Elon Musk’s claim about TSMC and Terafab is narrower than a chipmaking partnership. “Just discussions, but something may come of it,” Musk wrote in a post reported by Tom’s Hardware. His Saturday confirmation establishes that the parties have talked. It does not establish what they agreed to do, if anything.
That distinction matters because SpaceX presents Terafab as a way to secure more AI chips. A conversation with a potential manufacturing partner makes that goal worth watching. It supplies no evidence, by itself, of a factory commitment or additional chip capacity. The claim passes one test—whether talks occurred—and fails to clear two others: whether the parties have struck a deal and whether Terafab can produce chips.
What the TSMC report establishes
Tim Culpan reported exclusively in his Culpium newsletter on Friday that TSMC was exploring ways to work with Terafab. Culpan described TSMC owning and operating a facility as the most likely arrangement under consideration, while leaving open other forms of cooperation. That is reporting about a possibility, not a disclosed contract.
Musk’s Saturday response supports the central, limited point of that report: discussions happened. It does not confirm that TSMC chose an ownership structure, agreed to operate a plant, committed money or reserved production capacity. Treating his reply as confirmation of any one of those steps would turn an acknowledgment of contact into an announcement he did not make.
Tom’s Hardware reported that TSMC had not commented on the talks as of Saturday. Silence is neither an endorsement nor a rejection of a proposed arrangement. It leaves the company’s intended role, if any, outside the public record available here.
The strongest case for taking the report seriously is straightforward. Culpan reported more than a hypothetical pairing, and Musk confirmed that discussions took place. Negotiations have to begin somewhere. But the evidence that people are considering a plan is not evidence that either side has accepted its costs or obligations.
What SpaceX has committed to Terafab
The project already has a more substantial paper trail than the TSMC talks. SpaceX says it announced a collaboration with Tesla in March 2026 to develop Terafab. Its filing describes a general framework for future development, with individual projects still requiring separate negotiations and agreements. Timelines, milestones and capital spending had not been determined in that account.
SpaceX also says Intel joined Terafab in April 2026 and was expected to contribute expertise in chip design, fabrication—the manufacture of chips—and packaging, which connects and assembles chip components. That is a stated role, not evidence that Intel has completed manufacturing for the project. Nothing in the reported TSMC talks establishes that Intel has left or would be displaced.
SpaceX describes an effort spanning chip design, the fabrication of logic and memory chips, and advanced packaging. It gives Terafab a long-term goal of producing one terawatt of compute hardware annually. The number is a company target, not measured output. The available filings and reports do not establish installed Terafab production capability or chip output.
The caution is in SpaceX’s own disclosures. Its prospectus warns that Terafab may not succeed, says Tesla and Intel are not obligated to remain involved and expects SpaceX to keep sourcing a significant portion of its compute hardware from third parties. Those disclosures do not make the project pointless. They do make it difficult to treat its intended scale as a forecast of supply.
Free newsletter
Get the morning briefing
Start each day with the AI stories that matter and why — a short, free email from our newsroom.
Why the distinction matters for AI chips
SpaceX says Terafab is intended to expand its access to chip manufacturing and address potential future shortages. TSMC, in its 2025 annual report, expected robust AI-related demand in 2026 and described strong demand for leading-edge silicon and advanced packaging. That outlook helps explain why a possible manufacturing arrangement would attract attention. It says nothing about whether TSMC has capacity set aside for Terafab.
There is a credible strategic argument for pursuing such talks: SpaceX wants more secure access to compute hardware, and a manufacturing partner could be relevant to that ambition. Yet the argument is about incentives, not results. TSMC’s market outlook is its own forecast, and the available record contains no independent assessment showing that Terafab can manufacture chips at its stated scale.
The next evidence should be concrete. A public agreement could establish who would invest, build, operate or buy from a facility, and on what schedule. Evidence of installed equipment, qualified production and actual chip output would test the separate claim that the arrangement adds supply. Those are different thresholds; even a signed deal would not prove production capacity.
Verdict: talks confirmed; collaboration and chipmaking capacity unproven. Musk’s statement answers whether TSMC and Terafab have spoken, not what either will deliver. Until obligations are disclosed and production is demonstrated, Terafab’s proposed contribution to AI chip supply remains a plan rather than a measured result.