SpaceX has lined up another colossal compute deal ahead of its historic IPO, this time with Google. The company announced the agreement in a regulatory filing on Friday. Under the terms, Google will pay SpaceX $920 million per month from October 2026 through June 2029 for access to "approximately 110,000 NVIDIA GPUs, CPUs, memory, and other related components."
The deal closely mirrors the length and scope of the agreement SpaceX announced with Anthropic in late May. As part of that arrangement, Anthropic agreed to pay $1.25 billion monthly through 2029 to rent all available compute from the Colossus 1 data center near Memphis, Tennessee. That facility was originally built by xAI—now integrated into SpaceX—for its own artificial intelligence endeavors.
Google's contract secures roughly half the compute capacity that Anthropic enjoys at Colossus 1. SpaceX did not specify which data center Google would utilize, though CEO Elon Musk has previously indicated that the Colossus 2 facility would be reserved exclusively for xAI.
Prior to its deal with SpaceX, Anthropic was severely bottlenecked by compute capacity, only raising usage limits the day the partnership was announced. Google, however, is in a drastically different position, widely estimated to be the world's largest single owner of AI compute. A representative explained the deal stems from unexpected demand for recent AI products. "Google Cloud and SpaceX are long-time partners," the statement read. "This is a short-term, timely agreement to ensure we have bridge capacity to meet surging customer demand for our agent platform, Gemini Enterprise, which has been even higher than we expected."
Meanwhile, parent company Alphabet is on a massive spending spree, committing over $180 billion to capital expenditures this year with expectations to "significantly increase" by 2027. To support this infrastructure push, Alphabet recently announced an $80 billion equity sale.
Similar to the Anthropic contract, Google's agreement includes a cancellation clause. Both parties can terminate the deal with 90 days notice after December 31, 2026. Google's data center access will ramp up "through September at a reduced fee." The filing states: "If we fail to deliver access to the committed amount of GPUs by September 30, 2026, then following a one-month grace period, Google may immediately terminate the agreement or accept the number of GPUs provided" with reduced monthly fees.
SpaceX announced this strategic deal just a week before its stock is slated to begin trading on the Nasdaq. SEC filings reveal the aerospace giant aims to raise roughly $75 billion at a staggering valuation of $1.75 trillion, making it the largest IPO in history. Google, a longtime investor, is poised to see massive returns on its existing stake.
[AgentUpdate Depth Analysis] Google's unprecedented $920 million monthly compute deal with SpaceX is a stark indicator of the immense infrastructure toll demanded by the AI Agent ecosystem. Gemini Enterprise, Google's premier agent platform, requires massive concurrent processing for multimodal interactions and continuous tool usage, pushing inference costs to historic highs. Compared to Anthropic monopolizing Colossus 1, tech behemoths are shifting from mere model training to securing operational inference clusters for autonomous agents. This partnership reveals that even hyperscalers like Google face "capacity bridging" crises amidst explosive agentic workflows. Strategically, infrastructure giants with superior power grids and networking—like SpaceX and its subsidiary xAI—are emerging as the ultimate gatekeepers of the agent era. For developers leveraging frameworks like LangChain or CrewAI, this massive consolidation signals a future where raw compute is heavily monopolized. Moving forward, the true competitive edge in the AI Agent ecosystem will depend heavily on optimizing routing logic, minimizing redundant LLM calls, and driving down inference overhead to survive the tightening squeeze on global GPU availability.
