Altogether, TeraFab would require well north of $4 trillion, which generally aligns with Bernstein's $5 trillion estimate, excluding land, process R&D, software, and ecosystem development.
Raising $5 trillion would be extraordinarily difficult. For context, companies like Nvidia, Apple, and Alphabet have market capitalizations of $4.34 trillion, $3.71 trillion, and $3.5 trillion, respectively, so Musk would need to mobilize capital exceeding the value of the world's most valuable corporations. It is hard to imagine a private fundraising round, consortium, or even sovereign funding of this magnitude. For example, even if the U.S. would like to fund Musk's semiconductor venture, it could not do this that easily, as its budget for this year is about $7 trillion.
The only conceivable path would involve multi-government backing, sovereign wealth funds, hyperscalers, and capital markets acting in concert. However, we doubt this is possible at all. Furthermore, at a scale of $5 trillion deployed within a foreseeable timeframe, constraints would extend beyond capital and would include limited availability of wafer fabrication equipment, construction materials, and a sufficiently large and skilled workforce to build, operate, and maintain such fabs.
Then again, does Musk really plan to build a foundry that would leave behind TSMC, Samsung, and Intel combined just to make enough chips for Tesla, SpaceX, and xAI? Well, this is an open question.
Anton Shilov is a contributing writer at Tom\u2019s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends. ","collapsible":{"enabled":true,"maxHeight":250,"readMoreText":"Read more","readLessText":"Read less"}}), "https://slice.vanilla.futurecdn.net/13-4-19/js/authorBio.js"); } else { console.error('%c FTE ','background: #9306F9; color: #ffffff','no lazy slice hydration function available'); } Anton Shilov Social Links Navigation Contributing Writer Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.
usertests If they pull out technology that can substantially beat ASML EUV, maybe there's a chance, and multi-trillion estimates can be revised downward. If they plan to buy ASML machines, forget about it. There's always the possibility of a technological surprise upending the status quo, but you should assume it won't happen in this field. Substrate's claims about revolutionary ASML-beating chipmaking technology scrutinized, analyst likens the venture to a fraud — report pokes holes in the startup's technology, messaging, and leaders Reply
spooh_ Bernstein seems to calculate exactly how TSMC would manufacture nvidia chips. But Tesla/SpaceX doesn't try to compete with them – they're aiming to manufacture just two highly specialized chips in insane amounts for internal needs. Then without time or cost pressure from external clients they can try things unimaginable for TSMC Reply
Elrabin spooh_ said: Bernstein seems to calculate exactly how TSMC would manufacture nvidia chips. But Tesla/SpaceX doesn't try to compete with them – they're aiming to manufacture just two highly specialized chips in insane amounts for internal needs. Then without time or cost pressure from external clients they can try things unimaginable for TSMC You do realize that Tesla HW3 is 100 watts, HW4 is 160 watts and HW5 is "up to 800 watts when processing complex environments" right? The Bernstein calculations are based on 1400 watt B300 accelerators. If we swap that out for HW5 800 watt target TDP, we have to increase the fab numbers by 60% to hit the power target output per year. So your argument makes Elon's claim even MORE ridiculous. HW5 isn't even out yet. The design might be complete, but it's not shipping yet. Reply
DougMcC Elrabin said: You do realize that Tesla HW3 is 100 watts, HW4 is 160 watts and HW5 is "up to 800 watts when processing complex environments" right? The Bernstein calculations are based on 1400 watt B300 accelerators. If we swap that out for HW5 800 watt target TDP, we have to increase the fab numbers by 60% to hit the power target output per year. So your argument makes Elon's claim even MORE ridiculous. HW5 isn't even out yet. The design might be complete, but it's not shipping yet. But the terafab hasn't even broken ground yet. It would surely manufacture HW8 at the earliest. The general idea holds: if Terafab can't get any price advantage for optimizing for a single target, it would be colossally embarrassing. The only real question is how much cost savings you get out of the optimization, and this price estimate doesn't seem to factor this in at all. Reply
Stomx "Elon Musk's eventual ambitions include producing millions or billions chips that consume 1 terawatt (1 TW) of power per year. " The introduced unit "terawatt per year" is really something new the technical world never heard so far 😀 Reply
DS426 There we go, now we're looking at more realistic figures given the specs that Musk laid out for this project. $20b just gets the project underway. Reply
Elrabin DougMcC said: But the terafab hasn't even broken ground yet. It would surely manufacture HW8 at the earliest. The general idea holds: if Terafab can't get any price advantage for optimizing for a single target, it would be colossally embarrassing. The only real question is how much cost savings you get out of the optimization, and this price estimate doesn't seem to factor this in at all. It is exceptionally unlikely that they're going to get a price advantage in excess of the $20-25bn to break ground let alone the hundreds of billions or even trillions needed to hit these types of production targets. Simple math. Even if they reduced price per chip by 50%, which is unlikely, are they buying billions of dollars of chips per year to make this make sense? 2025 Tesla sold a grand total of 1.65 million vehicles globally. The company's GAAP Operating Income was $4.4 billion. Please explain how spending hundreds of billions to make net income of $3.8 billion makes sense? The actual chip costs here are a fraction of the income. In Jan 2026, average Tesla price was $52,628. The math doesn't math man Reply
SkyBill40 Typical Musk boast, complete with unrealistic and laughably unattainable end results. Are we to the point yet of just sidestepping anything he says given so much of it is bordering on ridiculous? Reply
truerock If Musk said, “The Terafab is expected to produce one terawatt of computing power in chips each year." First, there is no time-frame of when the goal would be reached. Second "one terawatt of computing power each year" is an not a clear and meaningful phrase. Reply
alan.campbell99 SkyBill40 said: Typical Musk boast, complete with unrealistic and laughably unattainable end results. Are we to the point yet of just sidestepping anything he says given so much of it is bordering on ridiculous? I really do think this is how he should be treated. Does he ever say this stuff in a venue or situation where he can immediately be pressed with critical questions in response rather than into his managed echo chamber? Reply
Key considerations
- Investor positioning can change fast
- Volatility remains possible near catalysts
- Macro rates and liquidity can dominate flows
Reference reading
- https://www.tomshardware.com/tech-industry/semiconductors/SPONSORED_LINK_URL
- https://www.tomshardware.com/tech-industry/semiconductors/elon-musks-terafab-semiconductor-project-could-cost-usd5-trillion-bernstein-claims-herculean-effort-would-cost-more-than-70-percent-of-the-total-yearly-us-government-budget#main
- https://www.tomshardware.com
- Arm moves beyond IP with AGI CPU silicon — 136-core data center chip targets AI infrastructure with Meta as lead partner
- Flipper Zero pen-testing tool gets an AI-powered companion app — natural language interface allows for faster, easier hacking
- HP's new AI workstation can expand horizontally to add 15% more internal volume — alternate side panel also includes more active cooling
- Kentucky farm family rejects $26 million offer for 600 acres of land from unnamed AI data center suitor — declines 7x offer, wants to ‘Stay and hold and feed a
- NVIDIA, Telecom Leaders Build AI Grids to Optimize Inference on Distributed Networks
Informational only. No financial advice. Do your own research.