Nvidia

Nvidia is a company that makes chips for running AI. It was founded in 1993 by three people, including Jensen Huang, and was originally known for gaming graphics cards. Those graphics cards turned out to be unusually well suited to AI computation, which is how the company got to where it is today.

It's easy to mistake Nvidia for a competitor to ChatGPT or Gemini, but it doesn't play in that fight. In the gold rush metaphor, it's not the one panning for gold — it's the one selling pickaxes and jeans. OpenAI, Google, and Meta all buy Nvidia chips to train their models.

That said, Nvidia does release its own AI models, like Nemotron and Cosmos. But these aren't attempts to run a chatbot business — they're closer to demos showing off "what you can build with our chips." It's fair to see them as models built to sell more chips.

Two scenes from 2026 capture Nvidia's standing well. It brought in the founder and core team of Groq, an inference-chip company that had been challenging it head-on, through a $20 billion licensing deal. And it guaranteed OpenAI 4.25 gigawatts of power supply at an Ohio site for 20 years. Absorbing a rival's brainpower while guaranteeing a customer's growth — that's a position well beyond just selling pickaxes.

"Nvidia unveiled a new AI model" — this usually doesn't mean a ChatGPT competitor has arrived, but that a demo model has been released to showcase the performance of its own chips.