Google has agreed to pay SpaceX $920 million per month from October 2026 through June 2029 for access to approximately 110,000 NVIDIA GPUs, CPUs, memory, and related components, according to a regulatory filing published Friday. The deal ranks among the largest compute rental agreements disclosed to date, and comes just one week before SpaceX stock is expected to begin trading on the Nasdaq exchange.
The arrangement is similar in structure and duration to one SpaceX announced with Anthropic in late May. Under that agreement, Anthropic committed to paying SpaceX $1.25 billion per month through 2029 to rent compute from one of the Colossus data centers near Memphis, Tennessee — facilities originally built by xAI, now part of SpaceX, for its own artificial intelligence work.
Also read: Android 17 arrives with bubble bar multitasking and deeper Gemini AI integration
IPO timing and financial context
SpaceX filed paperwork with the Securities and Exchange Commission indicating it aims to raise approximately $75 billion at a valuation of around $1.75 trillion, which would make it the largest initial public offering in history. Google, a longtime investor in SpaceX, holds a stake in the company that is expected to be worth more than $100 billion after the IPO, according to the filing.
The compute deal with Google further strengthens the financial relationship between the two companies ahead of the public listing. Google’s monthly payments will begin in October 2026, roughly three months after the earliest possible termination date.
Also read: Consumers say 'AI' in brand messaging is a turnoff, WordPress VIP survey finds
Cancellation clause and flexibility
Both SpaceX and Google retain the option to terminate the agreement with 90 days notice after December 31, 2026. The inclusion of a cancellation clause mirrors the Anthropic deal, suggesting SpaceX is structuring its compute rental contracts with flexibility for both parties as demand for AI infrastructure continues to evolve rapidly.
The agreement covers hardware components including CPUs, memory, and related equipment in addition to the 110,000 NVIDIA GPUs, though the filing did not specify which GPU models are included. NVIDIA’s H100 and upcoming B200 chips are the most common choices for large-scale AI training clusters.
Industry implications
The deal underscores the intensifying competition for AI compute capacity among major technology companies. Google, Microsoft, Amazon, and Meta have all announced significant capital expenditure increases for AI infrastructure in 2025 and 2026. By renting compute from SpaceX, Google gains access to dedicated GPU clusters without the lead times required to build its own data centers.
SpaceX’s Colossus data centers, originally developed by xAI, represent some of the largest concentrated AI compute installations in the world. The Memphis facility alone was reported to house over 100,000 GPUs during its initial buildout phase.
This story is developing. Check back for updates.
Be the first to comment