The correction landed on X with the force of a door slamming. A user had posted that the most likely outcome for Terafab, Elon Musk’s planned AI chip complex in Texas, was that Taiwan Semiconductor Manufacturing Co. would end up owning and running the plant. “No, we will build and run the fab,” Musk replied. “Let there be ZERO doubt about that.” He added that TSMC might sublease part of the site “if they want, but nothing more than that.”
The exchange, reported Wednesday, closed off a week of speculation that the world’s largest contract chipmaker was about to walk into the project as its operator. The message was unmistakable: the factory belongs to Tesla and SpaceX, and no outside foundry is going to run it.
The same day, on the other side of the world, Intel chief executive Lip-Bu Tan offered the counterweight. Speaking at an event in Tokyo marking Intel’s 50 years in Japan, Tan told Bloomberg that Intel would remain involved in Terafab, reaffirming a role it first took in April as the project’s first named partner. The juxtaposition was a little awkward: Intel stays, while its far larger Taiwanese rival has been reduced, in Musk’s telling, to a possible tenant.
Terafab began in March as a joint effort of Tesla, SpaceX and xAI, the AI startup that has since merged into SpaceX, to build what Musk has called a “gigantic chip fab.” In August the companies named Grimes County, Texas, about an hour from Houston, as the site and committed $16.8 billion to a first phase. The stated goal is one terawatt of computing capacity a year, chips destined for Tesla’s Optimus robots and Cybercabs and for the space-based data centers SpaceX is developing.
Intel’s contribution has been the process technology. Musk said on a recent earnings call that Intel’s 14A process would be the manufacturing backbone for Terafab’s output, and in April Intel committed to design, fabricate and package the chips, work that on paper handed Intel the operational core of a project Musk had sold as a break from outside chipmakers.
That is where the ambiguity now sits. Neither Tesla nor SpaceX has ever produced a silicon wafer. Tan gave no detail on what shape Intel’s continuing involvement would take, and Musk’s statement reads as a desire to own and direct the site himself, with Intel supplying process technology rather than operating the plant. In that arrangement, Musk gets the control and Intel takes the technology risk.
The speculation about TSMC arose for a reason. Intel, once the unquestioned leader in chip manufacturing, has spent several years trying to rebuild that position after losing ground to TSMC and Samsung in the race to the smallest process nodes, and its turnaround has been uneven. To industry watchers, the idea that Musk’s project might quietly slip into TSMC’s hands was a natural reading of a partnership with a foundry that has struggled to execute.
The timing matters because TSMC has been weighing a broader expansion of its U.S. footprint, separate from the $265 billion it has already committed to an Arizona campus. Reports had suggested Terafab could become an anchor customer for a prospective TSMC factory in Texas. Musk’s correction makes clear that if such a plant is built, TSMC would be building it for itself, not inheriting his.
Musk’s impulse runs the other way. He has spent years arguing that the companies he controls should own their most important parts, from rocket engines to battery cells, rather than rent them from suppliers. Terafab is that argument applied to the component on which everything else now depends: the AI chip. Whether his companies can actually make one is the question the next few years will answer.
The episode also carries a personal dimension. Musk has spent months positioning Terafab as proof that his companies can break their dependence on suppliers that control the most sought-after technology in the world. Bringing Intel’s process inside his own walls is, in that telling, less about saving money than about removing a point of failure he cannot afford. The question the industry is now watching is whether that independence is real, or whether it is a promise made years before any wafer leaves the line.
For the chip industry, the episode shows how the balance of power has shifted. The shortage years turned foundries into kings; now the customer with the deepest pockets writes the terms. Musk is not asking TSMC to make his chips. He is saying he will make them himself, on Intel’s processes, and that the foundry that once would have been the obvious answer can rent space if it likes.
Analysts said the project remains years from producing a single chip, with a pilot line reportedly targeted at 3,000 wafers a month and silicon production from 2029, and that the absence of any wafer experience at Tesla or SpaceX is the largest open question. For now, though, the ownership question has been settled in the bluntest possible terms. Terafab is Musk’s to build, and Musk’s to run.
