Google pays SpaceX $920M monthly for xAI compute

According to a SpaceX regulatory filing reported by TechCrunch and CNBC, Google will pay SpaceX $920 million per month from October 2026 through June 2029 for access to roughly 110,000 NVIDIA GPUs, plus CPUs, memory, and related components. The 32-month agreement ramps up through September 2026 at a reduced fee. The filing lets Google terminate, or accept fewer GPUs at a lower fee, if SpaceX misses a September 30, 2026 delivery milestone, and lets either party cancel with 90 days' notice after December 31, 2026, per TechCrunch. Google said the deal provides short-term bridge capacity for its Gemini Enterprise agent platform. The compute sits at facilities operated by SpaceX, which absorbed xAI in a February 2026 merger, and the deal follows a larger late-May arrangement under which Anthropic agreed to pay SpaceX $1.25 billion per month for Colossus 1 capacity.
What happened
According to a SpaceX regulatory filing reported by TechCrunch and CNBC, Google will pay SpaceX $920 million per month from October 2026 through June 2029, about 32 months, for access to "approximately 110,000 NVIDIA GPUs, CPUs, memory, and other related components." Google's access ramps up through September 2026 at a reduced fee. The companies disclosed the agreement roughly one week before SpaceX's planned IPO.
Deal terms
The filing includes operational protections for Google. As TechCrunch quotes it: "If we fail to deliver access to the committed amount of GPUs by September 30, 2026, then following a one-month grace period, Google may immediately terminate the agreement or accept the number of GPUs provided," with a corresponding reduction in fees. Either party may also terminate with 90 days' notice after December 31, 2026. The arrangement mirrors, at smaller scale, a late-May deal in which Anthropic agreed to pay SpaceX $1.25 billion per month through 2029 for all available compute at the Colossus 1 data center near Memphis, capacity originally built by xAI, now part of SpaceX. TechCrunch notes Google's contract covers roughly half the compute Anthropic secured, and that SpaceX did not specify which data center Google would use.
Why it matters
Time-limited contracts for external GPU capacity have become a common way for cloud providers and AI labs to bridge supply gaps without immediate capital outlay. What is notable here is the buyer: Google is among the world's largest owners of AI compute, yet still chose to lease. In a statement reproduced by TechCrunch, Google said: "Google Cloud and SpaceX are long-time partners. This is a short-term, timely agreement to ensure we have bridge capacity to meet surging customer demand for our agent platform, Gemini Enterprise, which has been even higher than we expected." For practitioners, leasing externally hosted GPU pools changes scheduling, data-locality, and cost assumptions relative to self-managed or long-term reserved capacity.
The IPO backdrop
The deal lands as SpaceX prepares what filings frame as the largest IPO in history, reportedly seeking to raise around $75 billion at a roughly $1.75 trillion valuation. Google is a longtime SpaceX investor whose stake is expected to exceed $100 billion after the offering, and the two are reported to be exploring orbital data centers. The compute agreements with Google and Anthropic give SpaceX large, recurring revenue lines from xAI-built infrastructure just as it markets itself to public investors.
What to watch
Key signals include whether SpaceX meets the September 30, 2026 GPU-delivery milestone that triggers Google's termination rights, the pace of the ramp into October, and whether more hyperscalers and labs sign similar short-term, high-dollar bridge contracts. Follow-on regulatory filings and any statements from the parties should clarify delivered capacity and which facilities are involved.
Scoring Rationale
This is a major, roughly $29 billion infrastructure commitment, disclosed in a regulatory filing and corroborated by TechCrunch and CNBC, that reshapes GPU-supply and cloud-capacity dynamics for large AI workloads. Its significance is amplified by the buyer being Google, among the largest compute owners, and by its timing one week before SpaceX's record IPO. The score sits in the major band, just below industry-shaking, reflecting verified terms and broad relevance to practitioners planning large-scale training and inference.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems

