After the initial report on Elon Musk’s planned Terafab project, there were a whole lot of questions about how Musk was going to get his ambitious semiconductor fab project up and running. Now a very, very big piece of the puzzle (like, 85%) has been solved: Intel is going to build it for him.
In an unexpected turn of events, Intel on Tuesday said that it had joined Elon Musk’s TeraFab project. The announcement mentions Intel’s ability to develop, produce, and package advanced processors in high volumes, which could help Tesla, SpaceX, and xAI to get enough compute performance for next-generation AI and robotics applications. However, the announcement made in an X post you can expand below does not reveal how exactly Intel will help TeraFab.
“Intel is proud to join the Terafab project with SpaceX, xAI, and Tesla to help refactor silicon fab technology,” a statement by Intel reads. “Our ability to design, fabricate, and package ultra-high-performance chips at scale will help accelerate Terafab’s aim to produce 1 TW/year of compute to power future advances in AI and robotics. It was fun hosting Elon Musk at Intel this past weekend!”
The announcement is not accompanied by any press releases or SEC filings, which raises questions about the framework of the collaboration between Intel and TeraFab, as well as any possible legal bindings. In fact, the post in X is deliberately written in a way that barely reveals any concrete details about the structure of the partnership.
Officially, TeraFab is positioned as the “most epic chip-building effort ever” that is to combine “logic, memory and advanced packaging under one roof,” which implies localized production in a massive facility. Furthermore, the company is hiring managers to build a greenfield semiconductor fabrication plant in Texas. By contrast, Intel’s wording rather implies a virtual semiconductor production ecosystem, or even a consortium that involves chip design, manufacturing, and packaging at Intel and demand from Tesla, SpaceX, and xAI. How such a consortium would differ from a typical wafer supply agreement that large companies tend to have with their suppliers is something that is unclear at this point.
It makes perfect sense for one of only three global semiconductor manufacturers (and the only American company of the three) to help Musk realize his vision. Though nobody has spelled it out, my guess is it’s going to be very similar to the way Apple funds its manufacturer ecosystem through partner build-outs: Musk is going to pay Intel to build and run the fab, and Musk’s Telsa/SpaceX/xAI consortium is going to buy the dedicated output at a given price for x number of years, after which Intel can use the fab for other projects.
This makes a great deal of sense for both parties, especially if Intel has its 18A process dialed in. This is not a given, as Intel’s process screwups at previous nodes let TSMC and Samsung lap them in the sub-10nm race. But Intel yields at that node are reportedly rising.
And having Intel do all the heavy lifting means Musk’s new company doesn’t have to negotiate with Applied Materials, ASML, LAM, etc. Intel already has standing purchase agreements with all of them, and likely already knows what it wants and what the lead times will be to equip the new fab. Not to mention already knowing which contractors they’ll get to bid to build it.
If that’s actually what’s happening, that clears up a whole lot of questions about the project. Now, it doesn’t make Musk’s crazy “2027 volume production of 1 million wafer starts a months” any more feasible, but a fully functional Intel fab up and running at full-tilt by 2029 pumping out chips for xAI/SpaceX/Telsa is entirely feasible.
Tags: Austin, Elon Musk, Intel, Semiconductors, SpaceX, technology, Terafab, Tesla Motors, Texas, xAI
Not thrilled that the Biden era CHIPS act will get the credit for getting Intel to invest domestically and show the market that it can credibly build the Terafab facility.
Didn’t Trump collect 10% of Intel last year when they couldn’t meet their loan commitments? That was back at a market price of ~$20 or so. INTC closed today at $61.74. And people say that Trump isn’t a deal maker! Last April on Independence Day, INTC hit a low of about $19 due to the tariff announcements. Always remember, Trump is the dip buyer’s best friend.
Not that I know.
I’ve been out of the semi world for a bit, but I’m not aware of Intel having a lot of rad hard IC experience, nor much in space rated packaging.
I do know that at the beginning SpaceX was not really interested in space grade devices, nor any stories about single event upset tolerance, going instead for commercial grade stuff (even SRAM FPGAs) and just building multiply redundant redundancy at their board/module level.
Note this was from meetings quite early for them, so maybe they evolved their thinking since. But for this fab Intel might end up needing help on space/rad tolerant stuff from the sector of the semi design world that have devices sitting on other planets, especially for designing rad-tolerant at something as tiny as 18A.
My suspicion is that the Rad-Hard rec is going to fall by the wayside, at least at the beginning, since no one has a solution for that anywhere near sub-10nm right now. I suspect they’ll just use extra shielding, which is orders of magnitude cheaper, and maybe have a pilot line lab for developing Rad-Hard approaches.
Although if Musk can get down to Drexler’s “rod logic” nanotechnology, that’s probably robustly radiation resistant…
Yep, Uncle Sam bought 433.3 million share at $20.47 per share on 8/22/25
https://newsroom.intel.com/corporate/intel-and-trump-administration-reach-historic-agreement.
I paid $19.74 on 4/4/25. Heh
That is an interesting question.
You’re building a large data center in space. Two alternatives:
– Use space-rated chips in a loose array because they’re slow, so communication delay doesn’t matter much. No (or little) shielding required. Put lots of data centers (or huge ones) up to get the required compute.
– Use normal chips in a dense configuration (have to be because of lightspeed communication delay). Shield the whole thing and have enough redundancy and error checking that error rates are down around terrestrial levels. Probably need a better cooling solution.
Which one gets you to the compute level you want faster?
With oldspace this wouldn’t even be a question because there’s not enough money in the world to do either.
But with Starship (assuming it works like Musk says it will), one or both may actually be possible.
Of course if Starship is somehow a bust, it doesn’t work. How much are you willing to bet on that though?
Also see Starlink and the really enormous constellation. It’s actually about the hardest service to disable via anti-satellite measures. Shoot one down and you’ve changed nothing. About the only way to disable it is to shoot enough down that you end up with a Kessler problem. And then everyone loses access. But the ones with enough launch to do that don’t want a Kessler endpoint, because orbit is too useful for them.
If this comes off it will make Musk “beyond bank.”
A lot of people don’t realize the degree to which cities are welcoming data centers because it’s a huge tax windfall with notable but extremely manageable downsides **and little to no infrastructure burden.**
5 guys run your data center, there’s little to no traffic there, but you’re on the hook tax-wise for the value of that DC. That value, given the hardware in there, is INSANE.
Is space expensive? Oh yes. Is it more expensive than massive property taxes? Doubtful once amortized.
On shielding, this is what Starship’s cost-to-orbit, especially at their internal bill rate, makes practical. Honking heavy shielding was unpossible with legacy costs to orbit their mass, so devices and packaging designs had to be more clever, much more tested, and thus many multiples more expensive. If you can fly bog standard Intel inside chips in commercial packaging solutions because you can afford to surround them with the equivalent of heavy lead plates, space electronics economics change radically.
Whatever they are, they’re not going to be bog standard Intel CPUs, they’re going to be AI-focused chips optimized for tensor operations. In Terafab, Intel will (I suspect) be operating like a foundry with a single customer.
[…] being built to produce AI chips for Elon Musk’s Telsa/SpaceX/xAI consortium, was going to be built and run by Intel, but the location was up in the air. Now, according to regulatory filings, it looks like the […]