NVIDIA Backs 5GW AI Factories

SpaceXAI’s 220,000-GPU supercluster, reportedly built in just 19 days, has set the industry on fire, showing just how quickly AI infrastructure can scale—if you have the muscle. Meanwhile, NVIDIA’s five-gigawatt “AI factory” partnership with IREN signals a new phase: the company isn’t just selling chips, it’s shaping how and where the world’s biggest AI projects get built, starting with a massive campus in Texas. The stakes? Billions in investment, and a race to see who can deliver capacity before the grid, cooling tech, and network connections hit their limits.

But here’s the catch: none of these mega-deals are guarantees. NVIDIA’s agreement with IREN is an “intent,” not a done deal—the real test is whether equipment, power, and software all arrive on time. At the same time, NVIDIA is making a major software play, rolling out Fleet Intelligence and NCCL Inspector for real-time fleet monitoring. The move is designed to lock in customers by making NVIDIA hardware harder to swap out, but raises new governance and procurement questions as more companies rely on these tools to wring performance from their infrastructure.

Featuring insights from Jensen Huang and reporting from HotHardware, this episode unpacks how NVIDIA is becoming the heartbeat of global AI—if it can keep up with the breakneck pace, shifting alliances, and escalating power needs.

Powered by Apisod.com