Google and SpaceX seal $30B AI compute deal through 2029

This deal cements compute as the choke point in the race to AGI. Google is effectively renting a second hyperscale backbone from SpaceX/xAI, locking in access to roughly 110,000 high‑end Nvidia GPUs at a time when everyone from startups to sovereigns is scrambling for capacity. For Alphabet, framing this as “bridge capacity” for Gemini Enterprise signals two things: demand for agentic AI is outpacing its own data center build‑out, and management is willing to outsource core infrastructure rather than lose enterprise workloads to rivals.

For SpaceX and its xAI division, the agreement validates a strategic pivot from pure launch and connectivity into selling raw intelligence infrastructure. Pre‑IPO, a contracted revenue stream on the order of $30 billion gives public investors a clean, AI‑linked growth narrative that doesn’t rely on launching its own competitive foundation models. More broadly, this is a preview of a world where a few capital‑intensive players own the physical substrate of AI — orbital and terrestrial — while model labs become, in effect, large tenants. That concentration of compute makes the frontier AI race faster, but also more brittle: if a handful of infrastructure landlords go down, or realign politically, a huge chunk of global AI capacity moves with them.