SpaceX is reportedly planning to manufacture its own GPUs, the chips that power artificial intelligence. The revelation comes from excerpts of its S-1 registration, a document companies file with the U.S. Securities and Exchange Commission before going public.
As reported by Reuters, SpaceX lists “manufacturing our own GPUs” among its biggest capital expenditures in the future. This comes a month after Elon Musk announced its own TeraFab chip factory focused on developing chips that can survive the harsh conditions of space and power its orbital AI data centers.
Why is SpaceX making its own chips?
The short answer is supply. In its TeraFab announcement event, Elon Musk stated that even if they buy all existing chipsets, it will only cover 2% of their future requirements.
In the filing, SpaceX has further warned prospective investors that it does not have long-term contracts with many of its chip suppliers, and there is no guarantee it can secure enough compute hardware to support its growth.
On paper, designing and producing your own chips is the obvious solution if you cannot buy enough. But chipmaking is one of the most complex endeavors a company can take on, and SpaceX is not exactly a semiconductor company, at least not yet.
Is making GPUs even realistic?
Honestly, it’s a massive challenge. In the same S-1 filing document, SpaceX has warned that its plans for orbital data centers may not achieve commercial success. Advanced chip manufacturing is about as complex as it gets, with thousands of tightly controlled steps that have to go perfectly every single time.
TeraFab has a long way to go before it can master these complexities. There’s a reason ASML is the only company that sells photolithography machines, and TSMC has a virtual monopoly over high-end chip production.
Musk has stated that Terafab will handle every step of chip production, including design, fabrication, packaging, and testing, all under one roof. Whether SpaceX can pull this off remains to be seen.