• 2 min read
China’s open AI push is eroding America’s edge
A new argument says US AI firms are betting on closed models with little moat, while China is gaining ground through open weights and lower costs.
Image: Hacker News
China’s open-weights AI strategy is starting to look like a real threat to the US lead in artificial intelligence. Writing on his blog, Ben Werdmuller argues that American companies have built their businesses around closed, centralized services, even though the models themselves are easy to replace and offer little durable lock-in.
His core claim is simple: the real moat is not the model, but the enterprise services around it — contracts, integrations, and workflow features. In practice, he says, users can move from ChatGPT to Claude with minimal disruption, especially in software engineering, where models are often accessed through an API and can be swapped without changing the broader workflow.
That matters because open technologies tend to win at the infrastructure layer. Werdmuller argues that open weights models, while not the same as open source, are still portable and permissionless: companies can host them where they want, modify them for their needs, and build on top of them without relying on a single vendor.
Recommended reading
Tu-214 wing redesign seen as too costly to pursue
How export controls may have backfired
According to Werdmuller, US export controls on GPUs and restrictions on sharing some kinds of data with Chinese servers have limited China’s ability to offer global, centralized AI services in the style of OpenAI or Anthropic. But rather than stopping Chinese firms, he says, those constraints pushed them toward a more effective distribution model: releasing models openly and turning a compute disadvantage into an ecosystem advantage.
He points to a recent report by Robert Hart in The Verge, which said Moonshot and Alibaba unveiled models they claim can compete with the best from OpenAI and Anthropic at a fraction of the cost. Werdmuller also cites a16z partner Martin Casado, who said in The Economist that there is an 80% chance any given startup is using Chinese models.
Werdmuller acknowledges concerns about Chinese censorship, including how models respond to subjects such as Tiananmen Square. But he argues the bigger strategic irony is that China, often seen as restrictive, is pushing more portable AI technology while US firms are keeping tight control over theirs.
He warns that this mismatch could have wider consequences if AI spending, now a major driver of the US economy, drops sharply. His preferred alternative is more support for public AI, federated services, and open research — areas he says have momentum, but not enough backing.
Enterprise Editor
Marcus follows the money. He covers enterprise software, cloud architecture, and the tectonic shifts in Big Tech strategy. He translates dense earnings calls and complex M&A activity into actionable insights about where the industry is actually heading. If a tech giant makes a silent pivot, Marcus is usually the first to notice.
via Hacker News
