Patel says Nvidia's biggest risk is competitors catching up with Huang gone

Huang met every Nvidia supplier quarterly to woo them, Patel says

Nvidia has long been the dominant semiconductor company; its chips are the most in-demand in the world, and the company was for a time the most valuable in the world. With Nvidia’s leader, Jensen Huang, out, NPR’s Scott Detrow asked Dylan Patel, founder and CEO of the AI and semiconductor research firm SemiAnalysis, whether that leaves an opening for Nvidia’s competitors to catch up.

Detrow noted that Patel and others had estimated Nvidia is probably a year, two years ahead of its competitors, and asked whether other companies can catch up with Huang gone. “I think that’s the biggest risk at least,” Patel replied.

Over the last handful of years, Nvidia’s lead has actually lessened, Patel said, and the company has lost market share, albeit the market has grown so fast that Nvidia’s revenue and profits have skyrocketed. “Now, with, you know, many barbarians at the gate, when sort of Jensen is needed most, he sort of is disappearing,” Patel said.

Detrow asked whether Huang was needed most by the company or by the industry. Most by the company, Patel said: the industry has for a long time tried to roll its own solution, at least among the biggest customers. Google, Microsoft, Meta and Amazon all have been working on their own AI chips for a number of years, and yet they all still relied on Nvidia, though their reliance is reducing to some extent — a goal Patel said is theirs and has been their goal. “The only way Nvidia can keep their market share and stay on top of the world is if they out execute everyone,” Patel said, meaning they must keep innovating and bringing new technology to the market. “But without Jensen, you know, the founder and the visionary of the company, it’s going to be a lot harder for them to keep on top.”

Detrow said he sometimes feels a CEO’s role in a company is overblown, then turned to the supply side, asking how Huang’s personal relationships with supply chain managers in Korea and Taiwan were so key to Nvidia getting what it needed. “There’s no other company in the supply chain, including in the semiconductor supply chain, where the CEO is, every quarter, meeting with all his suppliers,” Patel said. Chips are such a complex supply chain, Patel said, and Huang has hundreds of suppliers, talking to every single one of them — going to Korea, going to Taiwan, going to China, meeting with them, having drinks with them, going to local fried chicken spots or night food markets, “everything to woo them.” Given an industry where products are improving by 2x a year, “at an incredible pace,” Nvidia needs its suppliers to have full focus and put their best engineers on Nvidia’s product, Patel said, and “this just keeps Nvidia ahead of everyone.”

Detrow called the really risky big bets Jensen Huang took a big part of the reason the company dominated the field, and asked whether Patel saw another CEO taking those bets, or another company trying to establish itself taking the types of bets he took. Patel said there are many visionaries within the company, but the challenge is whether any are willing to bet the entire company. Huang developed CUDA — “the software that makes Nvidia what it is” — in 2010, when everyone who was a technologist at the time and everyone who was on Wall Street at the time thought it was a waste of money and a waste of effort, and it turned out to be the bedrock of all AI software, Patel said. “Is someone going to do something that is such a waste of money that, for a decade looks like a terrible bet?” Patel asked. “It really takes a visionary, and I don’t know if too many people will make those kinds of bets.”

If someone asked whether Nvidia without Jensen a decade from now would be the one bringing the most important new technology to market, Patel said: “I’d probably wager no.”

Detrow asked whether Patel had any sense of what some of Nvidia’s next big bets could be. Patel described the direction Huang had been steering Nvidia toward: the company started making chips, then started making servers, then started making systems, and now is really focused on the end applications — helping companies make medical devices, satellites and entire systems that include AI as a part of it. If Nvidia does not do this, most small and medium-sized healthcare companies, and even many companies that are large but not the largest, can’t afford to build their own systems out of AI, Patel said.

That, Patel said, is where Jensen was heading: in automotive, in healthcare, in manufacturing, making it so medium-sized and small companies can all start deploying AI throughout their businesses to make them more efficient, increase production, make their products better and make people healthier. Without that, only the biggest companies can do this, and “the profit motive sort of makes it so, you know, all of the profit is captured by a few, not diffused to the whole ecosystem.”

Detrow closed the segment by identifying his guest as Dylan Patel, founder and CEO of SemiAnalysis, and thanking him for talking through the question.