Join for quick access. Enter your email below and we'll send confirmation plus sign you up to our newsletter.
By submitting your information, you confirm you are aged 16 or over,
have read our
Privacy Policy
and agree to the
Terms & Conditions. Geographical rules apply.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
You are now subscribed
Your newsletter sign-up was successful
Join the club
Get full access to premium articles, exclusive features and a growing list of member rewards.
An account already exists for this email address, please log in.
Google reportedly in talks with SpaceX to launch its orbital data centers — partnership could mark a historic turning point and boost upcoming IPO
Putting AI servers in space has been discussed as a holy grail of sorts for some time now. The economics of an orbiting data center would benefit from always-available solar power, even considering the relative difficulty in cooling the rack units. The main issue is the stratospheric price tag of lifting that compute to orbit. Now, though, according to a Wall Street Journal report, Google believes that SpaceX might be able to make the dream real.
According to the report, Google is in talks with SpaceX and a few other contenders about this strategy, though given how Elon Musk's orbital enterprise has steadily become by far the main player in commercial launches, it's the clear front-runner in those talks. Google's move may be related to the company's Project Suncatcher initiative, revealed last November, that intends to send satellites laden with Google Tensor Processing Units (AI chips) into orbit starting in 2027.
This news has the potential to boost the impending SpaceX IPO to infinity and beyond. That offering is expected to be the largest of all time, and was already expected to reach stratospheric levels of $1.5 to $1.7 trillion. As of this writing, neither company has offered any comment on the presumably ongoing negotiations.
It's worth noting that SpaceX recently struck a partnership with Anthropic that could include "multiple gigawatts of orbital AI compute capacity", and that it filed an application last January with the FCC to launch up to a million satellites for datacenters, so SpaceX would doubtless be happy for another client in this space.
The notion of space AI datacenters has long been derided as a fever dream, even by OpenAI honcho Sam Altman himself, given the financial delta-V required to place thinking rocks in orbit. Estimates pin the theoretical launch cost for SpaceX itself at around $2,700 per kilogram, an amount that works out to a best-case scenario of $3,400/kg for a customer, assuming a completely stuffed rocket — something that's hard to achieve in practice.
That reason is precisely why SpaceX's February 2026 price table lists $7,000/kg as a standard rideshare price, to fill in the gaps and maximize profit during a launch (or minimize losses, depending on how you slice it).
The math for Google's Project Suncatcher says that the financial equilibrium for space datacenters sits at around the $200/kg mark, not even in the same galaxy as current figures. Yet the economics of SpaceX's Falcon 9 rockets are driving that cost down. One such Falcon 9 recently launched for the 34th time in a row, and some analysts think it's literally a matter of space-time until five to six reuses of the same ship are enough to offset its production cost. After that, in theory, the only major expenses are fuel, maintenance, and launchpad utilization.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
It's still hard to say if low-Earth-orbit meme generation will become a reality, but it's a reasonable enough conclusion that SpaceX is currently the only entity that can pull it off. The firm has made 165 launches in 2025, more than the rest of the world combined, up from 134 in 2024. Likewise, it has put 14,844 payloads in orbit in total, and it's reportedly only 218 units away from having launched as many satellites as every country on Earth since space became reachable.
Bruno Ferreira is a contributing writer for Tom's Hardware. He has decades of experience with PC hardware and assorted sundries, alongside a career as a developer. He's obsessed with detail and has a tendency to ramble on the topics he loves. When not doing that, he's usually playing games, or at live music shows and festivals.
Google is also one of the major owners of SpaceX makes sense. Why would they use anyone else they are more than a “partner”. They are an “owner” of SpaceX?
Relative difficulty of cooling. Hmm, looking up on NASA docs the ISS PVR can reject up to 14kW of heat into space using about 42 square meters of panel, 740.7 kg weight.
As for power generation, it would seem that with some additional new arrays for the ISS there is 215kW available from it's arrays. From I can find on a quick search a TPU v3 pod of 512 chips draws 100-120kW.
Just to try and visualise the scale here, I'm also not sure I saw any mention of these chips needing to be radiation-hardened and/or whatever else is needed for them to operate in orbit.
I have wondered at times what, if any industry, could be economically and practically put into orbit instead of taking up space and other concerns on the surface. AI slop factories are nowhere on such a list .
Am I the only one who thinks putting AI in space is a horrible idea? I mean...we really are going to just put all the most advanced AI out of reach in a place where we literally can't pull the plug and just hope it goes cool?
Starting to wonder if the AI is already sentient and just playing us.
Relative difficulty of cooling. Hmm, looking up on NASA docs the ISS PVR can reject up to 14kW of heat into space using about 42 square meters of panel, 740.7 kg weight.
As for power generation, it would seem that with some additional new arrays for the ISS there is 215kW available from it's arrays. From I can find on a quick search a TPU v3 pod of 512 chips draws 100-120kW.
One problem is that photovoltaic power generation scales a lot better than heatsinks. To scale up a heatsink, you need effective ways of transporting the heat, which starts to become bulky and heavy, the longer the distances are that we're talking about.
