QUICK ANSWER
SpaceX and Google signed a Cloud Service Agreement on June 5, 2026. Google pays $920 million per month from October 2026 through June 2029 — approximately $30 billion total — for access to ~110,000 NVIDIA GPUs at SpaceX's Colossus data centers. Reduced fees during ramp-up through September. Terminable by either party with 90 days' notice after December 31, 2026. Google's stated reason: "bridge capacity" for surging Gemini Enterprise demand. Disclosed in SpaceX's S-1 Amendment No. 2, filed one week before SPCX begins trading on Nasdaq on June 12.
What the SEC Filing Actually Says
SpaceX disclosed the Google agreement in Amendment No. 2 to its Form S-1 registration statement, filed with the SEC on June 5, 2026 — the same day the deal was signed. The filing is a free writing prospectus under Rule 433, which means it is a legally binding disclosure obligation for a company in registration. The exact language from the filing:
"On June 5, 2026, we entered into a Cloud Service Agreement with Google LLC with respect to access to compute capacity. The compute capacity provided includes approximately 110,000 NVIDIA GPUs, CPUs, memory, and other related components. Pursuant to the agreement, the customer has agreed to pay us $920 million per month from October 2026 through June 2029, with capacity ramping up through September at a reduced fee. 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 pro rata reduction in the monthly fees. After December 31, 2026, the agreement may be terminated by either party upon 90 days' notice. The customer will retain ownership of, and intellectual property rights in, its content, AI models, and related data."
The timing is deliberate. Disclosing this agreement in an S-1 amendment one week before IPO pricing on June 11 is a strategic choice. New significant customer contracts must be disclosed to investors before they buy shares. The Google deal also materially strengthens the investment thesis for SPCX: SpaceX is no longer just a rockets and Starlink company — it is a commercial AI compute infrastructure company generating revenue from two of the world's largest AI labs simultaneously.
The Two Colossus Contracts Side by Side
| Metric |
Anthropic Deal |
Google Deal |
| Signed |
Announced May 6, 2026; S-1 disclosed May 20 |
June 5, 2026 (S-1 amendment same day) |
| Monthly fee |
$1.25 billion/month |
$920 million/month |
| Term |
Through May 2029 |
October 2026 through June 2029 |
| Total contract value |
~$40 billion |
~$30 billion |
| GPU count |
220,000 (Colossus 1, all capacity) |
~110,000 (roughly half) |
| Termination |
90 days' notice (either party) |
90 days' notice after December 31, 2026 |
| Data center |
Colossus 1, Memphis TN |
Not specified (Elon Musk has said Colossus 2 reserved for xAI) |
| Customer's stated reason |
Core model training and inference infrastructure |
"Bridge capacity" for surging Gemini Enterprise demand |
As TechCrunch noted, Google's deal appears to be paying for roughly half the amount of compute that Anthropic has access to at Colossus 1. SpaceX did not specify which data center Google would use. Elon Musk has previously suggested Colossus 2 would be reserved for xAI. If Colossus 1 (220,000 GPUs) is split between Anthropic and Google, that accounts for the full Colossus 1 capacity with two major tenants. The Mississippi facility under development and any future Colossus clusters would then be available for xAI or additional tenants.
Google's Explanation - Gemini Enterprise Demand
A Google Cloud spokesperson told Business Insider: "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." The "bridge capacity" framing is notable. Google is one of the world's largest data center operators - it spent $175-185 billion in capex in 2026. Needing to rent 110,000 GPUs from SpaceX as "bridge capacity" signals that Gemini Enterprise demand has outpaced even Google's own infrastructure buildout pace, which has been among the most aggressive in the industry.
The irony is not subtle. Google is paying Elon Musk's SpaceX - which includes the merged xAI business and its Grok models competing directly with Gemini - $920 million per month to run Gemini Enterprise workloads. The compute infrastructure market has reached a point where even the biggest hyperscalers are willing to rent capacity from competitors because the supply constraints are that severe. Google is simultaneously building its own data center in West Memphis, Arkansas, which will eventually provide owned capacity near the same geographic area as Colossus.
