Musk confirms talks with TSMC over Terafab, which could leave Intel's 14A node out

Elon Musk has confirmed that talks are under way with TSMC about a possible collaboration on Terafab, the semiconductor manufacturing project intended to supply Tesla, SpaceX and xAI exclusively. The confirmation came on Saturday in a post on X, in which Musk wrote tersely that these are "just conversations, but something may come of them," without giving further details. News of the negotiations first emerged on Friday through Culpium, journalist Tim Culpan's newsletter, while TSMC has made no comment on the matter.

According to Culpan's analysis cited by Tom's Hardware, the scenario he considers most likely (referred to here as 'Plan A') would have TSMC build and operate the new plant, while SpaceX or Terafab itself would put up capital, commit to buying a guaranteed volume of chips, or combine both formulas. The exact ownership structure would remain undefined. This arrangement would echo the joint ventures TSMC already runs in Japan (JASM) and Germany (ESMC), where local partners provide capital and demand while TSMC supplies the process technology and runs the fab.

A second possibility floated ('Plan B') would reverse the roles: SpaceX would retain majority ownership of Terafab, with TSMC investing less but contributing its operational experience and its manufacturing processes. This model would be an absolute first for TSMC, the world's largest contract chipmaker. The article insists that, with nothing official, both 'Plan A' and 'Plan B' are merely Culpan's speculation and not the only options on the table; there could be other variants not yet considered.

One relevant detail provided by Culpium is that rumors about TSMC's intention to build a fab complex in Texas would, on this reading, be tied to the aim of supporting Terafab.

TSMC's potential entry raises questions about Intel's role in the project. So far, Intel is the only named technology partner for Terafab, slated to supply its 14A process (1.4 nm class) to manufacture the chips for SpaceX, Tesla and xAI. If TSMC gets involved bringing its own manufacturing node, Intel's 14A process could be left out of the project.

The piece also offers an alternative reading centered on advanced packaging, an area where TSMC is the industry leader: Terafab could manufacture the chips in Texas, presumably with Intel's 14A process, and then task TSMC with packaging them in the same region, a more limited role than that of main manufacturer.

Another factor complicating the picture is patents. Intel has a broad cross-licensing agreement with TSMC, but Terafab has no equivalent deal, so manufacturing chips with a node developed by Intel could infringe some TSMC patent. That circumstance would explain why Musk would want to keep TSMC close, to the point that it might come in as a partner or investor in Terafab.

Finally, the article underlines that Musk is not seeking to turn Terafab into a foundry competing with Intel Foundry, Samsung Foundry or TSMC itself. His goal would be to secure uninterrupted supply with room for custom silicon for his companies, aimed at artificial intelligence workloads, vehicles and space projects. Given that broad strategic goal, the article's author (Anton Shilov) suggests that Musk is probably keeping flexibility in both tactics and strategy to achieve it.

It is worth stressing that, as of today, these are all preliminary talks confirmed very tersely by Musk, with no details on timelines, investment figures or schedule, and without any confirmation from TSMC.