Musk Unveils Terafab: Tesla and SpaceX's $20B Chip Mega-Facility

The global semiconductor business is not being used by Tesla（TSLA） or SpaceX at all. In a livestream on X, Elon Musk announced that Tesla, SpaceX, and xAI will work together to build Terafab, the "largest chip manufacturing facility ever." It will be built in Austin, Texas, and is expected to cost at least $20 billion.

Musk himself made it clear why: "We either build the Terafab or we don't have the chips." We're going to build the Terafab because we need the chips.

Why Existing Supply Chains Aren't Enough

The size of the project comes from a supply gap that Musk sees as catastrophic. Musk said that the total capacity of current chip manufacturers is only about 2% of what Tesla and SpaceX will need for future computing needs, even though they have relationships with TSMC, Samsung, and Micron, the three biggest chipmakers in the world.

You can't fill that gap by getting more wafers. Terafab's stated goal is no less than one terawatt of annual computing output. This amount is meant to support the growth of self-driving cars, humanoid robots like Optimus, and an orbital data center network that SpaceX has already asked the Federal Communications Commission to build with a million satellites.

The factory will make two different chip architectures: an edge-inference chip that works best with the Optimus robot and robotaxi fleets, and a radiation-hardened, high-power variant for SpaceX's space-based infrastructure. Musk says that these hardware needs are so specific that no external roadmap can reliably meet them.

The Vertical Integration Play — and the Capital Stack Behind It

Musk is building up his tech business in a bigger way with Terafab. In February, SpaceX bought xAI. Since then, Tesla has invested $2 billion in xAI and added the Grok chatbot to its range of cars. The facility's goal of making 2-nanometer chips would put it on the cutting edge of semiconductor production, a field that is currently controlled by TSMC. To reach this level of engineering skill, very few companies in the world have shown it.

Some of the money for these projects will likely come from SpaceX's planned $50 billion IPO, which is expected to be a record-setting offering. Now, Terafab's reputation is tied to that IPO because SpaceX will use orbital data centers running on Terafab chips as a key part of its growth story for public market investors.

The Credibility Question Musk Can't Outrun

An asterisk is attached to the statement, which must be considered for responsible analysis. Musk has a terrible history of forecasting the delivery dates of game-changing devices. The Hyperloop was never completed.

The Cybertruck was far more expensive than the $40,000 that was promised, and it wasn't released until years later than anticipated. For over a decade, fully autonomous vehicles have been "one year away."

The majority of individuals would concur that producing semiconductors is the most challenging and costly industrial process. Intel struggled to catch up to TSMC in terms of 2nm technology despite investing a significant amount of time and money in the process.

Building a new facility that targets that node and simultaneously serves space-based and ground-based applications is an extremely challenging engineering task, regardless of the sponsor's financial resources.

Terafab is the type of announcement that clearly divides Tesla investors. Bulls see a vertically integrated robotics and AI powerhouse protecting itself from supply chain vulnerabilities, creating a structurally sound moat if execution is successful.

Bears envision another Musk moonshot consuming funds that Tesla's primary EV business need as the firm deals with growing margin pressure and competition. For a project this ambitious, the $20 billion price tag is a floor, not a ceiling.

Whether Terafab becomes a competitive advantage or a warning footnote will depend on timeline discipline rather than vision.