How much compute will Google buy from SpaceX?

Google will pay SpaceX for AI compute—what the deal covers

Google has agreed to pay SpaceX $920 million per month for compute capacity as part of an AI infrastructure arrangement tied to Google’s Gemini Enterprise needs. Multiple filings and follow-up reporting describe this as a major capacity boost, with the compute sourced from SpaceX facilities using Nvidia GPUs.

The key signal is scale and purpose: the deal is framed as bridging capacity and meeting surging demand for Gemini Enterprise rather than being a small pilot. That matters because AI workloads are increasingly constrained by data-center capacity—especially high-end GPU availability—and companies are competing not only on model performance but also on throughput and time-to-deploy.

What’s explicitly described

• Monthly price: $920 million
• Compute type: access to Nvidia GPU infrastructure
• Business goal: additional capacity for Gemini Enterprise

The reporting also emphasizes that the agreement spans a defined period running through mid-2029, indicating Google is planning for continued infrastructure needs, not just short-term overflow.

Why it matters for the industry

This arrangement effectively positions SpaceX as a compute infrastructure player—selling datacenter-like capacity to a major AI customer—while Google treats the partnership as a way to keep operations responsive during rapid growth. It also highlights how AI “supply chain” issues increasingly include not only chip procurement but also power, cooling, and GPU hosting arrangements.

Taken together, it’s another sign that the highest-growth AI systems are pulling demand through the entire data-center stack, and large firms are willing to pay premium prices to secure compute fast.