The story says Google signed a deal to pay SpaceX $920 million per month from late 2026 through mid-2029 for access to roughly 110,000 NVIDIA GPUs and related hardware at xAI data centers in Tennessee. That follows a larger Anthropic agreement for the same infrastructure. On its face, this looks bizarre. Google is supposed to be one of the companies with the deepest AI infrastructure stack in the world. SpaceX is supposed to be a space company. The filing also includes a big escape hatch. After this year, either side can walk away with 90 days notice.
The useful read is straightforward. Compute is still supply constrained, data center buildouts are delayed by power, permitting, and hardware bottlenecks, and xAI built a lot of capacity fast. Google is buying bridge capacity because it needs GPUs now, not because xAI suddenly won the model race. Many comments landed on the same blunt conclusion: this deal says more about idle xAI hardware and weak
Grok demand than about any new SpaceX moat. SpaceX is acting like a
GPU landlord because frontier model work is not soaking up the cluster.
Where the comments got sharper was on valuation. A lot of people saw the timing as the real story. SpaceX has been pitching itself as mostly an AI and infrastructure company rather than a launch business, and these contracts pad reported revenue right before an IPO that many commenters already think is detached from datacenter economics. Several people pushed back on the viral math that every extra dollar of revenue should be multiplied by the current price-to-sales ratio. That is not how valuation works. Still, the broader skepticism held: renting GPU capacity is a lower-margin, more competitive business than the IPO narrative implies, and the short cancellation window makes these agreements look more like temporary relief valves than durable proof of a trillion-dollar AI franchise.
A second strand focused on what the deals imply operationally. Google using outside NVIDIA clusters does not mean TPUs failed. It means demand outran available supply and companies will buy whatever compute they can get. The quoted price per GPU hour looked high to many readers, but others noted that speed, interconnect, and immediate availability matter more than sticker price in a shortage. The Memphis site’s power source also bothered people. Google has public carbon-free data center goals, while xAI’s facilities have drawn criticism for relying on trailer-mounted gas turbines. That gave the whole arrangement a more cynical cast: Google gets capacity without fighting the local political and environmental battles itself.
The dominant conclusion was not that the contract is fake. It was that two things can be true at once. Google probably does need the compute. SpaceX probably does want the revenue optics. The deal is economically real enough to matter in the near term, but it does not rescue the bigger claim that a company now leaning heavily on datacenter leasing deserves software-like multiples forever.