Washington's export controls assumed AI would stay expensive and American, but China just made the technology too cheap and too good to contain.
The Summary
- China's AI models are rapidly gaining global competitiveness, forcing the US to rethink its standard protectionist strategy of chip export controls
- Trump's own AI advisors are now publicly attacking US AI companies, revealing fractures in America's AI policy response
- The "AI for All" strategy, commoditizing frontier models through open releases, is undermining the assumption that cutting off chips would keep China behind
The Signal
China's AI offensive is exposing the core flaw in US containment strategy: you can't gate-keep something that refuses to stay gated. The playbook, restricting advanced chip exports to slow China's AI development, worked under one premise. That cutting-edge AI would remain capital-intensive, proprietary, and concentrated in a handful of Western labs.
But Chinese labs aren't playing that game. They're releasing competitive models openly, often matching or beating US counterparts on benchmarks, while requiring less compute. DeepSeek's recent releases shocked researchers by achieving frontier performance at a fraction of the expected cost. When you can't buy the best chips, you get creative with the ones you have.
"The rapidly increasing global competitiveness of China's artificial intelligence models is ringing alarm bells in Washington."
The fractures inside the Trump administration tell the real story. David Sacks, Trump's AI and crypto czar until recently, joined other current and former advisors in publicly attacking America's leading AI companies over the weekend. The targets: the same firms Washington ostensibly wants to support against Chinese competition.
The contradiction is stark. Here's what's happening:
- US policy still treats AI leadership as a national security asset to protect
- But Chinese labs are commoditizing that asset by giving it away
- American companies want to compete globally, which means engaging with China
- Hardliners see any engagement as strengthening the adversary
This isn't a policy debate. It's cognitive dissonance at the cabinet level. You can't simultaneously treat AI as a strategic weapon AND allow the market dynamics that made American tech dominant. China figured out the exploit: if the US won't let you buy the shovels, you flood the market with free gold.
The "AI for All" framing is deliberate. It positions China as the democratizing force against American monopolies. Never mind the irony of an authoritarian state championing openness. What matters is the narrative works in Jakarta, Nairobi, São Paulo, everywhere the US is trying to build AI alliances. Those countries now have access to models that would have cost millions to develop or license from OpenAI or Anthropic.
The Implication
The chip export controls were the wrong tool because they solved yesterday's problem. They assumed AI would scale vertically, bigger models, more chips, higher costs. China is forcing horizontal scaling instead: better algorithms, more efficient training, open distribution. If that continues, the US either locks down its AI companies completely, killing their commercial viability, or accepts that the AI race won't be won by containment.
Watch what happens next with open source AI policy in the US. If Washington clamps down on American labs releasing weights, it hands China the entire "AI for All" narrative. If it doesn't, it admits the export controls failed. Either way, the assumption that you can out-spend and out-chip your way to AI dominance just got a lot shakier.