the part of the calculation most underappreciated by investors. AI accelerators are typically written down over five years, and he argued that the pace of architectural change means fleets must be replaced rather than extended. “You've got to use it all in five years because at that point, you've got to throw it away and refill it,” he said.
Recent financial-market criticism has centred on similar concerns. Investor Michael Burry, for example, has raised questions about whether hyperscalers can continue stretching useful-life assumptions if performance gains and model sizes force accelerated retirement of older GPUs.
The IBM chief said that ultimately, he expects generative-AI tools in their current form to drive substantial enterprise productivity, but that his concern is the relationship between the physical scale of next-gen AI infrastructure and the economics required to support it. Companies committing to these huge, multi-gigawatt campuses and compressed refresh schedules must therefore demonstrate returns that match the unprecedented capital expenditure that Krishna outlined.
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.
Luke James is a freelance writer and journalist. Although his background is in legal, he has a personal interest in all things tech, especially hardware and microelectronics, and anything regulatory.
-
tennis2 Where is all the investment money coming from? That's my question.Reply
Even if we're only talking about the datacenters' hardware/infrastructure, this is a massive amount of money. Where is it being divested from? -
Eximo It isn't, borrowed against stock price collateral. A giant house of cards with the only real winner being the large capital firms getting the interest while it is still growing.Reply
They all think that they can find a way to make profit eventually. -
chaz_music Replytennis2 said:Where is all the investment money coming from? That's my question.
Even if we're only talking about the datacenters' hardware/infrastructure, this is a massive amount of money. Where is it being divested from?
Individuals and investment firms are throwing money at this with great abandon without doing any reasonable P/E ratio studies. This is worse than the Dotcom bubble as the investments are huge. My neighbor maxed out his home equity load to invest in this. The only true use-case that can support this level of investment and give a reasonable return is for military applications. Even medical diagnostics and robotic surgeries cannot support the amount of investment funding.
We are also forgetting the negative impact it is having on ordinary life, such as:
- Higher utility costs (electric, NG and other fuels, water - evaporative cooling, etc.)
- Availability of certain electronic components and subsystems (DRAM, SSDs, etc.)
- Buildouts of server farms are going to briefly absorb available electrical system materials such as wiring, generators, transformers, substation components, power stations, and the workers to install these.
A very frightening possibility:
With AI startups digging for any profitable usage market, AI is going to be used to drive up any commodity such as housing, cars, food, and others. AI can already make stock market decisions and buy up food futures (orange juice, pork bellies, beef, tobacco, .sugar) resulting in inflation that did not exist at this level before. AI in a capitalistic economy. This can also happen in the medical industry. -
SomeoneElse23 I'm hoping the chat agents the public sees and uses aren't a good representation of the gazillions of dollars being poured into this "AI".Reply
There's either some nefarious about this, or incredibly dumb. -
redgarl AI will be the force generation of capital at the minute proper implementation will become a reality.Reply
When robotic will merge with AI, it will be over. We are talking about Defense industry, self-driving technologies, chemical research, medicine, health industry... basically any white collar and warehouse jobs.
It is not about making money for now, but boosting efficiency and cutting cost, which is already a reality.
We are talking about the biggest companies in the world, they are going to find a way to maintain their objectives to win the AI race, especially with the government backing them.
Everyone denying it will be left in the dust. -
redgarl Reply
Most of the companies involved in the AI race are the biggest market cap companies anyway. The money is not an issue at all. OpenAI is backed by MS owning 27% of the company, not to mention OpenAI having a capital venture of 10% in AMD.tennis2 said:Where is all the investment money coming from? That's my question.
These companies want to win the race because they know it will pay off in the future, like Amazon did 10-15 years ago. Unlike the dotcom bubble, the finances are solid, and products do exist. -
Eximo Reply
Yes, products exist. The problem is that EVERYONE is trying to be THE company with that product. A few are going to succeed, but a whole lot aren't. Acquisitions will absorb some of it, but a lot of people/investors/suppliers etc are not going to survive. A lot of successful companies came out of the dotcom bubble, but a lot didn't, and they are the source of concern.redgarl said:Most of the companies involved in the AI race are the biggest market cap companies anyway. The money is not an issue at all. OpenAI is backed by MS owning 27% of the company, not to mention OpenAI having a capital venture of 10% in AMD.
