GOOGLE RENTS ELON MUSK’S AI HARDWARE — $920 MILLION A MONTH TO SPACEX FOR 110,000 NVIDIA GPUS GOOGLE ITSELF CANNOT BUILD FAST ENOUGH Google has agreed to pay SpaceX $920 million every month from October 2026 through June 2029 for access to roughly 110,000 NVIDIA GPUs sitting inside a SpaceX-controlled data center. The deal totals approximately $30 billion over the life of the contract and was disclosed in a regulatory filing on June 5th. The compute in question traces back to xAI, Elon Musk’s AI venture, which built the Colossus 1 facility near Memphis to train its Grok models. SpaceX absorbed xAI in February 2026 and the hardware came with it. Anthropic signed a parallel deal two weeks earlier, agreeing to pay SpaceX $1.25 billion a month for full Colossus 1 capacity. Google is taking roughly half that amount from what appears to be a separate facility. In a statement, Google called the arrangement bridge capacity, saying demand for its Gemini Enterprise platform has exceeded expectations. Consider what that admission actually means. Google is widely estimated to be the world’s largest private owner of AI compute, with hundreds of thousands of its own custom TPU chips. And it still ran short. The compute shortage driving this deal is not a temporary blip. Alphabet has committed over $180 billion in capital expenditures this year and says spending will rise significantly again in 2027. The deal also closes one week before SpaceX begins trading publicly at a $1.75 trillion valuation. Keywords: Google SpaceX compute deal, AI GPU shortage, SpaceX Colossus data center, Gemini Enterprise capacity, AI infrastructure spending