The biggest companies on Earth are at war over AI — OpenAI, Google, Meta, Anthropic, xAI — and nearly all of them buy their weapons from one supplier: Nvidia. A company that almost went bankrupt three times.
It began in 1993 at a Denny's. Its first chip flopped; to survive, Jensen Huang flew to Japan, told Sega the chip they'd paid for didn't work — and asked them to pay anyway. They did. Then, in 2006, Huang made a bet that nearly broke the company on purpose: CUDA, software that turned gaming chips into general computers. It crushed profits, the stock collapsed, and Wall Street begged him to stop for a decade. He refused. In 2012, two of those gaming cards trained the AI that started the deep-learning revolution — and the lonely bet became the only game in town.
The genius wasn't the chip; it was CUDA, the software every researcher learned and couldn't leave. When ChatGPT hit, Nvidia touched $4 trillion — the most valuable company in history. In a gold rush, sell shovels. Nvidia sells the shovels to everyone.
(The one caveat: a lot of that demand runs in a circle — Nvidia invests in OpenAI, which buys Nvidia chips. Boom, or bubble? No one knows yet.)