There is currently a trade tension between the US and China over soybeans.
America exports huge volumes of soybeans; China, in turn, has levers it can pull.
But while the two nations squabble over agriculture, another contest is brewing – one with even higher stakes: the contest over who controls the future of artificial intelligence.
Let’s step back a little.
You may have heard of China’s “Great Firewall,” strict internet controls, surveillance, censorship. So you might expect its approach to high-tech things would be secretive, tightly guarded.
But when it comes to AI, China is doing something counterintuitive: it’s giving away large chunks of its tech – open-sourcing complex models, publishing weights, encouraging a community to build and modify.
Why would a nation so protective of information open the gates to its crown jewels?
The Open-Source Gamble
Chinese AI firms have embraced open-source models. That means the technical nuts and bolts – the model architecture, the “weights” (the internal parameters), sometimes even the training code – are made available, so anyone can use them, adapt them, build on them.
In contrast, US heavyweights like OpenAI lean toward a closed-source model: you access the intelligence via APIs or subscriptions, but you don’t get the engine.
Chinese consumers have historically resisted paying for software services. So even their Netflix equivalent starts at less than half the entry price of its US counterpart.
In AI, Chinese companies “have been slashing API pricing for more than a year now.” The result is an AI “race to the bottom.”
China gives away more, charges less (or nothing) to get adoption. This is not ideal for margins, but it is superb for reach.
Speeding it up
Beijing views open source as essential to speeding progress.
Export controls can choke off chips and hardware, but it’s much harder to strangle lines of code. In essence, open source is harder to contain.
And China sees it as a tool to diffuse AI broadly across its economy, even if private firms sacrifice some profits. The state’s goal is adoption, not just monetization.
Meanwhile, Beijing frames AI not as a zero-sum “first to AGI” arms race, but more like electricity: infrastructure to be woven into every industry.
So yes, they’re handing the tech out.
And as China’s models proliferate globally, it also extends soft power: “AI for all,” they say, for the Global South and beyond.
The ascent of China’s Open AI
A year ago, OpenAI blocked certain Chinese developers from accessing GPT models, implicitly declaring that their crown-models had to be shielded from competition.
Soon, though, that strategy backfired.
China’s open-source offerings surged. Models like DeepSeek, Alibaba’s Qwen family, and MiniMax began to outrank or rival the Western giants in benchmark tests.
For example, MiniMax-M1 is fully open source under Apache 2.0 – “businesses can take it and use it … without restriction or payment.”
Meanwhile, in global AI model marketplaces like OpenRouter, DeepSeek’s models grabbed notable market share.
Even Nvidia’s Jensen Huang has complimented Chinese models: calling them “world-class” and noting that they serve as catalysts for progress globally.
Suddenly, the West is facing a choice: double down on closed, expensive models, or adjust.
OpenAI reportedly rushed to release free models of its own, with Sam Altman saying: “It was clear that if we didn’t do it, the world was gonna head to be mostly built on Chinese open-source models.”
What’s next
Open-source code spreads like wildfire; it’s immune to many forms of geopolitical strangulation.
As open models built in China become foundational, those who adopt them get built into China’s ecosystem: standards, updates, derivative models, libraries. It makes the playing field tilt subtly.
For the US and its tech champions, the reaction is messy.
Some push open models of their own, others double down on control for safety.
But the more China’s open models gain global traction, the harder it is to reassert proprietary dominance.
