SpaceX is Planning to Spend Up to $500 Billion on Data Centers. Here's Why Elon Musk is Pushing All-In
Key Points:
- SpaceX plans to expand its AI compute capacity from 1.4 gigawatts to between 6 and 10 gigawatts, potentially spending $300 billion to $500 billion in 2027 alone, driven by lucrative deals with Anthropic and Google generating tens of billions annually per gigawatt.
- Major tech players like Microsoft, Google, Amazon, Anthropic, and OpenAI are engaged in an aggressive AI infrastructure arms race, with Microsoft alone adding 31 new data centers in Q4 and expecting $175 billion in capex for 2026.
- High inference margins, such as Anthropic’s 85% gross margin and GPU payback periods under a year, are fueling rapid AI infrastructure investments, with upcoming IPOs from SpaceX, OpenAI, and Anthropic potentially unlocking trillions in market value to accelerate the buildout.
- Taiwan Semiconductor (TSMC) is a critical bottleneck for AI infrastructure expansion due to its limited advanced node capacity, despite raising capex to $60-$64 billion and investing $265 billion in Arizona facilities to meet soaring demand through 2029-2030.
- Key risks to watch include potential drops in spot compute pricing, TSMC’s capacity constraints, and the possibility that supply could eventually outpace demand, which could undermine the current high return economics and trigger a market correction in AI infrastructure stocks.