The U.S. wants Asia on its AI. China's cheaper models are winning

The United States wants to be Asia's primary AI supplier. China is offering cheaper, open-source alternatives. The outcome of this contest carries real implications for the chipmakers, cloud providers, and AI model companies that count on the region's growth.

At the Asia-Pacific Economic Cooperation Digital Weeks in Chengdu, China, this month, the U.S. presence was subdued. A Commerce Department official, Bill Guidera, promoted the American AI Exports Program. Politico reported earlier this month that the department had received only 78 applications for that program, less than expected. The Commerce Department pushed back, saying the volume exceeded expectations. The White House did not comment.

Google and Meta had booths at the event. Their focus was on molecular AI and small-business tools, not large language models. Google's government affairs vice president, Wilson L. White, made only passing references to Gemini. After him, a Tencent vice president highlighted the company's Hunyuan large language model and a cloud project in Thailand.

China is hosting APEC this year. President Xi Jinping announced at the World AI Conference on July 17 that China would provide 5,000 AI training opportunities for developing countries and set up cooperation centers in Southeast Asia. Later in Chengdu, the 21 member economies, including the U.S., agreed to back open-source AI with "strong security."

"The endorsement of open-source models with strong security assurance gives China's open-weight strategy greater regional legitimacy, especially across emerging Asian economies where deployment cost and technological sovereignty are major considerations," said Wei Sun, principal analyst at Counterpoint Research.

Gary Dvorchak, managing director at The Blueshirt Group, said the U.S. strategy is to stop China from becoming the leading AI supplier for Asia. China's models are cheaper. The U.S. offers a more complete stack from chips to models, he said. The cost gap is widening. New Chinese models have launched with capabilities similar to U.S. models at a fraction of the price.

The risk for investors is a slow erosion of U.S. market share in Asia's AI infrastructure. Nvidia, whose chips are the standard for training, has a direct stake. The company's stock fell 3.55% today to $190.01, with an Alpha Score of 70 (Moderate). Alibaba, which opens its Qwen model for developers, has an Alpha Score of 54 (Mixed). Tencent, whose Hunyuan model is gaining adoption, scored 52 (Mixed).

A permanent shift in Asian AI procurement would affect not just Nvidia. Companies like Google, Meta, and Microsoft compete in the cloud and model layers. The region's linguistic diversity adds another complication. More than 1,300 languages are spoken in Southeast Asia alone. Governments are spending billions on locally tailored systems, said Pak-Sun Ting, CEO of Votee AI, a startup working with at least five governments on custom models. Ting noted that entities in Southeast Asia tend to use Nvidia chips for training. They may use other chips for inference.

"While the U.S. and China are fiercely competing in AI technology and diplomacy through distinct approaches, they ultimately cannot fully decouple from one another," said Yue Su, principal economist at the Economist Intelligence Unit. She pointed to the light U.S. presence at APEC as partly a scheduling conflict with a San Francisco AI summit that same week.

A full decoupling is unlikely. The risk of the U.S. losing share in the fastest-growing AI market outside North America is real. The next catalysts are the APEC leaders' summit in November and any new export controls on AI chips. For now, the market's best source of confidence is the technical lead of Nvidia's hardware. That lead is not guaranteed to hold cost pressure.