SpaceX just locked in another massive compute deal, and this time the customer is Google. According to TechCrunch AI, the company disclosed the agreement in a regulatory filing on Friday: 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 hardware. The timing is no accident. SpaceX revealed the deal just one week before its stock is expected to start trading on the Nasdaq.
That’s the headline number. What stands out here is who’s buying.
Why Google renting compute is the real story
Google isn’t some compute-starved startup. Some estimates name it as the world’s largest single owner of AI compute. So when a company that builds its own data centers at planetary scale starts renting capacity from a rocket company, that tells you something about how tight the market really is.
Google’s own explanation, as reported by TechCrunch AI, is demand it didn’t see coming. “Google Cloud and SpaceX are long-time partners,” the company said in a statement. “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.”
Read that again. Even Google is scrambling for “bridge capacity.” The AI agent boom is outrunning the infrastructure built to serve it, and the biggest players are paying premium rent to keep up.
How this compares to the Anthropic deal
This is the second giant SpaceX compute deal in weeks. Here’s how they line up:
- Anthropic (late May): $1.25 billion per month through 2029, renting all available compute from the Colossus 1 data center near Memphis, the facility xAI originally built for itself before becoming part of SpaceX.
- Google (now): $920 million per month, for roughly half the compute Anthropic has access to at Colossus 1.
The difference in motive matters. Anthropic was genuinely capacity-constrained, and raised its usage limits the same day that deal landed. Google is coming from a position of strength and just needs a stopgap. Same supplier, very different reasons.
SpaceX didn’t say which data center Google will use. CEO Elon Musk has previously signaled he’d reserve the Colossus 2 facility for xAI.
The fine print
A few terms worth flagging:
- Cancellation clause: Both sides can walk with 90 days’ notice after December 31, 2026. Neither party is locked in for the full term.
- Ramp-up: Google’s access scales up through September at a reduced fee.
- Delivery penalty: If SpaceX fails to deliver the committed GPUs by September 30, 2026, then after a one-month grace period Google can terminate outright or accept fewer GPUs at a reduced monthly fee.
That delivery clause is telling. It means even SpaceX isn’t certain it can stand up 110,000 GPUs on schedule, and Google wrote itself an exit just in case.
The money behind the money
Google’s parent, Alphabet, is on a spending tear. It has already committed to more than $180 billion in capital expenditures this year and expects that to “significantly increase” in 2027. To fund it, Alphabet recently announced an $80 billion equity sale. Renting from SpaceX is one more line in a very expensive year.
For SpaceX, the timing is everything. The company is aiming to raise around $75 billion at a valuation near $1.75 trillion, which would make it the largest IPO in history. Stacking up recurring revenue from Google and Anthropic right before the bell is a strong look for investors. Google, for its part, is a longtime SpaceX backer, and its stake is expected to be worth more than $100 billion after the IPO.
What to watch next
Two things. First, whether these short-term “bridge” deals become permanent fixtures, because nothing in AI infrastructure has stayed temporary lately. Second, the bigger swing: SpaceX and Google are reportedly in talks to build orbital data centers, a centerpiece of SpaceX’s post-IPO plans. Compute in space sounds far off, but a month ago so did Google renting GPUs from a rocket company.
For the full breakdown, check the original report at TechCrunch AI.