Xi Jinping's debut at the World AI Conference last week wasn't just a speech. It was the opening move in a contest over who gets to set the rules for artificial intelligence globally, and China arrived with serious weight behind it.
On July 17, standing before an audience in Shanghai, Xi framed China's AI ambitions in the language of inclusion: "AI development should not be a solo performance by a single country, but a symphony of international cooperation." The day before, 29 nations had already signed on to make that symphony official, formalizing the World Artificial Intelligence Cooperation Organization, headquartered in Shanghai. UN Secretary-General António Guterres attended the signing. The founding roster includes Russia, Brazil, Cuba, Venezuela, Indonesia, Pakistan, Kazakhstan, Serbia, Belarus, and a bloc of 10 African and 12 Asian countries. The United States isn't one of them.
This is a direct answer to Washington's export-restriction playbook. The Biden and Trump administrations both bet that controlling access to advanced chips and frontier AI systems would keep adversaries locked out and allies dependent on US technology. China's counter is to make those restrictions irrelevant, not by cracking the chip embargo, but by building a parallel ecosystem where it doesn't matter.
The governance bloc is the diplomatic layer. The technical layer is more consequential. Zhipu AI released GLM-5.2 on June 13 under an MIT license with no regional restrictions, meaning any government, university, or startup anywhere in the world can download it, self-host it, fine-tune it, and deploy it commercially, with no dependence on American cloud providers or API terms. No strings attached. Independent benchmarker Artificial Analysis now ranks GLM-5.2 the number-one open-weight model in the world and fourth overall, with benchmark results that match or beat OpenAI's GPT-5.5 on several long-horizon coding tests at roughly one sixth of the cost. The model runs on a mixture-of-experts architecture with about 744 billion total parameters, 40 billion active per token, and carries a one-million-token context window.
For a government in sub-Saharan Africa or Southeast Asia choosing between US-controlled API access and a freely downloadable Chinese model with permissive licensing, the choice isn't abstract. It's a procurement decision. China is making that decision easy. The $295 billion infrastructure plan Bloomberg reported in June reinforces the bet. According to Bloomberg, Beijing is preparing to spend roughly 2 trillion yuan over five years on a nationwide network of AI data centers, with state firms China Mobile and China Telecom operating the bulk of them. Critically, the plan mandates that at least 80% of the technology, including AI chips, come from domestic suppliers, with Huawei as the primary beneficiary. That's a deliberate squeeze on Nvidia, which has already seen its H20 chip sales to China cut off under US export controls. Beijing isn't trying to import the stack. It's building its own.
What this means if you're building or investing in AI #
For founders and investors, the WAICO framework creates a compliance headache that didn't exist six months ago. If you're building an AI product for international markets, you now have to make choices that were previously irrelevant: which governance regime your product is aligned with, which model weights you're running, and whether your infrastructure sits in a jurisdiction that has signed onto US-aligned frameworks, Chinese-aligned frameworks, or neither. Those choices will affect your access to customers, your regulatory exposure, and your ability to raise from US institutional investors who are increasingly subject to outbound investment restrictions. The Global South adoption play is real. Many of the WAICO founding members have large populations, rapidly growing technology sectors, and governments that have shown zero appetite for taking sides in the US-China technology rivalry. They want access to AI tools, and China is offering them a frictionless path: open-weight models they can run themselves, governance frameworks they helped shape, and data center infrastructure backed by Chinese state capital. That's a compelling package, and it doesn't require trusting Beijing unconditionally. It just requires preferring availability over restriction.
Frankly, the US export-control strategy was always a bet on maintaining a wide enough capability gap that restrictions would sting. GLM-5.2's benchmark performance suggests that gap is narrowing faster than Washington anticipated. A model that rivals the top closed-weight systems, ships free, and comes with no API terms to violate is a harder thing to contain than a chip. Chips need fabs. Weights need a hard drive.
The deeper irony is that the open-source ethos, long associated with American internet culture and Silicon Valley permissiveness, is now being deployed most aggressively by Beijing. China didn't invent the MIT license. But it's using it more strategically than anyone else right now.
Also read: BloombergNEF just revised its US data center power forecast 83% higher in seven months • Claude Fable 5 helped crack the Jacobian Conjecture after 87 years of failure • Curative CEO canceled a $600,000 Salesforce contract after vibecoding a replacement CRM in two months