To secure additional computing power for its Gemini Enterprise subscribers, Google has signed a massive $30 billion AI deal with SpaceX, which owns Elon Musk's xAI. According to an SEC filing by SpaceX, the company will receive $920 million a month from Google starting this October through June 2029, in exchange for computing power from xAI's data centers.
The agreement grants Google access to 110,000 NVIDIA GPUs, along with associated CPUs and memory resources. SpaceX noted in the filing that if it fails to deliver this designated GPU capacity by September of this year, Google has the right to either terminate the contract immediately or accept the available GPUs at a discounted rate after a one-month grace period.
Since Google operates an extensive global network of proprietary data centers and is continuously building more, the deal raises questions about its reliance on SpaceX. A Google Cloud spokesperson told CNBC and The New York Times that this represents "a short-term, timely agreement" to provide "bridge capacity" to satisfy the surging enterprise demand for Gemini.
Google is not the only AI competitor relying on SpaceX's compute infrastructure. Anthropic has also contracted with SpaceX for access to xAI’s Colossus 1 data center, agreeing to pay $1.25 billion monthly through May 2029. These details surfaced in SpaceX's IPO filings, with the company scheduled to go public on June 12 in what is expected to be the largest IPO in history.
[AgentUpdate Depth Analysis] The aggressive land grab for compute by Google and Anthropic underscores a key shift in the AI Agent landscape: "Compute supply chain sovereignty is now the ultimate competitive moat over algorithmic uniqueness." Even Google, with its proprietary TPU infrastructure, faces immediate "bridge capacity" deficits when scaling enterprise-grade AI Agents. This bottleneck highlights how multi-agent architectures and complex workflows exponentially scale inference demands. For the AI Agent ecosystem, while developer frameworks democratize Agent creation, production deployment remains tightly bottlenecked by hardware. The monetization of xAI's Colossus 1 infrastructure proves that speed-to-market in data center scaling is currently the most valuable currency in AI, forcing Agent startups to design with strict compute-efficiency in mind to survive rising token costs.
