China dominates cheaper AI in Asia as U.S. struggles at APEC

The 21 member economies of the Asia-Pacific Economic Cooperation endorsed open-source artificial intelligence with strong security at a meeting in Chengdu, China, on July 23, 2026, handing Beijing a diplomatic victory as Washington struggled to sell its own AI technology in the region.

Washington left few public traces of its involvement in the APEC Digital Weeks, held in Chengdu in July 2026, while China used its host role to showcase AI capabilities and secure backing for cheaper Chinese models. Gary Dvorchak, managing director at The Blueshirt Group, said the American strategy is to stop China from becoming the leading AI supplier in Asia and the world.

Beijing sent Vice Premier Zhang Guoqing to the ministerial-level talks on July 23 to push for tech standards, securing the open-source endorsement the same day. The agreement gives China’s open-weight strategy greater legitimacy, especially in emerging Asian economies where deployment cost and technological sovereignty matter, said Wei Sun, principal AI analyst at Counterpoint Research.

Chinese President Xi Jinping had laid the groundwork on July 17, announcing that China would provide 5,000 AI training opportunities to developing countries and establish cooperation centers in Southeast Asia.

At a July 24 forum during the Chengdu event, Bill Guidera, a deputy under secretary at the U.S. Department of Commerce, promoted the American AI Exports Program, calling it “a brilliant design” that lets buyers acquire a full U.S. tech stack. But Politico reported earlier in July that the program had drawn only 78 applications; an International Trade Administration spokesperson said the volume exceeded expectations.

Google and Meta were the only U.S. companies with booths on the show floor as of July 23, and they focused on niche applications rather than large language models. Google’s Wilson L. White made only passing references to its Gemini model in a speech the next day, while Tencent Vice President Cai Guangzhong highlighted growing adoption of his firm’s Hunyuan LLM and a cloud project in Thailand.

Analysts expect a hybrid AI landscape in Asia, not a clean split between U.S. and Chinese spheres. With more than 1,300 languages in Southeast Asia alone, governments are spending billions on localized AI, often training systems on Nvidia chips while also leveraging Chinese open-source models, said Votee AI CEO Pak-Sun Ting. Yue Su, principal economist at the Economist Intelligence Unit, noted that although competition is fierce, the U.S. and China cannot fully decouple. She added that the light U.S. presence in Chengdu was partly because of competing events such as a San Francisco AI summit on July 24 where South Korean President Lee Jae-myung met with OpenAI’s Sam Altman, Anthropic’s Dario Amodei, and Nvidia’s Jensen Huang.