Eric Schmidt: China Is Now Just Six Months Behind the U.S. in AI

The U.S. lead over China in artificial intelligence has collapsed from roughly two years to six months, according to Eric Schmidt, former Google CEO and current CEO of Relativity Space. Schmidt, who helped design the chip export restrictions meant to hold Beijing back, delivered the assessment in a live interview with Robert Safian at the Masters of Scale Summit 2026, recorded for the Rapid Response podcast.

"Now it looks like they're six months behind, and there are many studies that show this," Schmidt said. The chip limits, he admitted, did not hold. "I assumed they would work for about 10 years."

Schmidt's explanation is blunt: Chinese engineers found a way to run frontier-class algorithms on far less compute. "They're so smart they figured out a way to do the kind of algorithms we develop, and use the knowledge we develop, with essentially one-tenth of the energy and, in some factors, less performance," he said.

What changed in the past year?

Schmidt pointed to collapsing costs. "The cost of software has gone from very expensive to almost nothing," he said, predicting an explosion of agents and software built on top of that cheap foundation. He framed the U.S.-China contest as "the defining competition of your generation, my generation, of our age," because both countries are chasing systems with superhuman reasoning. His example: Waymo's autonomous cars already see better than human drivers, "especially because they don't drink."

The stakes, in his words, are generational. "The race for intelligence is literally the race of a millennium," he said, arguing that decisions made now will shape the lives of today's children and grandchildren.

Where does China hold the advantage?

Energy. Schmidt said China has "infinite energy, most of it renewable," and none of the seven-year interconnection queues that stall U.S. grid projects. He noted China is building data centers "massively." Chinese capital markets, though more restricted, still fund startups: DeepSeek, he said, "just did a huge raise."

One detail stands out. Schmidt said U.S. investors are barred from participating. "It's illegal for us to invest in DeepSeek. I know because I tried," he said.

The U.S., by contrast, has talent and money but not power. Schmidt cited estimates that the country needs 60 to 80 gigawatts of additional capacity to get through 2030 or 2031. At roughly $50 billion per gigawatt — about one nuclear plant's worth — the implied buildout costs $3 trillion to $4 trillion. "The last time I checked, we weren't building them at a very high rate," he said.

On funding, he was wry: "Where in the world could you get $50 billion on a text message?"

Is open source the problem?

Schmidt said China's open-weight strategy was rational — a way for a trailing player to win users fast — and it is working, as DeepSeek's fundraising shows. American alternatives exist: he cited Reflection AI's newly released 500-billion-parameter model, Nvidia's Nemotron, and Google's Gemma 4, all downloadable.

But he flagged what he called an unsolvable security gap. Closed models are controlled by companies subject to national law. Open-weight models carry guardrails, but a technique called "obliteration" strips them out, leaving a raw model usable for cyberattacks. "I don't fully understand how we're going to navigate through this open-source, open-weight stuff when it gets into the hands of terrorists," Schmidt said.

Can both countries win?

Schmidt rejected the zero-sum framing. "China is not our enemy," he said, noting that competitors sometimes create entirely new markets together. His preferred path — U.S. and Chinese elites agreeing on ground rules — is, in his words, "not possible with the current administration."

For now, Schmidt leaves the industry with a two-horse race, a six-month gap, and no agreed mechanism to slow either horse down.