SpaceX Plans to Develop its Own GPU, Warning Investors About Chip Supply and Costs
SpaceX is potentially tackling one of the biggest challenges in the chip industry: manufacturing the key components that power artificial intelligence – graphics processing units (GPUs). Ahead of a potential IPO this summer with a target valuation of $1.75 trillion, the company has warned potential investors that it plans to invest heavily in developing AI and other technologies.
According to excerpts from SpaceX’s S-1 registration statement, the company lists “manufacturing our own GPUs” as one of its ongoing “significant capital expenditures.” Companies must submit S-1 filings to the U.S. Securities and Exchange Commission before going public to disclose their risks and financial condition.
SpaceX and its xAI division previously announced a partnership with Tesla to jointly develop Terafab – an advanced AI chip manufacturing center in Austin, Texas, planned and overseen by CEO Elon Musk.
While Musk has said the project will produce chips for cars, humanoid robots, and space data centers, many details, including the types of AI chips to be produced (such as GPUs), remain unknown.
There are many different designs for AI chips. For example, Nvidia primarily produces GPUs, which are versatile chips good at performing a variety of data processing tasks. Alphabet’s Google has taken a different approach with its Tensor Processing Units (TPUs), which are specifically optimized for certain functions crucial for building AI models and running chatbots like Anthropic’s Claude.
It is currently unclear when SpaceX plans to produce its own chips, or which companies – Terafab’s developer or its partner Intel – will be responsible for the manufacturing technology inside the factory.
Musk told analysts on Wednesday that by the time Terafab scales, Intel’s next-generation 14A manufacturing process “may be quite mature or available,” and “that seems like the right path.”