These companies want to win the race because they know it will pay off in the future, like Amazon did 10-15 years ago. Unlike the dotcom bubble, the finances are solid, and products do exist. -
redgarl Reply
You are... so LATE to it... automation tools are already used for decades and don't involve AI. That's why you see bear markets, like 3 weeks ago, being so bipolar. You are just trying to find something to help your narrative. Matter of fact, AI is there to stay and it will get better and better at a faster rate than the adoption of the smart phone.chaz_music said:AI can already make stock market decisions and buy up food futures...
I am already using it all the time instead of doing searches because it provides me links I can verify and offer me a good inquiry above 80% of the time.
I was looking at buying a car and I investigated if the engine was reliable. I was able to find if there was recalls, reports and life expectancy of vehicles hosting the engine easily. In 5 minutes, I found more information then if I searched for 2 hours. And this is only for information amalgamation.
Just look at what is happening with image recognition and generation lately. All content creators in the industry, either hollywood, game studios or social medias, are all having to face the reality that the technology is forcing them to adapt radically.
We are not even talking about Defense, Robotics and Security. I can anticipate what to expect in 10 years. It is going to be a brutal shift. The Ukrainian war has shown us that Modern Warfare is shifting everything that was the standard of the industry.
I cannot really understand people denying it beside pure denial and bias toward repercussion linked to their hobby as PC builders. -
redgarl Reply
If you want to have the pulse of the industry, you need to check at the semiconductor industry. Everything is in high demand. Waffer, GPUs, Memory, CPUs, server, storage... it is all there and all obvious.Eximo said:Yes, products exist. The problem is that EVERYONE is trying to be THE company with that product. A few are going to succeed, but a whole lot aren't. Acquisitions will absorb some of it, but a lot of people/investors/suppliers etc are not going to survive. A lot of successful companies came out of the dotcom bubble, but a lot didn't, and they are the source of concern.
What is funny about the whole bubble theory, is that it is coming from SHORTERS who are betting against the technology.
Matter of fact is those companies are the biggest money maker in the world. It is not BS money. They are investing in each others because there is an American and Chinese forefront for the AI race. Also, their intend are different. Some are focusing on software, other hardware, some want to maintain a Close ecosystem, others are working for implementing an Open ecosystem.
What we are seeing is a vertical integration instead of the usual horizontal expansion.
The only thing people are having for argument is the circular aspect of the transactions, which is going against the circular nature of the economy anyway. Money is not created, it is distributed, unless the government is printing money.
I don't see the logic behind it beside a bunch of shorters trying to influence and manipulate the markets. They even stated that life cycling of hardware should be around 24 months, which is ludicrous for datacenters, but they are spreading their misconceptions because they are bias to make their narratives with huge money implication for a bet against the markets. -
Eximo Reply
I see you are focusing on the semi-conductor industry. That puts your posts in a different light.redgarl said:If you want to have the pulse of the industry, you need to check at the semiconductor industry. Everything is in high demand. Waffer, GPUs, Memory, CPUs, server, storage... it is all there and all obvious.
What is funny about the whole bubble theory, is that it is coming from SHORTERS who are betting against the technology.
Matter of fact is those companies are the biggest money maker in the world. It is not BS money. They are investing in each others because there is an American and Chinese forefront for the AI race. Also, their intend are different. Some are focusing on software, other hardware, some want to maintain a Close ecosystem, others are working for implementing an Open ecosystem.
What we are seeing is a vertical integration instead of the usual horizontal expansion.
The only thing people are having for argument is the circular aspect of the transactions, which is going against the circular nature of the economy anyway. Money is not created, it is distributed, unless the government is printing money.
I don't see the logic behind it beside a bunch of shorters trying to influence and manipulate the markets. They even stated that life cycling of hardware should be around 24 months, which is ludicrous for datacenters, but they are spreading their misconceptions because they are bias to make their narratives with huge money implication for a bet against the markets.
They will do quite well regardless, because they can easily shift gears to the next desirable hardware product, along with established markets. Though there will be a period when their biggest competitor is their own used hardware being available.
The worry is all the companies over-leveraging themselves to buy that hardware. When that bubble bursts, we well have general economic fallout. Yes, some people who invest in the right companies will win, but lots of people will not. You are right, it is wealth transfer. But lots of losers and only a few winners does not a healthy economy make.