Blab Tech
Google to Pay SpaceX $920 Million a Month for Compute Power Through 2029
In a deal that could reshape the AI hardware race, Google will pay Elon Musk’s SpaceX $920 million each month for access to compute resources from October 2026 through June 2029, according to a regulatory filing released on Friday. The contract, the second such agreement SpaceX has entered with an AI competitor in weeks, totals roughly $30 billion.
Under the terms, SpaceX will provide Google with about 110,000 Nvidia graphics‑processing units, along with central‑processing units, memory chips and related components. If SpaceX fails to deliver the Nvidia chips by September 30, Google may terminate the contract with a one‑month grace period. Either party can also end the arrangement with 90‑day notice, a clause mirrored in SpaceX’s earlier deal with Anthropic.
Google’s spokesperson said the deal is intended to supply “bridge capacity” for the company’s rapidly growing AI services. In its most recent earnings report, Alphabet reported that Google Cloud’s backlog—work that has been contracted but not yet recognized as revenue—had nearly doubled from the prior quarter to more than $460 billion. The spokesperson added that demand for Gemini Enterprise, Google’s enterprise‑grade AI platform, had exceeded expectations.
SpaceX’s involvement in AI infrastructure follows a series of moves aimed at monetizing its data‑center capabilities. The company’s xAI subsidiary, formed after the February merger of SpaceX with the AI firm, has been positioning itself as a compute‑infrastructure provider. xAI’s facilities in Memphis, Tennessee, and Mississippi are expanding, and the company has secured a deal with the startup Cursor that gives it the right to acquire Cursor for $60 billion or pay a $10 billion breakup fee.
The technical scope of the Google‑SpaceX partnership is significant. The 110,000 Nvidia GPUs, combined with CPUs and memory, could represent well over 100 megawatts of computing power when using Nvidia’s H200 chips. That amount of power is enough to energise roughly 75,000 homes at any given moment.
The relationship between Google and SpaceX is both collaborative and competitive. Earlier this year, SpaceX disclosed that Google owned a 6.11 percent stake in the company at the end of 2025. After SpaceX’s merger with xAI, Google’s share is estimated to be around 5 percent, according to Bloomberg calculations. In addition to the compute deal, the two companies have discussed launching Google’s test products for orbital data centers. A person familiar with the matter said in May that Google was exploring deals with other launch providers for its Project Suncatcher initiative.
The deal follows SpaceX’s recent agreement with Anthropic, which also involved renting compute capacity to an AI firm. SpaceX’s strategy appears to be to secure long‑term revenue streams from AI workloads while it prepares for a potential initial public offering later this year.
At present, the contract is in effect, with payments scheduled to begin in October 2026. Google has not yet disclosed the exact pricing structure beyond the monthly fee, and the status of the Nvidia chip delivery remains to be confirmed. The partnership will likely influence the competitive dynamics of AI infrastructure, as both companies seek to leverage their respective data‑center assets to meet the surging demand for large‑language‑model training and inference.
The agreement represents a significant financial commitment from Google to secure external compute resources and underscores the growing importance of dedicated AI hardware in the cloud services market. SpaceX’s expansion into data‑center operations and its stake in Google further intertwine the two companies’ interests as they navigate the evolving landscape of AI infrastructure.
Under the terms, SpaceX will provide Google with about 110,000 Nvidia graphics‑processing units, along with central‑processing units, memory chips and related components. If SpaceX fails to deliver the Nvidia chips by September 30, Google may terminate the contract with a one‑month grace period. Either party can also end the arrangement with 90‑day notice, a clause mirrored in SpaceX’s earlier deal with Anthropic.
Google’s spokesperson said the deal is intended to supply “bridge capacity” for the company’s rapidly growing AI services. In its most recent earnings report, Alphabet reported that Google Cloud’s backlog—work that has been contracted but not yet recognized as revenue—had nearly doubled from the prior quarter to more than $460 billion. The spokesperson added that demand for Gemini Enterprise, Google’s enterprise‑grade AI platform, had exceeded expectations.
SpaceX’s involvement in AI infrastructure follows a series of moves aimed at monetizing its data‑center capabilities. The company’s xAI subsidiary, formed after the February merger of SpaceX with the AI firm, has been positioning itself as a compute‑infrastructure provider. xAI’s facilities in Memphis, Tennessee, and Mississippi are expanding, and the company has secured a deal with the startup Cursor that gives it the right to acquire Cursor for $60 billion or pay a $10 billion breakup fee.
The technical scope of the Google‑SpaceX partnership is significant. The 110,000 Nvidia GPUs, combined with CPUs and memory, could represent well over 100 megawatts of computing power when using Nvidia’s H200 chips. That amount of power is enough to energise roughly 75,000 homes at any given moment.
The relationship between Google and SpaceX is both collaborative and competitive. Earlier this year, SpaceX disclosed that Google owned a 6.11 percent stake in the company at the end of 2025. After SpaceX’s merger with xAI, Google’s share is estimated to be around 5 percent, according to Bloomberg calculations. In addition to the compute deal, the two companies have discussed launching Google’s test products for orbital data centers. A person familiar with the matter said in May that Google was exploring deals with other launch providers for its Project Suncatcher initiative.
The deal follows SpaceX’s recent agreement with Anthropic, which also involved renting compute capacity to an AI firm. SpaceX’s strategy appears to be to secure long‑term revenue streams from AI workloads while it prepares for a potential initial public offering later this year.
At present, the contract is in effect, with payments scheduled to begin in October 2026. Google has not yet disclosed the exact pricing structure beyond the monthly fee, and the status of the Nvidia chip delivery remains to be confirmed. The partnership will likely influence the competitive dynamics of AI infrastructure, as both companies seek to leverage their respective data‑center assets to meet the surging demand for large‑language‑model training and inference.
The agreement represents a significant financial commitment from Google to secure external compute resources and underscores the growing importance of dedicated AI hardware in the cloud services market. SpaceX’s expansion into data‑center operations and its stake in Google further intertwine the two companies’ interests as they navigate the evolving landscape of AI infrastructure.
