Google Signs $920M Monthly Compute Deal With SpaceX

SpaceX said in a regulatory filing that Alphabet unit Google will pay $920 million per month from October 2026 through June 2029 for access to approximately 110,000 Nvidia GPUs, along with CPUs, memory, and other components, according to Bloomberg, CNBC, TechCrunch, and Euronews. The filing describes a capacity ramp with reduced monthly fees while SpaceX brings hardware online, and a delivery clause allowing Google to terminate the contract or accept fewer GPUs with a pro-rata fee reduction if SpaceX does not meet the GPU commitment by September 30, 2026, CNBC reports. Bloomberg and other outlets calculate the pact at roughly $30 billion over its term.

Per reporting in CNBC and TechCrunch, the agreement covers GPUs and supporting servers and mentions that capacity will ramp through September at a reduced fee; the filing gives both parties the option to end the agreement after December 31, 2027, provided 90 days notice. TechCrunch and CNBC note this deal resembles the late-May arrangement between SpaceX and Anthropic, where SpaceX agreed to provide large-scale compute capacity to a major AI developer.

Industry coverage in TechCrunch, Bloomberg, and Tom's Hardware places the deal in the context of SpaceX's February merger with xAI and the group's rapid buildout of Colossus-class data centers. Tom's Hardware, citing Reuters, reports that the combined annual value of SpaceX's recent compute deals could exceed the company's 2025 revenue from Starlink, launch services, and AI, a point highlighted in multiple outlets as a meaningful revenue shift ahead of SpaceX's planned IPO.

Editorial analysis - technical context: Companies buying third-party data-center capacity to bridge short-term peak demand is a growing pattern as model sizes and inference workloads spike. For practitioners: large offsite capacity deals commonly introduce challenges around network egress, workload colocation, and consistent hardware mixes; teams often need to validate performance on mixed-generation GPU fleets and plan for staged capacity ramping rather than instantaneous availability.

Industry analysis: The deal is reported as a near-term response to surging enterprise demand for agentic and large-context services. CNBC quotes a Google Cloud spokesperson saying the agreement helps "ensure we have bridge capacity to meet surging customer demand for our agent platform, Gemini Enterprise." Public reporting frames this transaction alongside Anthropic's contract with SpaceX and the firms' emphasis on securing predictable access to thousands of accelerators.

Monitor the SpaceX regulatory filings for further detail on which data centers host the capacity and on the mix of GPU generations deployed. Observers should also follow published utilization results or performance benchmarks from early customers to assess how mixed fleets (H100, H200, GB200 and others) perform together. Finally, watch public filings and market commentary around SpaceX's IPO for how underwriters and investors value contracted, recurring compute revenue relative to legacy aerospace and Starlink income.