Alphabet’s Google commits $920 million per month for SpaceX compute capacity, highlighting how AI infrastructure spending is reaching new scale

Alphabet’s Google has agreed to pay SpaceX $920 million a month for AI compute capacity, according to a report that cited a regulatory filing. The arrangement underscores how AI demand is pushing major tech companies to secure large volumes of processing power, often through complex, multi-year infrastructure deals.

The agreement is described as spanning roughly 32 months, with pricing set at the $920 million monthly rate. The timeline runs from October of this year through June 2029, with “capacity ramping up through September at a reduced fee,” according to the same reporting that pointed to the filing.

The filing described Google’s planned use of about 110,000 Nvidia graphics processing units (GPUs), along with central processing units, memory, and other components housed at SpaceX’s data centers. GPUs are specialized chips used to train and run machine-learning models, including large language models that power many AI services.

The deal also includes leverage for Google if SpaceX cannot deliver the committed amount of GPUs by a deadline. If SpaceX fails to provide access to the required GPU capacity by September 30, 2026, Google can either end the agreement immediately or accept fewer GPUs at a reduced fee after a one-month grace period. After this year, either party can terminate the agreement with 90 days’ notice.

A Google Cloud spokesperson told CNBC that the contract was intended to provide “bridge capacity” to meet surging customer demand for its agent platform, Gemini Enterprise. Gemini Enterprise refers to subscription offerings for large businesses that want managed access to Gemini-based capabilities for enterprise use cases.

For investors trying to gauge the cost structure of AI initiatives, the scale of the commitment matters. At $920 million a month, the financial magnitude is large enough to be noticed in quarterly spending trends, especially in periods when companies are simultaneously investing in model development, data centers, and cloud services.

The timing of the reporting also ties the commitment to broader attention on SpaceX’s plans for an IPO, which has been discussed publicly as a potentially record-setting event. That context has made the Google deal stand out, because it points to SpaceX’s ability to monetize compute capacity rather than relying solely on rocket launches.

Still, much remains undisclosed in the public reporting. Details such as the exact performance specifications of the rented systems, how billing changes during ramp periods beyond the described “reduced fee,” and how the agreement interacts with Google’s internal data center plans were not fully spelled out in the cited coverage, leaving room for uncertainty about the ultimate economic impact beyond the headline commitment.