.png){kind=logo}
SpaceX has told investors that its AI expansion may push it toward one of tech’s toughest jobs: making GPUs. In its S-1 filing, the company listed possible in-house GPU manufacturing as a major cost associated with AI and related technologies. The disclosure arrived ahead of a summer IPO that could value SpaceX at about $1.75 trillion.
An S-1 is the registration document companies file with the U.S. Securities and Exchange Commission before going public. In the filing, SpaceX pointed to large capital needs as it builds out AI capabilities. At the same time, it warned that chip supply may not keep pace with its growth.
The filing also linked that effort to Terafab, an AI chip complex under development in Austin, Texas. SpaceX is building the site with xAI and Tesla. Elon Musk has said the project targets chips for cars, humanoid robots, and space-based data centers.
Even so, major details remain unclear. SpaceX has not said what type of AI chips Terafab will actually produce. One central issue is whether the company means traditional GPUs or a broader class of AI processors.
That uncertainty matters because different chip designs serve different goals. NVIDIA focuses on GPUs, which power heavy compute tasks. By contrast, Google uses TPUs built for narrower AI jobs such as model training and chatbot inference.
The company has not publicly revealed what it is working on. It has also not said when in-house chip production might begin. In addition, it remains unclear which Terafab groups, or partner Intel, would handle fabrication technology inside the plant.
Musk offered one clue during Tesla’s Wednesday analyst call. He said Intel’s next-generation 14A manufacturing process should look mature, or close to full readiness, by the time Terafab reaches scale. He also said that path “seems like the right move.”
The S-1 made clear that supply remains a concern. SpaceX said it lacks long-term contracts with many direct chip suppliers. That gap could create problems as demand for AI computing hardware rises.
The company also told investors it expects to keep buying a significant share of its computing hardware from third-party suppliers. As a result, Terafab may not meet its goals on schedule. The filing added that those goals may not be achieved at all.
Meanwhile, the broader chip challenge remains steep. Designing advanced processors is difficult, and manufacturing them at scale is even harder. NVIDIA, for example, leads in GPU design but still relies on TSMC in Taiwan for production.
TSMC built that position over years of investment and repeated execution. Advanced chip production uses exotic materials and more than a thousand tightly controlled steps. Years of iPhone chip output gave TSMC the experience needed to keep advanced processors flowing at scale.
Read More: SpaceX Gains Option to Buy AI Startup Cursor for $60 Billion Ahead of Planned IPO
Alongside the hardware plan, SpaceX tied its AI push to Cursor, the coding startup linked to a $60 billion deal. This week, SpaceX said it had secured the right to acquire Cursor. Before that, CNBC reported that Microsoft had explored a deal.
Microsoft later stepped back. Now it is trying to widen the reach of its own AI tools. It has made progress with GitHub Copilot, yet the AI coding market is now led by Cursor, Anthropic, and OpenAI.
Microsoft still plays a major role in that market as both investor and cloud provider. It has invested billions in Anthropic and OpenAI, and both companies are committed to heavy spending on Microsoft Azure.
Soon after the SpaceX announcement, a company post said SpaceX AI and Cursor were working closely to build “the world’s best coding and knowledge work AI.” Cursor chief executive Michael Truell then wrote on X that he was excited to partner with the SpaceX team to scale Composer.
The deal also moved quickly. The agreement came together late in Cursor’s fundraising process, and prospective investors did not expect the deal. In the weeks before the announcement, SpaceX had already given Cursor access to compute.
SpaceX told investors that its AI expansion may require in-house GPU production, linking the effort to Terafab in Austin and a broader chip strategy. The filing also warned of supply risks and ongoing dependence on third-party hardware, placing execution and timing at the center of investor attention.
