China's open-weights AI strategy is winning
China's Open-Weights AI Strategy is Winning
By Ben Werdmuller | July 20, 2026
While American AI remains locked behind proprietary walls, China is pivoting toward an open-weights strategy that is effectively seizing the lead. The US approach—prioritizing closed systems—is not only destined for failure but could potentially destabilize the broader American economy.
The Illusion of the "Model Moat"
Many believe that the AI model itself is the primary competitive advantage. However, proprietary models are a moat. In reality, the actual "moat" consists of the enterprise ecosystem surrounding the model:
- Complex service contracts and legal deals.
- Deep integration with existing corporate infrastructure.
- Specific "quality of life" features tailored for business workflows.
Because models are often accessed via an API, switching costs are remarkably low. An engineer can simply swap one endpoint for another while keeping the same prompt.
# Example of how easily a developer can switch providers
curl https://api.openai.com/v1/chat/completions \
-H "Authorization: Bearer $OPENAI_API_KEY" # Swap this...
curl https://api.anthropic.com/v1/messages \
-H "x-api-key: $ANTHROPIC_API_KEY" # ...for this.
Strategic Comparison: US vs. China
| Feature | United States Strategy | China Strategy |
|---|---|---|
| Access | Closed / Proprietary | Open-Weights |
| Distribution | Centralized Services | Permissionless / Portable |
| Primary Goal | First-order Profits | Ecosystem Dominance |
| Moat Focus | Brand & Lock-in | Ubiquity & Integration |
The Logic of Open Weights
China faces significant hurdles: they lack the same compute access as the US, and international regulations often prevent users from sending sensitive data to Chinese servers. Consequently, they cannot compete with OpenAI or Anthropic on a centralized service level.
Instead, they have turned this weakness into a distribution advantage. By releasing open-weights models (which are portable and permissionless, though not strictly "open source"), China ensures their tech is the foundation for global innovation.
The Closing Performance Gap
Previously, the US maintained a lead because frontier models were vastly superior. That advantage is evaporating.
"Moonshot and Alibaba unveiled models they claim can go toe-to-toe with the best from OpenAI and Anthropic at a fraction of the cost."
This rapid acceleration suggests that the US lead is shrinking just as AI becomes the cornerstone of geopolitical power and national security. The impact is already visible in the startup world. As noted by a16z partner Martin Casado, the probability of a startup utilizing Chinese models is staggering:
The Great Paradox
There is a striking irony here: China, a society known for being strictly controlled, is embracing an open technological strategy. Meanwhile, the US—the champion of the open internet—is seeing its leading companies hoard their technology.
While there are valid concerns regarding how Chinese models might be tuned to reflect government narratives (e.g., their handling of Tiananmen Square), the US strategy of "locking down" tech with no real moat is a losing game.
Economic Risks and the Path Forward
The US economy is currently heavily leveraged on AI spending. If the market realizes that closed models are commoditized and the "bottom falls out," the economic repercussions will be severe.
To pivot, the US needs to align its incentives toward ecosystem benefits rather than immediate corporate profits. We must support:
- Public AI initiatives
- Federated services
- Open, transparent research
Further Reading & Context
The need for collaborative infrastructure extends beyond AI. For instance, the "PIT Crew for news" initiative seeks to protect journalism through similar collaborative, cross-disciplinary technology capacity.
