The company that built the world’s most powerful AI infrastructure just admitted it isn’t enough. And it’s writing a $32 billion cheque to a rocket company to fill the gap.
In one of the most extraordinary business arrangements in tech history, Google has agreed to pay SpaceX about $920 million a month for access to compute capacity — leaning on the rocket company’s infrastructure rather than relying only on its own data centres. The deal, disclosed on June 5 via SpaceX’s IPO filing, is a staggering admission hidden inside a staggering number: even Google cannot build AI infrastructure fast enough to keep up with its own demand.
Table of Contents
The Deal: Every Number That Matters
| Detail | Figure |
|---|---|
| Monthly Payment | $920 million |
| Contract Start | October 2026 (reduced fees from now until Sept) |
| Contract End | June 2029 |
| Total Value | ~$32 billion |
| GPUs Covered | ~110,000 Nvidia GPUs |
| Also Includes | CPUs, memory, and related components |
| Infrastructure | xAI data centres (post SpaceX-xAI merger) |
| Exit Clause | 90 days’ notice after December 31, 2026 |
| GPU Delivery Deadline | September 30, 2026 (or Google can terminate) |
Why Google — Which Owns Massive Data Centres — Is Renting From SpaceX
This is the question that makes this story genuinely remarkable, and the answer cuts to the heart of the 2026 AI landscape.
A Google representative described the deal as a result of unexpected demand for its recently launched AI products: “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 numbers back that up. Google Cloud revenue rose 63% year over year in Q1 2026, and cloud backlog nearly doubled quarter over quarter to more than $460 billion. That backlog is the real story — Google has more committed future revenue than it currently has the infrastructure to serve. This arrangement lays bare a paradox at the heart of the AI boom: even the companies best equipped to build their own computing cannot build it fast enough.
For context on the broader AI infrastructure race, read our AI and technology coverage on TechnoSports.
The SpaceX-xAI Connection: How a Rocket Company Became an AI Landlord
The Google agreement marks the second massive infrastructure deal announced by SpaceX following its merger in February with xAI, Elon Musk’s artificial intelligence company, in a transaction that valued the combined entity at $1.25 trillion.
The xAI merger gave SpaceX something it never had before: a vast GPU data centre footprint originally built to train Grok, xAI’s AI model. That infrastructure — most notably the Colossus 1 facility near Memphis, Tennessee — is now being rented out to the highest bidders in the AI industry at eye-watering rates.
Anthropic agreed to pay SpaceX $1.25 billion per month through 2029 to rent all the available compute from its Colossus 1 data centre. Google’s deal appears to be paying for roughly half the amount of compute that Anthropic has access to at Colossus 1.
In the span of weeks, SpaceX has gone from rocket company to the most significant GPU landlord on Earth — renting compute to Google and Anthropic, two of the most direct competitors to its own xAI division. The strategic complexity of that arrangement is remarkable.
The Roles Have Completely Reversed
There is a delicious irony buried in the history here. In May 2021, Google Cloud and SpaceX announced a partnership to deliver data and cloud services at the network edge, with SpaceX locating Starlink ground stations within Google data centre properties. In that arrangement, Google was the infrastructure provider. Now Google is the buyer.
Five years ago, SpaceX needed Google’s infrastructure to run Starlink. Today, Google needs SpaceX’s infrastructure to run Gemini. The power dynamic has inverted completely — and it happened faster than almost anyone predicted.
The Protection Clauses: Google Isn’t Taking This on Faith
At $920 million per month, Google has negotiated hard exit ramps that reveal exactly how nervous the company is about Nvidia GPU delivery timelines.
If SpaceX fails to deliver the committed GPU capacity by September 30, 2026, Google may terminate after a one-month grace period, or accept the available chips with proportionally reduced payments. After December 31, 2026, either side can exit with 90 days’ notice. Those clauses suggest Google negotiated protection against a real risk: that the promised hardware does not arrive on schedule — a reflection of how tight the supply of advanced Nvidia chips has become.
The exit clauses are not boilerplate. They are a live acknowledgement that Nvidia GPU supply chains remain one of the most constrained and unpredictable variables in all of technology right now.
Alphabet Is Spending at a Scale That Defies Comprehension
This deal doesn’t exist in isolation. Alphabet has already committed to more than $180 billion in capital expenditures this year and has said it expects that to “significantly increase” in 2027. To help fund that, Alphabet recently announced an $80 billion equity sale.
To put $180 billion in annual capex in perspective: that is more than the entire GDP of many mid-sized economies, deployed in a single year, almost entirely toward AI infrastructure. The $32 billion SpaceX deal is one line item in that broader spending surge.
In jumping into the infrastructure leasing market, SpaceX is also competing with a group of companies commonly called neoclouds, that includes CoreWeave and Nebius — whose stocks got hammered on the day of the SpaceX-Google announcement before bouncing back. The message from the market: a SpaceX-scale competitor entering cloud compute is a threat worth pricing in immediately.
We recently covered the SpaceX IPO — the largest in history — where this compute business is now a central part of the company’s revenue story heading into its public market debut. Also worth reading: our breakdown of China’s CXMT reaching HBM3 — the memory that makes all these Nvidia GPUs actually function.
Sources: SpaceX Form S-1 Amendment No. 2 (June 3, 2026), TechCrunch, CNBC, Bloomberg
