Source: Content compiled from fortune

Nvidia is a $3 trillion behemoth in the world of AI chips. But the massive impact Nvidia has brought to the market hasn't scared off a host of challengers, from established chipmakers to young startups.

"The market needs competition, and that's why we're so attracted to it," Mark Papermaster, executive vice president and chief technology officer of Advanced Micro Devices, said at Fortune's Brainstorm AI conference on Tuesday.

AMD started out decades ago as a maker of microprocessors for PCs, but has also developed a strong business designing and selling GPUs, AI-friendly chips that are at the heart of Nvidia's booming business. While AMD's GPUs were once aimed at video games and entertainment applications, the company is now targeting the AI training market directly.

"We went from almost zero AI revenue in 2023 to an estimated $5 billion this year. The market needs competition, and that's what we're focused on," Papermaster said.

Andrew Feldman, co-founder and CEO of Cerebras Systems, believes there are so many players because the AI market is so big. It's not just Nvidia or AMD, but hyperscalers or startups "seeing the opportunity to explode and wanting to develop their own chips."

"It's because of the surge in demand," Feldman said. "We see it on the training side, we see it on the inference side - and I don't think any of us can keep up, to be honest."

Feldman says about 

Feldman went on to say that competition is good not only for competitors but also for consumers, because it means the AI in their applications will run faster and cheaper. "Mark's GPU runs faster than Nvidia's," he said, referring to AMD's Papermaster. "Our accelerators run 75 times faster than Nvidia's in many cases," he claimed.

Speed matters. Feldman referred to the internet era before broadband, calling it a disaster.

"Once we had broadband, all of a sudden you had new applications, you had streaming, you had all this fun stuff, and engagement was high," he said. "I think that's what's happening with AI right now, as speeds go up, you're going to get into the broadband era of AI inference."

It's worth noting that generative AI is still in its early stages. "There's a lot of innovation ahead of us," said Altera CEO Sandra Rivera. For Rivera, it's not about taking on Nvidia but about building the best device for what you want to do, because chips aren't one-size-fits-all. For example, as AI models and applications evolve into different versions, the processing power needed to run it will reside on "edge" devices like smartphones and home appliances, as well as inside data centers.

"Competition is good," she said. "It makes you sharper, it makes you more focused, and it certainly addresses a very broad range of customer needs."

Market for AI chips

The demand for AI chips is also creating opportunities for investors. Cerebras filed paperwork for an initial public offering in September, but Feldman declined to comment on the status or timing of the listing. Rivera said Intel-owned Altera still plans to go public in 2026. Intel acquired Altera in 2015 but has been beset by business challenges that led to the resignation of CEO Pat Gelsinger last week. Despite the changes at Intel, Rivera said Intel still plans to sell its stake in Altera and conduct an IPO in 2026. "We are in the middle of that process and there is a lot of interest, which is encouraging," she said, later adding, "Last week's news was sad on many levels, but I'm trying to keep the company and employees focused on the things we control."

View more at EASELINK