The U.S. is officially seeing its AI lead fade fast against China. Once miles ahead, American AI models now only barely outperform Chinese rivals. DeepSeek, Alibaba Cloud, Baidu, and Tencent have closed the gap to single-digit percentage differences in key benchmarks by late 2024.
The Trump and Biden administrations both agreed: winning AI dominance over China is a must. The strategy: slow China’s tech exports while pushing U.S. AI innovation with light oversight and investments in chips, energy, and government AI adoption. It worked—for a while.
But China’s massive state-backed push is paying off. Their Next Generation AI Development Plan, AI education, and public spending on semiconductors and data centers have shifted the game. DeepSeek and Qwen are now matching top U.S. models. AI in Chinese manufacturing is also winning, with Xiaomi’s factory cranking out electric vehicles every 76 seconds using 700+ AI robots. 
Export controls on advanced chips are losing steam, too. China uses shell companies and stockpiles to bypass bans while boosting homegrown chip development. They’ve even innovated software tricks to get more from existing hardware.
The U.S. can still fight back. OpenAI’s $500 billion Stargate project, plus huge investments from Amazon, Meta, Microsoft, and Google keep America competitive. But Washington knows it can’t rely on dominance forever. A national AI Action Plan is set for July rollout.
Officials are now planning for a world where U.S. and Chinese AI ecosystems coexist. That means new standards for AI evaluation beyond just raw performance—looking at cost, transparency, and ease of switching between models. The goal: stop users from getting locked into one ecosystem.
The U.S. also wants to build better tools to compare outputs from different AI models and keep applications flexible to swap underlying AI engines quickly.
Sharing data with Chinese AI builders won’t be banned outright either. The government recognizes some Chinese models might outperform U.S. ones in areas like medical diagnosis. Sharing anonymized or masked data could help tap those benefits while managing security risks.
Here’s Jake Sullivan, Biden’s National Security Adviser, warning in October 2024:
> “The United States risked ‘squandering [its] hard-earned lead’ if it did not ‘deploy AI more quickly and more comprehensively to strengthen [the country’s] national security.’”
Washington knows losing outright isn’t game over. But falling behind without a plan to adapt could cripple U.S. AI industries as China surges ahead economically and militarily.
The AI race just got a lot messier.
