Google commits $920M monthly to SpaceX for 110,000 GPUs through 2029

The Deal: $10.1B in GPU Access

Google will pay SpaceX $920 million per month from October 2026 through June 2029 for access to approximately 110,000 NVIDIA GPUs, CPUs, memory, and related components, according to TechCrunch AI reporting on a regulatory filing. The 32-month contract totals roughly $10.1 billion and includes a ramp-up period at reduced rates through September 2026, followed by full-price commitment from October onward.

Both parties retain termination rights with 90 days’ notice after December 31, 2026. If SpaceX fails to deliver the committed GPU allocation by September 30, Google may immediately exit or accept reduced capacity with corresponding fee reductions—a performance guarantee unusual in infrastructure-as-a-service contracts.

Context: Colossus Capacity and Competitive Positioning

The agreement arrives one week before SpaceX’s anticipated Nasdaq debut at a $1.75 trillion valuation, raising approximately $75 billion. According to TechCrunch AI, this deal mirrors the structure of SpaceX’s May agreement with Anthropic, in which Anthropic committed to $1.25 billion monthly through 2029 for Colossus 1 data center capacity. Google’s 110K-GPU allocation represents roughly half Anthropic’s access, suggesting SpaceX is segmenting its existing infrastructure across multiple hyperscaler customers or reserving additional capacity for future deals.

Google’s rationale—stated by a company representative to TechCrunch AI—frames the arrangement as a short-term stopgap: “This is a short-term, timely agreement to ensure we have bridge capacity to meet surging customer demand for our agent platform, Gemini Enterprise, which has been even higher than we expected.” The characterization underscores that even vertically integrated compute owners like Google face capacity constraints when demand spikes faster than capital deployment can accommodate.

Alphabet has committed to more than $180 billion in capital expenditures during 2026 and signaled that spending will “significantly increase” in 2027. The $920M monthly burn for external GPU rental—despite Google’s historical position as one of the world’s largest AI compute holders—reflects the breadth and velocity of generative AI adoption among enterprise customers.

Why This Matters

The SpaceX–Google transaction signals that GPU scarcity remains the binding constraint in the AI infrastructure market, even for cash-rich incumbents with substantial internal capacity. Third-party data center operators like SpaceX can now command premium long-term commitments from cloud providers, validating the strategic value of vertically integrating compute ownership alongside space infrastructure. For teams evaluating cloud vendor stability and resource availability in 2027, this deal demonstrates that even Google’s commitments to Gemini Enterprise customers may depend on external, vendor-controlled capacity allocation—a dependency worth factoring into architecture and procurement decisions.