SpaceX CEO Elon Musk has warned that the company's rapidly expanding AI computing business could experience declining revenue if GPU capacity becomes more widely available and affordable, highlighting the financial risks surrounding multibillion dollar infrastructure agreements. His comments follow speculation that SpaceX could significantly increase the prices it charges customers such as Anthropic if shortages of computing capacity continue. However, Musk emphasized that the same market conditions could also move against infrastructure providers.
In a November 8 post on X, Musk responded to venture capitalist Chamath Palihapitiya, who discussed SpaceX's computing agreement with Anthropic during the All In podcast. Palihapitiya argued that increasingly limited data center capacity could allow SpaceX to terminate its existing agreement with 90 days of notice and negotiate new pricing at approximately 2 to 3 times its current rates. Under that scenario, AI developers requiring substantial computing resources might have limited alternatives.
"It works both ways, of course. If compute becomes plentiful and cheap, SpaceX revenue will drop. No crying in the casino for either party."
— Quote by: Elon Musk, SpaceX CEO.
The agreement represents a substantial component of SpaceX's expanding AI operations. According to the company's official SEC filing, Anthropic agreed to pay $1.25 billion monthly for access to approximately 325,000 NVIDIA GPUs across the Colossus and Colossus II data centers through May 2029. The arrangement could generate approximately $42.5 billion over the specified period, accounting for reduced payments during the initial deployment. However, either company can terminate the agreement with 90 days of notice, meaning the projected revenue is not guaranteed.
A separate Reuters investigation into Anthropic's financial disclosures revealed additional computing agreements potentially worth up to $84.5 billion through 2029. These arrangements also feature cancellation provisions, introducing uncertainty over how much revenue SpaceX will ultimately collect.
The situation reflects a broader challenge facing AI infrastructure providers. Companies are investing billions in NVIDIA accelerators, advanced memory, electricity infrastructure, and cooling systems based on expectations of continued demand. While limited GPU availability can support premium pricing, additional manufacturing capacity and competing data centers could eventually reduce the cost of renting computing resources.
SpaceX is simultaneously expanding its own AI infrastructure, including plans to deploy more than 1 million additional NVIDIA GPUs. The company is also pursuing semiconductor manufacturing ambitions through Terafab, potentially reducing its dependence on external suppliers over the longer term.
For now, Musk's comments describe a potential market risk rather than an announcement of declining revenue or changes to the Anthropic agreement.
Musk's warning highlights a fundamental risk surrounding the AI infrastructure boom: companies are investing enormous amounts of capital in computing hardware whose rental value depends heavily on continued shortages and demand growth.
If AI computing becomes more affordable, developers and smaller businesses could benefit from lower operating costs and better access to powerful hardware. However, infrastructure providers holding expensive GPU deployments could face declining margins and longer investment recovery periods.
For PC gamers and hardware enthusiasts, greater availability of computing resources would be a positive development, particularly if it eventually reduces pressure on GPU manufacturing and memory supply. The technology industry should prioritize sustainable expansion and affordable computing rather than relying on artificial scarcity or continuously rising infrastructure prices.
Could falling AI computing prices eventually benefit PC gamers through cheaper GPUs and memory, or will data center operators continue absorbing most of the available hardware?
