Google Is Paying SpaceX $920 Million a Month for AI Compute — the Deal Disclosed in an SEC Filing

Google will pay SpaceX $920 million per month for access to AI compute capacity under a multi-year agreement disclosed in an SEC filing on Friday, June 6. The deal runs from October 2026 through June 2029, with capacity ramping up through September at a reduced rate before the full monthly price kicks in. At full term, it could net SpaceX approximately $30 billion.

Google gets access to roughly 110,000 Nvidia GPUs — along with associated CPUs, memory, and related infrastructure — hosted at SpaceX's Colossus data centers. A Google Cloud spokesperson described the arrangement to the New York Times as "a short-term, timely agreement to ensure we have bridge capacity" to meet surging demand for its Gemini Enterprise agent platform.

Why Google Is Renting From SpaceX

The backstory is straightforward: Google is short on compute. Gemini Enterprise — its AI platform aimed at businesses deploying AI agents at scale — has seen demand outpace what Google's own data center buildout can supply in the near term. Rather than wait for new capacity to come online, Google is renting from whoever has chips available now.

SpaceX has chips because it built them for xAI, Elon Musk's AI lab, which has struggled to compete at the frontier against OpenAI, Anthropic, and Google itself. The Colossus data centers in Memphis, Tennessee, and Mississippi were constructed at extraordinary speed — Musk publicly claimed a 19-day buildout for the initial Colossus cluster — and the capacity now exceeds what xAI can productively absorb at its current scale. Leasing excess capacity to Google and Anthropic is a rational way to monetize the infrastructure investment while xAI continues development.

SpaceX Already Has a Bigger Deal With Anthropic

The Google agreement is SpaceX's second major AI infrastructure deal. Anthropic signed a $1.25 billion per month contract with SpaceX for access to 220,000 GPUs at Colossus — a larger arrangement by both GPU count and monthly fee. Combined, Google and Anthropic are now paying SpaceX over $2 billion per month for compute capacity, positioning SpaceX as one of the dominant AI infrastructure providers globally, alongside AWS, Azure, and Google Cloud itself.

The timing of the Google disclosure is notable: SpaceX filed the agreement with the SEC as part of its IPO preparation, with a market debut expected next week at a valuation above $1.7 trillion. Google holds approximately five percent of SpaceX — a stake that gives it a direct financial interest in a strong IPO. The compute deal is genuinely useful for Google regardless, but the disclosure also functions as a high-profile advertisement for SpaceX's AI infrastructure business to prospective IPO investors.

Termination Clauses and Uncertainty

The agreement includes provisions that limit its certainty. Google can terminate immediately if SpaceX fails to deliver the committed GPU capacity by September 30, 2026. After December 31, 2026, either party can exit with 90 days' notice. Both exits suggest neither company is treating this as a long-term infrastructure relationship — Google's cloud buildout will eventually close the capacity gap, and SpaceX's own AI ambitions may reclaim the compute. The deal is a bridge, not a foundation.

What the agreement demonstrates concretely is that AI compute demand has grown large enough to make SpaceX — a rocket and satellite company — a credible cloud infrastructure competitor, at least for specific, large-scale capacity needs. The AI infrastructure race is no longer solely about who builds the best chips; it's about who can deploy enough of them, fast enough, where customers need them.