Google rents 110,000 GPUs from SpaceX as Gemini demand strains capacity

AI compute demand is now large enough that even Google is buying bridge capacity from SpaceX. TechCrunch reported at 11:57 AM PDT on June 5, 2026, that Google agreed to pay SpaceX $920M per month for access to roughly 110,000 NVIDIA GPUs, CPUs, memory, and related components. The full-rate period runs from October 2026 through June 2029.

The agreement was disclosed in a SpaceX regulatory filing. Capacity will ramp through September at a reduced fee, and both parties can terminate the agreement with 90 days’ notice after December 31, 2026. If SpaceX does not deliver access to the committed GPUs by September 30, 2026, Google can end the agreement after a one-month grace period or accept fewer GPUs with lower monthly fees.

The scale is notable because Google is not a compute-poor AI startup. It is widely viewed as one of the largest owners of AI infrastructure, yet a company representative told TechCrunch the deal is a short-term answer to stronger-than-expected demand for recent AI products, including Gemini Enterprise. Alphabet has already committed more than $180B in capital expenditures this year and expects that figure to rise significantly in 2027.

For SpaceX, the contract strengthens a second story around its pre-IPO business: not just rockets and Starlink, but rentable AI infrastructure. The Google deal follows a similar Anthropic agreement worth $1.25B per month. The important question is whether these deals are temporary overflow capacity or the beginning of a normal pattern in which frontier AI companies combine owned data centers with massive external compute leases.