Schmidt: U.S.-China AI gap narrows to six months

Eric Schmidt, former Google CEO and current Relativity Space CEO, stated that the AI performance gap between the U.S. and China has narrowed from one to two years to roughly six months. He attributed China's progress to algorithmic innovations that achieve comparable performance with one-tenth the energy of U.S. models, a result of restricted access to advanced chips. Schmidt noted that while the U.S. leads in frontier model performance and safety controls, China leads in model cost, with token costs three to five times cheaper. He highlighted specific releases, including Reflection AI’s 500 billion parameter model, NVIDIA’s NemoTron, and Google’s Gemma 4.

Schmidt argued that China’s strategy of open-source and open-weight models is gaining global traction, whereas the U.S. relies on closed-source models costing $100 million to $500 million each. He warned that open-weight models can be stripped of safety guardrails via a technique called obliteration, posing cyberattack risks, and predicted a significant cyber event within 12 months to illustrate this vulnerability. Schmidt argued that the U.S. must maintain its lead in frontier capabilities and safety while acknowledging China's structural advantages in cost and deployment speed.