Overview
- SpaceX disclosed a multi‑year cloud services agreement with Google that commits roughly $920 million per month at full rate, with Google to use about 110,000 NVIDIA GPUs plus CPUs and memory on SpaceX infrastructure.
- The contract includes a ramp period at reduced fees before full payments start in October 2026 and requires SpaceX to make the committed GPUs available by September 30, 2026 or face a one‑month grace period followed by potential contract termination or pro‑rata fee reductions.
- Google keeps ownership of any models, data, and intellectual property processed on the leased machines, using the capacity as bridge compute for its Gemini models and cloud customers rather than transferring AI ownership to SpaceX.
- When combined with SpaceX’s earlier Anthropic arrangement, analysts say the two deals could shift investor views by adding many billions in potential recurring revenue, though reported annualized and lifetime figures depend on full delivery and no early terminations.
- Operational and disclosure questions remain: reporters and analysts are flagging where the GPUs will be hosted, which GPU generations will be supplied, and whether SpaceX can meet tight build and deployment schedules ahead of its planned public listing.