Eric Schmidt says gap between China and US in AI is now negligible

Eric Schmidt thinks the AI race between the US and China is much closer than Washington would like.

In May 2026, the former Google CEO said China’s leading AI models now trail top US systems by under six months. In 2025, that gap was estimated at one to two years.

That is a lot of ground to make up in roughly a year. It is even more striking because China did it while US restrictions cut off access to advanced American chips.

A gap measured in nanoseconds

Schmidt, who now chairs the Special Competitive Studies Project (SCSP), did not soften the assessment. He described the remaining difference in AI development timelines as negligible, measured in “nanoseconds.”

Schmidt cautioned that those controls are beginning to fail in their intended purpose.

Schmidt credited the catch-up to software innovation and workarounds that keep Chinese models competitive on weaker hardware. He specifically pointed to models like DeepSeek as examples of that approach.

Not just one man’s opinion

Schmidt is not alone in landing on the six-month figure. Independent analyses in mid-to-late 2026 also put the US lead in frontier AI models at approximately six months.

A report from the Institute for Progress reached that estimate. Demis Hassabis of DeepMind has also said Chinese AI firms are currently around six months behind Western capabilities.

Those same analyses point to a technique called distillation as part of China’s playbook. In simple terms, a model learns from the outputs of a more capable model, a bit like a student learning from a skilled teacher’s worked answers instead of deriving everything alone.

Chinese developers are using distillation to partly make up for their hardware shortfalls.

The bubble question

Schmidt paired his warning about China with a confident take on the AI economy. In his view, AI is not a bubble.

Throughout 2025 and 2026, Schmidt has argued that customer demand for automation backs up the spending. He has cited business operations, healthcare and logistics as industries where that demand shows up.

He has even suggested AI may be “under-hyped” relative to the scale of its expected impact.

Why the timing matters

The two halves of Schmidt’s message pull in interesting directions. One says AI is a durable, real economic engine. The other says the US lead in building that engine is thinner than many assumed.

The 2025 estimate of one to two years gave policymakers some breathing room. A gap described in “nanoseconds” offers very little.

What this means

For investors, Schmidt’s “not a bubble” stance offers some reassurance to anyone nervous about AI valuations. His case points toward long-term support for companies tied to automation demand and the infrastructure and hardware buildout behind it.

If export controls are faltering, as Schmidt warns, Washington will face pressure to rethink its approach.

China’s progress suggests that clever software can partly substitute for top-tier chips, which complicates any thesis that hinges entirely on hardware scarcity.

What to watch next is whether the six-month benchmark holds, shrinks further or widens again. Expert assessments from groups like the Institute for Progress, along with comments from figures like Hassabis, will be the scoreboard. US policy decisions on chip controls will help decide which direction the number moves.