Tesla will retain operational control of its Terafab chip-manufacturing project, Elon Musk has confirmed, even as the company has held discussions with Taiwan Semiconductor Manufacturing Co. (TSMC) about potential collaboration.

The statement narrows but does not close the door on a partnership with TSMC, the world's largest contract chipmaker. Musk's position, as carried in trade press, is that Tesla will run Terafab even if the talks with TSMC produce a deal.

What is Terafab?

Terafab is Tesla's in-house semiconductor initiative, framed by Musk as part of the company's push to vertically integrate AI compute. The project is intended to give Tesla direct access to advanced-process AI silicon, reducing dependence on third-party foundries for the processors that run its autonomous-driving software and training infrastructure.

The build-out extends an existing pattern. Tesla already designs inference silicon for its Full Self-Driving (FSD) computer and runs its Dojo training platform on custom chips. Terafab takes the company one step further, into the physical fabrication of the chips themselves.

Why does Tesla want to keep control?

Operating authority over Terafab gives Tesla direct control over three variables that determine competitive position in AI hardware: which process node to use, how much capacity to allocate, and when new designs move into production. In the foundry model, the customer pays for wafers but does not own the line, the tools, or the recipes.

TSMC has served as Tesla's primary foundry partner across multiple generations of AI silicon and remains the only contract manufacturer with high-volume production at the leading edge. Any Terafab plan therefore has to be weighed against the cost and time required to build a competitive fab from scratch.

What do the TSMC talks cover?

Neither company has disclosed the substance of the discussions. The plausible scenarios range from TSMC providing process-design kits and engineering support, to a long-term capacity agreement under which Tesla buys wafers from TSMC while running Terafab independently, to a narrower technology-licensing arrangement.

Musk's confirmation that Tesla will retain control of Terafab is consistent with all three. Tesla runs the fab; TSMC supplies whatever piece of the puzzle Tesla decides it still needs.

Why this matters commercially

The economics of advanced semiconductor manufacturing make pure captive production hard to justify. Industry estimates routinely place the cost of a leading-edge fab at well over $20 billion, with multi-year construction timelines. Terafab's viability depends on Tesla bringing enough of its AI chip volume in-house to absorb that capital outlay — a calculation that shifts as the company's external demand for training compute continues to scale.

What to watch next

The next concrete signal will be whether Tesla announces a formal partnership with TSMC covering Terafab, and whether the company begins site preparation or equipment orders within the timeline Musk has previously discussed. Musk's statement preserves Tesla's optionality on capacity while keeping the operational levers inside the company.