The overhyped showdown between the US and China for ‘AI supremacy’ (whatever that means), usually features cutting-edge chips like Nvidia’s H100 GPUs—the silicon brains powering models like ChatGPT. But as even China-hating tech investor Kyle Bass admits, “AI chip cost is a tiny portion of AIDC (AI data center) buildout cost.” America (whose entire advantage in chip innovation stems from ethnic Chinese!) won’t win if Washington ignores the bigger picture: the U.S. cannot compete with China on the cost of delivering exaFLOPS1 of AI computing power.
AI data centers aren’t just warehouses of humming servers; they’re colossal engineering feats. A single hyperscale facility can cost $2 billion and span football fields, guzzling 100 MW, as much electricity as a small city. Chips make up only 10–20% of that tab—say, $100–200 million for a cluster delivering 100 EFLOPS. The rest?
