Google will pay SpaceX $920 million per month for AI computing resources.

      **Summary**: Google will pay SpaceX $920 million each month for access to 110,000 Nvidia GPUs until mid-2029, marking the second significant compute agreement SpaceX has finalized ahead of its anticipated $75 billion IPO. Google refers to this as bridge capacity to support the rising demand for Gemini Enterprise.

      SpaceX has entered its second major compute agreement within a month, this time with Google. According to a regulatory filing released on Friday, Google will pay SpaceX $920 million monthly from October 2026 through June 2029 for access to around 110,000 Nvidia GPUs, CPUs, memory, and other related components, totaling approximately $30 billion over the contract’s duration.

      This deal comes just one week before SpaceX is expected to trade its stock on Nasdaq, aiming to raise about $75 billion at a valuation of roughly $1.75 trillion, potentially making it the largest IPO in history.

      **The second Colossus agreement**

      This agreement is structured similarly to SpaceX's recent deal with Anthropic, where Anthropic agreed to pay $1.25 billion monthly through 2029 for the entire output of Colossus 1, a Memphis data center originally built by xAI for its chatbot Grok before it merged with SpaceX in February.

      Google's agreement seems to cover nearly half the compute resources that Anthropic has at Colossus 1. SpaceX did not clarify which data center Google would utilize. CEO Elon Musk has indicated that the Colossus 2 facility may be reserved for xAI’s projects.

      Both agreements include a 90-day cancellation option after December 31, 2026. Google’s access will increase gradually through September at a reduced rate. Should SpaceX fail to deliver the promised GPU capacity by September 30, 2026, Google can terminate immediately after a one-month grace period or accept available hardware at a reduced rate.

      **The need for Google**

      This deal is significant as Google is not lacking in compute resources. Some sources identify it as the largest single owner of AI compute, largely due to its custom TPU chips. The company has already allocated over $180 billion for capital expenditure in the current year, with an expectation for a significant rise in 2027. Alphabet has recently announced an $80 billion equity sale to support this spending.

      A Google spokesperson referred to the SpaceX deal as a temporary solution. "This is a timely short-term agreement to ensure we have bridge capacity to meet the increasing customer demand for our agent platform, Gemini Enterprise, which has exceeded our expectations," the company stated.

      This indicates that Gemini Enterprise is utilizing GPU capacity more rapidly than Google’s data center expansion can keep up with. For a company investing $180 billion annually in infrastructure, renting Nvidia GPUs from a rocket company is an unusual acknowledgment of demand exceeding supply.

      **Transforming costs to revenue for SpaceX**

      For SpaceX, these compute agreements are turning a cost center into a revenue generator. While xAI developed Colossus to train Grok, user engagement with the chatbot has reportedly declined. Instead of leaving the GPUs unused, SpaceX is leasing them to the two companies in urgent need of compute resources: Anthropic, which was heavily resource-constrained before the agreement, and Google, which is expanding its infrastructure but cannot do so quickly enough.

      With the contracts from Anthropic and Google, SpaceX is now generating approximately $2.17 billion in monthly revenue from compute, amounting to around $26 billion annually. This figure will be a significant part of the IPO prospectus. Additionally, Google has been a long-time investor in SpaceX, with its stake expected to exceed $100 billion post-listing. The companies are also discussing the development of orbital data centers, a key element of SpaceX’s plans following the IPO.

      The overall picture is striking: the world’s largest search company is renting Nvidia GPUs from the world’s largest rocket company, as demand for AI compute has surpassed even the most ambitious infrastructure expansions. Whether this demand is sustainable, or indicative of a peak that will later correct, remains a concern in relation to SpaceX’s $1.75 trillion valuation.