Intel partners with Musk: here's the concrete role in the mega chip factory Terafab
On April 7, Intel officially announced its participation in Terafab, the $20–25 billion AI chip project launched by Elon Musk in March alongside Tesla, SpaceX, and xAI. The Santa Clara company communicated its entry through a post on X, stating that its ability to "design, manufacture, and package high-performance chips at scale" will accelerate Terafab's aim to generate 1 terawatt of computing power annually. Intel's stock reacted with an increase of over 2% on the trading session following the announcement.
Intel is proud to join the Terafab project with @SpaceX, @xAI, and @Tesla to help refactor silicon fab technology. 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.
— Intel (@intel), April 7, 2026
The agreement was reached after a meeting between Intel CEO Lip-Bu Tan and Elon Musk, which took place at the company's campus the previous weekend and was documented by an official photo published by the company. This partnership is significant not only for Intel but especially for the concrete feasibility of Terafab: neither Tesla nor SpaceX has ever manufactured a chip, and building a semiconductor fab is among the most complex infrastructure projects imaginable. Musk had already raised the issue openly during an earnings call: "Can someone else build these things? It’s very hard to build things." The answer, at least for now, is Intel.
What is Terafab?
Terafab was presented by Musk on March 21, 2026, during an event at the former Seaholm Power Plant in Austin, Texas. The project is a joint venture between Tesla, SpaceX, and xAI (the latter acquired by SpaceX in an all-stock deal) and involves constructing a vertically integrated manufacturing plant at the North Campus of Giga Texas for the prototyping phase, with the full-scale facility in a yet-to-be-defined location. The stated goal is to consolidate every phase of semiconductor production under one roof: chip design, lithography, manufacturing, memory production, advanced packaging, and testing. This capacity, according to Tesla, currently does not exist in any other facility in the world.
The technological target is the 2-nanometer node, the most advanced process currently entering commercial production. The initial output capacity target is 100,000 wafers per month, with the long-term ambition of reaching 1 million wafers monthly. Musk has summed up the necessity with a direct formula: “All the production plants on Earth only produce about 2% of what we need. Either we build Terafab, or we won’t have chips.”
Two categories of chips, and the space gamble
Terafab will produce two distinct lines of chips. The first includes the inference chips for Tesla vehicles and humanoid robots (Optimus), including the upcoming generation AI5 and subsequent iterations. It should be noted that even before the announcement of Terafab, Tesla had already shifted the high-volume production of AI5 to mid-2027, partially due to a delay of about six months accrued by Samsung on the 2nm node that directly impacts the roadmap of the next chip, AI6. Small-batch production of AI5 is still expected in 2026. The second category concerns the so-called D3 chips, specifically designed for AI satellites in orbit that SpaceX intends to set up as space data centers.
Indeed, the space component is the boldest aspect of the entire vision: Musk has stated that 80% of the computing capacity produced by Terafab will be allocated for orbital applications, against 20% for terrestrial use. The technical reasoning is that solar irradiance in space is about five times greater than that on the Earth’s surface, and that cooling in a vacuum makes thermal management more efficient than in traditional data centers. Musk concluded, stating that orbital computing could become economically competitive with terrestrial computing within 2–3 years.
Intel: the missing partner
For Intel, the partnership comes at a critical strategic moment. The company is seeking to establish itself as a foundry for third-party customers after years of struggling to find sufficient volumes to make its advanced nodes profitable. Since the appointment of CEO Lip-Bu Tan, Intel has been working to attain the necessary volumes to bring customers to the table, and a client the size of Terafab is exactly the type of customer it needs.
Intel’s role is not limited to mere manufacturing, covering the entire value chain from architecture to finished product. The operational responsibility for constructing and managing the plant will fall directly on Intel, effectively establishing the core of the company’s repositioning as a pillar of American AI infrastructure.
This is not a trivial matter: building a chip fab typically requires several years, investments exceeding $20 billion, thousands of precision machines, and a highly specialized workforce that takes decades to develop. TSMC has spent $165 billion to position its factories in Arizona to move towards 2nm, and production at that node is not expected before 2029. Despite its recent difficulties, Intel is the only player outside of TSMC and Samsung capable of operationally starting such a path in the United States, as the White House observes with interest (including direct investment, given its 10% stake in Intel).