Nvidia Is Infrastructure, Not Just a Chip Company
A piece calling Nvidia 'the central bank of AI' got traction, and the HN discussion immediately made it more interesting. The key pushback: a real central bank can expand supply monotonically. Nvidia cannot. TSMC's capacity constraints mean Nvidia is more like a central bank that cannot print money fast enough during a crisis. Someone asked what TSMC is in the analogy, and nobody had a clean answer.
A separate thread on Nvidia's claim that every dollar it invests returns $100 got the reaction it deserved: skepticism. The circular financing dynamic, where Nvidia invests in AI companies that then buy Nvidia GPUs, is real and people are noticing it. One commenter simply wrote 'are those $100 in the room with us right now.'
The pattern here is that Nvidia's structural position in the AI economy is becoming legible to a broader audience. It is not just a semiconductor company. It sets the pace, the price, and indirectly the strategy of every AI company that depends on H100s and B200s. That is a different kind of power than market share.
So what?
Any startup whose unit economics depend on GPU availability or pricing is exposed to Nvidia's supply decisions in a way that resembles exposure to a central bank rate decision. Founders should model GPU cost sensitivity explicitly and track alternative inference providers, including AMD and custom silicon efforts, as a hedge.