Google Is Renting 110,000 Nvidia GPUs From SpaceX for $920 Million a Month

Google is turning to SpaceX for a major boost to its AI computing capacity, with the company agreeing to pay $920 million per month for access to a large pool of Nvidia GPUs and other data-centre hardware.

The agreement was disclosed in a SpaceX filing with the US Securities and Exchange Commission. Under the terms of the deal, Google will receive access to approx. 110,000 Nvidia GPUs, along with CPUs, memory and other related components.

Google will pay the full $920 million monthly fee from October 2026 through June 2029. Capacity will gradually ramp up during the months leading to October, with Google paying a reduced fee during that initial period.

Google Needs More Computing Power for Gemini Enterprise

The deal is designed to give Google additional capacity as demand for its AI products continues to grow. In a statement shared with TechCrunch, Google described the agreement as a short-term measure intended to help the company meet unexpectedly strong demand for Gemini Enterprise, its AI platform for businesses.

Google already operates some of the largest AI infrastructure in the industry, but the rise of generative AI has pushed major technology companies to secure as much computing capacity as possible. Training and running increasingly advanced AI models requires vast numbers of specialised processors, particularly Nvidia GPUs.

The scale of the new SpaceX deal reflects just how intense that demand has become. At $920 million per month, the agreement would be worth more than $11 billion per year if it continues at the full rate.

The SpaceX Deal Includes an Exit Clause

Despite the headline figure, the agreement includes several conditions. SpaceX is expected to provide access to the agreed number of GPUs by September 30, 2026. If it misses that target, Google can terminate the contract after a one-month grace period or accept a smaller number of GPUs in exchange for a proportional reduction in the monthly fee.

After December 31, 2026, either company can also end the agreement by providing 90 days’ notice.

The filing states that Google will retain ownership of its content, AI models and related data. In other words, SpaceX is providing the infrastructure, but Google will maintain control over the AI systems and information running on it.

SpaceX Is Becoming a Major AI Infrastructure Provider

The Google agreement arrives shortly after SpaceX secured another significant compute deal with Anthropic, the company behind the Claude AI assistant.

Anthropic agreed to use the full computing power of SpaceX’s Colossus 1 facility in Memphis, Tennessee. The data centre houses more than 220,000 Nvidia processors and is expected to provide Anthropic with 300 megawatts of additional capacity. Reuters estimates that SpaceX’s compute agreements with Google and Anthropic are worth around $26 billion annually when combined.

The timing is also notable. SpaceX is preparing for a major US stock market debut, and the Google deal gives the company another substantial source of potential revenue outside its more familiar rocket-launch and satellite-internet businesses.

SpaceX has not disclosed which specific data centre Google will use. However, the agreement makes one thing clear: the race for AI compute capacity is becoming expensive enough that even Google is willing to spend close to $1 billion per month to secure additional hardware.

FAQ

How much will Google pay SpaceX for AI computing capacity?

Google will pay SpaceX $920 million per month from October 2026 through June 2029, following an initial ramp-up period with reduced fees.

How many Nvidia GPUs will Google receive access to?

The agreement includes access to approximately 110,000 Nvidia GPUs, along with CPUs, memory and other related hardware.

Why is Google renting computing capacity from SpaceX?

Google said the additional capacity will help it meet stronger-than-expected demand for Gemini Enterprise, its AI platform for businesses.

Written by

Share

Copied