Jensen Huang says Nvidia is happy selling chips to open and closed AI
Nvidia does not mind who wins the AI race.

The short version
- CEO Jensen Huang says their chips power both sides.
- They just sent 25,000 graphics processors to Fireworks.ai.
- That helps companies train specialized models on their own servers.
- Your next software tool might run on cheaper custom hardware.
Why it matters to you. For your business, the race between open and closed models means cheaper, more specialized artificial intelligence tools are coming to your desk much sooner than you think.
Nvidia, the massive chipmaker whose graphics cards run the artificial intelligence boom, is happy to take your money no matter which software you use. According to The Register, Nvidia Chief Executive Jensen Huang said the company does not care whether open or closed AI models win the market as long as Nvidia hardware powers the infrastructure behind them.
Speaking at the Forge event in San Francisco on Tuesday, Huang sat down with Lin Qiao, the head of Fireworks.ai, which is a platform for companies to train and run custom software models. Fireworks just received a massive delivery of 25,000 Nvidia graphics processing units to support its expansion across Europe.
Huang explained that while closed models have their place, specialized intelligence trained on specific domains of knowledge holds most of the world's actual value. He noted that taking an open model and adapting it with your own domain knowledge creates a powerful tool for any business.

The conversation also touched on the soaring costs of running artificial intelligence, with companies warning that unchecked spending is getting out of hand. Genspark, an artificial intelligence workspace run by Kay Zhu, recently built a slide generator that matches top-tier quality at a fraction of the cost, charging 30 cents per million tokens instead of $5.
Looking ahead to next year, Huang predicted that agent harnesses and robotics systems will be the next major industry breakthroughs.
Photo: NVIDIA




