A reported SpaceX IPO filing suggests Google has committed nearly $920M per month to rent AI compute capacity from SpaceX, including around 110,000 NVIDIA GPUs from late 2026 through mid-2029.
That would put the deal at roughly $29B over the full term.
Why does this matter?
Google already builds its own TPUs. Alphabet is also planning massive infrastructure spending. Yet demand for Gemini Enterprise and agentic AI products appears to be growing faster than even Google can provision capacity internally.
Anthropic has reportedly signed a similar large-scale compute deal with SpaceX as well.
The broader signal is clear: AI infrastructure is becoming one of the biggest bottlenecks in tech.
Even the companies with the deepest pockets and strongest internal hardware strategies may need external compute bridges to keep up with demand.
