Google will pay SpaceX 920 million dollars a month for computing power - and to Elon Musk's company, a week before it goes public. The deal was disclosed in a regulatory filing on Friday and runs from October 2026 to June 2029. For that money Google gets access to around 110,000 NVIDIA graphics processors plus other equipment.

Let's translate what this means. Google - the company considered the world's single biggest owner of AI infrastructure - is renting someone else's servers. Why? Because, as it admits itself, demand caught it off guard. "This is a short-term agreement to provide bridge capacity for the surge in demand for our Gemini Enterprise platform, which is even higher than we expected," the company said. When even the biggest player has to rent, that says something about how hungry this industry is for electricity and silicon.

The deal resembles the one SpaceX struck with Anthropic at the end of May - there the figure was 1.25 billion dollars a month for the entire capacity of the Colossus 1 data center near Memphis, the one xAI (now part of SpaceX) originally built for itself. Google gets roughly half of what Anthropic takes. The same farm, two hungry clients, and an owner billing both sides.

The context behind all this is spending that's hard to grasp. Alphabet, Google's parent, has pledged over 180 billion dollars in capital expenditure this year alone, with "significantly" more in 2027. To fund it, it recently sold 80 billion in shares. This isn't a thrifty company counting every euro - this is a race where the stake is not to fall behind, at any cost.

And here's the point. SpaceX goes public in a week aiming to raise around 75 billion dollars at an estimated valuation of 1.75 trillion - the largest in history. Google, a long-time SpaceX investor, will hold a stake worth over 100 billion after the listing. So Google is paying billions for servers from a company in which it already holds billions. The money spins in a circle among a handful of names, and the electricity bill is paid by all of us through the price of every AI tool we use. Who ends up earning the most from this race - it's not hard to guess.