Why This Changes the SpaceX IPO Story
At $920 million per month from October 2026 through June 2029, the Google deal amounts to approximately $30 billion total. Combined with the Anthropic deal ($1.25 billion/month, ~$40 billion total), SpaceX's Colossus data centers are generating approximately $2.17 billion per month in contracted AI compute revenue from just two customers — an annualized run rate of roughly $26 billion from the AI infrastructure segment alone.
The SpaceX IPO was initially framed primarily as a Starlink story — the satellite internet business with 10.3 million subscribers and profitable operations. The Colossus contracts reframe the narrative entirely. SpaceX is now also a commercial AI infrastructure company with approximately $70 billion in contracted compute revenue across two long-term agreements. For institutional investors evaluating the $1.77 trillion IPO price, the Google deal is a material data point that strengthens the revenue visibility argument.
The risk disclosures have not changed. Both the Anthropic and Google contracts are terminable with 90 days' notice after December 31 and June 2026 respectively. Revenue concentration in two terminable contracts is a risk factor, not a guaranteed annuity. But the fact that SpaceX disclosed the Google deal in its S-1 amendment one week before IPO pricing — rather than waiting until after the IPO — means the deal's economics, including its termination provisions, are fully in the public record for investors to evaluate before buying shares on June 12.
What This Means for the AI Compute Market
The Google-SpaceX deal confirms a structural trend visible across all the major AI infrastructure stories of 2026: demand for GPU compute capacity is outpacing supply at every level of the market. ByteDance is planning $70 billion in AI capex. Amazon, Google, Microsoft, and Meta are collectively spending $500+ billion on data center infrastructure this year. SpaceX built Colossus 1 (220,000 GPUs) for xAI's training needs and is now renting it out at $2.17 billion per month because other AI labs need it more urgently than xAI does at this moment.
The price signal is the most important data point. At $920 million per month for 110,000 GPUs, Google is paying approximately $8,364 per GPU per month. At $1.25 billion per month for 220,000 GPUs, Anthropic pays approximately $5,682 per GPU per month. The higher rate for the smaller Google block likely reflects the shorter notice period (terminable from December 31 vs May 2029 for Anthropic) and the timing premium for immediate availability. These per-GPU rates represent the current spot market for large-scale AI compute in a supply-constrained environment.
Frequently Asked Questions
Are Anthropic and Google using the same data center?
Possibly. Anthropic's contract is for Colossus 1 specifically (220,000 GPUs near Memphis, Tennessee). Google's contract does not specify the data center. If Colossus 1 is being shared between Anthropic (220,000 GPUs) and Google (110,000 GPUs), that would imply overlapping allocations unless Anthropic's 220,000 GPU allocation has been revised downward. SpaceX has not clarified the data center assignment. Elon Musk has stated Colossus 2 is reserved for xAI.
Does this affect the SpaceX IPO price?
The deal was disclosed in the S-1 amendment, so it is already reflected in the information available to investors before June 11 pricing. At $920 million/month from a second major AI customer, the deal materially increases SpaceX's visible contracted revenue and strengthens the AI infrastructure narrative. Whether it changes the IPO price range depends on underwriter book-building conversations happening now. The fixed $135/share price was set before this deal was disclosed - expect the underwriters to gauge whether investor demand shifts in response.
Why would Google rent compute from SpaceX/xAI instead of building its own?
Google Cloud's spokesperson answered this directly: demand for Gemini Enterprise has exceeded expectations. Building a new data center from site selection to full GPU deployment takes 18-36 months. Renting from an existing facility with NVIDIA GPUs already installed takes weeks. The "bridge capacity" framing means Google is using SpaceX's Colossus infrastructure to meet near-term demand while its own construction pipeline catches up. Google is building a data center in West Memphis, Arkansas - near the Colossus Memphis facilities - which will provide owned capacity when it comes online.
Is there a conflict of interest in Google renting from SpaceX/xAI?
Yes, and it is visible. SpaceX now includes the merged xAI business whose Grok models compete directly with Google's Gemini. Google is paying SpaceX $920 million per month to run Gemini workloads on infrastructure that also hosts xAI's competing AI development. Both parties apparently concluded the compute access value outweighs the competitive conflict — and given Gemini Enterprise demand exceeding Google's own capacity, that calculus is straightforward from Google's perspective.