Google signs SpaceX Artificial Intelligence compute deal

Google is securing short-term compute capacity from SpaceX to support Gemini Enterprise demand. SpaceX could take in more than $30bn if the agreement runs its full 32-month term.

SpaceX has sealed another major deal to rent out compute for Artificial Intelligence, this time with Google. The terms were disclosed in a filing with the U.S. Securities and Exchange Commission. Google has committed to pay SpaceX $920mn per month for access to about 110,000 Nvidia GPUs, CPUs, memory, and other related components.

The arrangement is set to start in October and last until June 2029, according to SpaceX. Should SpaceX fail to provide access to the agreed number of GPUs by September 30, Google will have the option to terminate the agreement with a one-month grace period or accept the number of GPUs provided and pay a reduced pro rata fee. After Dec. 31, either party can end the agreement with 90 days’ notice. But should it run its full 32-month term, SpaceX stands to take in more than $30bn.

The Google agreement follows a similar arrangement with Anthropic, which was confirmed in early May. Under the terms of that deal, the Claude creator is renting 300 megawatts of compute capacity at SpaceX’s Colossus data centers in Tennessee for $1.25bn a month. SpaceX had hinted that more deals were likely when the Anthropic agreement was made public last month.

SpaceX’s IPO is expected as soon as this week, when it is seeking a valuation in the region of $1.75tn, and becoming a key provider of Artificial Intelligence infrastructure appears central to its strategy. The Financial Times calculated that the Anthropic and Google deals alone will contribute more than $26bn to the conglomerate’s annual revenue. Google said the agreement will provide bridge capacity to meet surging customer demand for Gemini Enterprise. At the start of June, Google parent company Alphabet launched an equity sale of $80bn as part of an effort to fund investments in its Artificial Intelligence compute infrastructure.

78

Impact Score

Google taps Intel for in-house Artificial Intelligence chips

Google’s potential order for in-house Artificial Intelligence accelerators would strengthen Intel’s contract manufacturing push as chip designers look beyond TSMC. Nvidia is also evaluating Intel technology for a multi-chip processor.

European Commission clarifies high-risk Artificial Intelligence rules

The European Commission’s draft guidance maps how high-risk Artificial Intelligence classifications may apply to agentic systems, regulated products, biometrics, employment tools, and financial services. The guidance signals narrow exceptions and a broad view of systems that materially influence regulated decisions.

NVIDIA confidential computing expands Apple private cloud compute

NVIDIA GPUs with Confidential Computing will support confidential inference for Apple’s Private Cloud Compute as it expands to Google Cloud. The collaboration brings NVIDIA Blackwell GPUs into the hardware security architecture behind some next-generation Apple Intelligence features.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.