In a move that underscores just how voracious the appetite for AI compute has become, Google has signed a landmark agreement to rent computing infrastructure from SpaceX, committing to payments of $920 million per month. The deal, disclosed in a regulatory filing on Friday, comes just one week before SpaceX's long-anticipated initial public offering.
Under the terms of the contract, Google will gain access to approximately 110,000 NVIDIA GPUs, CPUs, memory, and related hardware starting in October 2026. The arrangement runs through June 2029, making it a roughly 2.5-year bridge deal intended to shore up capacity while Google builds out its own infrastructure.
A Google spokesperson framed the agreement as a response to unexpected demand for its recently launched AI products. "This is a short-term, timely agreement to ensure we have bridge capacity to meet surging customer demand for AI services," the company said in a statement. Google Cloud and SpaceX have collaborated for years, and this deal deepens that relationship significantly.
The agreement closely mirrors a separate compute deal SpaceX struck with Anthropic in late May, under which Anthropic committed to paying $1.25 billion per month to access the full capacity of the Colossus 1 data center near Memphis, Tennessee. The Colossus facility was originally built by xAI, now part of SpaceX following Elon Musk's consolidation of his AI and aerospace ventures. By contrast, Google's deal appears to cover roughly half the compute resources that Anthropic secured, and SpaceX has not disclosed which specific data center Google will be using. Musk has previously indicated that the Colossus 2 facility is earmarked for xAI's own workloads.
The scale of these back-to-back compute arrangements positions SpaceX as a significant player in the data center leasing market, a role few would have anticipated for a company primarily known for rockets and Starlink satellites. For Google, which is widely considered one of the world's largest individual owners of AI compute, the decision to lease capacity from an outside provider illustrates just how dramatically demand has outpaced even hyperscaler projections.
Industry observers note that the timing relative to SpaceX's IPO is notable. With two major compute contracts totaling over $2 billion per month now on the books, SpaceX enters the public markets with a high-value recurring revenue stream from two of the most prominent names in AI. Analysts are watching closely to see whether other tech giants follow suit with similar arrangements as AI inference workloads continue to scale.
Why It Matters
This deal signals that even the largest cloud providers cannot build data center capacity fast enough to keep pace with AI demand. The normalization of hyperscalers renting compute from third parties at nine-figure monthly rates reflects a fundamental shift in how AI infrastructure is being sourced and financed. For enterprises evaluating their own AI strategies, the message is clear: GPU scarcity is a macro-level challenge, not just a startup problem.