SpaceX Plans to Manufacture Its Own GPUs at New Semiconductor Fab, S-1 Filing Reveals

Leaked excerpts from SpaceX's confidentially filed S-1 registration statement, cited by Reuters and reported in detail by Tom's Hardware, reveal that the company intends to manufacture its own graphics processing units at a new in-house semiconductor facility. The filing lists GPU production among the reasons for what SpaceX describes as substantial planned capital expenditures, as the company prepares for an initial public offering valuing it at $1.75 trillion.

The disclosure is the first confirmation that SpaceX's chip ambitions extend beyond using silicon developed at Tesla to include GPUs — a category of chip central to powering AI models and agentic workloads. SpaceX warned prospective investors that it does not hold long-term supply contracts with many of its direct chip suppliers and may face difficulty purchasing sufficient silicon to meet its growth targets, providing the strategic rationale for in-house production.

The GPU plan is connected to Terafab, a semiconductor manufacturing complex that Elon Musk is developing in Austin, Texas, in partnership with Tesla and xAI. On the same day the S-1 details emerged, Musk disclosed that Terafab will use Intel's 14A manufacturing process node — the most advanced process Intel has in development — suggesting that Intel will handle the fabrication technology inside the facility.

Musk told Tesla analysts that Intel's 14A process will likely be mature or ready for prime time by the time Terafab reaches production scale.

Ambiguity remains over whether SpaceX uses the term GPU to refer to conventional graphics processors or to specialized AI accelerators, a naming convention Tesla has applied inconsistently to its own AI5 and AI6 chips. The size of the investment and the timeline for first silicon have not been disclosed.

Read the original reporting at Tom's Hardware.