The artificial intelligence rivalry between the United States and China is expanding beyond models and advanced semiconductors into safety rules, data centers and capital markets. Washington has proposed a mechanism for the two countries to notify each other of major AI incidents, while OpenAI has called on the U.S. government to lead the development of international technical standards for advanced AI. In China, Alibaba has unveiled a domestically developed AI chip and plans to expand its global data-center capacity to more than 20 gigawatts by 2032.

The two countries are pursuing different strategies. The United States is seeking common rules for evaluating AI capabilities, managing risks and responding to major incidents. China is expanding an AI infrastructure stack that connects domestic chips, models and data centers.

Yet the two AI ecosystems are not fully separated. U.S. financial institutions continue to participate in fundraising for Chinese technology companies, while Chinese capital is still flowing into U.S. AI companies and semiconductor markets. Technology and national-security barriers are rising, but financial interdependence remains.

These developments come ahead of a meeting between U.S. President Donald Trump and Chinese President Xi Jinping in Washington on Sept. 24. Xi is making a state visit to the United States from Sept. 23 to 25. Trade, critical minerals and Taiwan are among the issues expected to be discussed, while AI competition and risk management have emerged as another area to watch.

The central question is whether the two countries can continue competing in advanced technology while establishing rules to manage AI risks that neither can easily contain alone.

U.S. Separates Technology Controls From AI Risk Management

That distinction became clearer during talks in New York between U.S. Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng. The United States proposed a notification mechanism under which the two countries could inform each other about major AI incidents with potential national-security implications.

The proposal is intended to create a direct communication channel for risks that could be difficult for either country to manage independently, including the misuse of AI in cyberattacks or biological threats and unexpected behavior by advanced systems.

“Moving from opaque to more transparency between the number one and the number two AI powers in the world is very important,” Bessent said.

China has not publicly accepted the U.S. proposal. The discussions therefore should not be interpreted as an agreement on bilateral AI safety cooperation. U.S. export controls on advanced AI chips and semiconductor manufacturing equipment are also being treated separately from the proposed incident-notification mechanism.

Washington is restricting China’s access to advanced technologies while leaving open the possibility of communication over risks that could affect both countries. Technology competition and risk management are being handled on separate tracks.

OpenAI Calls for U.S.-Led International Standards

OpenAI has proposed that the United States take the lead in developing international technical standards for advanced AI.

The proposal focuses on common methods for measuring model capabilities and risks, conducting safety evaluations, determining when human oversight is required, and classifying and reporting serious incidents. OpenAI has also proposed an international network through which national AI safety bodies could share evaluation and reporting standards, with South Korea included among the potential partner countries.

International standards extend beyond technical specifications. Rules defining which AI capabilities are considered dangerous, what testing systems must undergo and which incidents companies must disclose could influence development practices and market-access requirements.

This gives standards a strategic dimension. The United States is seeking influence not only through advanced models and semiconductors but also through the criteria used to evaluate and govern AI systems. If evaluation frameworks developed by U.S. companies and research institutions become the basis for international standards, AI developers elsewhere could eventually have to test and disclose their systems against those criteria.

The U.S. incident-notification proposal and OpenAI’s international standards initiative address different levels of the same problem. One concerns government-to-government communication when serious incidents occur; the other seeks common criteria for evaluating and reporting AI risks. Neither requires technological competition to stop.

China Builds Domestic AI Chips and 20GW Infrastructure

China is expanding the physical infrastructure behind its AI ambitions.

Alibaba has unveiled the Zhenwu V900, a new AI accelerator developed by its semiconductor operation. The company says the chip delivers a substantial performance improvement over its previous generation and is targeting mass production in the first quarter of 2027. Alibaba also plans to develop next-generation AI models with as many as 10 trillion parameters.

Its data-center ambitions are also expanding. Alibaba has set a target of increasing its global data-center capacity to more than 20GW by 2032.

The 20GW figure is a long-term target, not capacity already installed. Actual capital expenditure, electricity procurement and the pace of data-center construction will need to be monitored. The strategy, however, is taking shape around an integrated system of domestic AI chips, cloud infrastructure, data centers and large-scale models.

U.S. restrictions on China’s access to advanced AI semiconductors increase the strategic importance of domestic computing capacity. China’s response is to build infrastructure that reduces its exposure to external technology controls.

Technology Barriers Rise, but Capital Keeps Moving

Capital markets present a different picture.

According to LSEG data, U.S. banks have acted as bookrunners on 19 equity fundraising deals by Chinese advanced-technology companies this year, involving $17.2 billion. Chinese and Hong Kong investors’ holdings of U.S. equities have also surpassed $750 billion.

Chinese investment in U.S. AI startups has increased as well, rising from $436 million in 2023 to $8.9 billion this year.

Governments are reorganizing semiconductor and critical-technology supply chains, while companies are building increasingly domestic AI infrastructure. Global capital, however, continues to move across both markets. U.S. investors are seeking exposure to growth in China’s technology sector, while Chinese investors continue allocating capital to U.S. AI and semiconductor companies.

The U.S.-China AI rivalry therefore cannot yet be described as two fully separated ecosystems. Technology and national-security boundaries are becoming more restrictive, while financial interdependence persists.

That connection could become a source of financial risk if bilateral relations deteriorate sharply. It also raises the economic cost of an abrupt technological and financial separation.

Sept. 24 Summit Brings Two AI Strategies Together

The Trump-Xi summit on Sept. 24 is expected to cover trade, critical minerals, Taiwan and other bilateral issues. The extent to which AI will produce a concrete outcome remains uncertain.

Several specific signals will matter. The first is China’s response to the U.S. proposal for an AI incident-notification mechanism. The second is whether the two governments establish a regular AI safety dialogue or an emergency communication channel. The third is whether U.S. restrictions on advanced AI semiconductors remain separate from any limited cooperation on risk management.

If China responds to the incident-notification proposal and the two sides establish a regular AI dialogue while Washington maintains its semiconductor controls, a distinct structure in U.S.-China AI relations would emerge: competition in technology, separate infrastructure development and limited communication over risks that could affect both countries.

International standards add another dimension. The United States is seeking to translate AI evaluation and safety frameworks into international rules, while China is expanding domestic semiconductors and data centers to build independent computing capacity. Capital continues to cross between the two markets.

Model performance defined the early stage of the AI rivalry. Advanced semiconductors, data centers, electricity and capital subsequently became strategic assets. Incident response, safety evaluations and international technical standards have now entered the competition.

There is currently little basis to conclude that the Sept. 24 summit will ease the broader U.S.-China AI rivalry. The immediate test is whether the two governments can establish a minimum communication mechanism for major AI incidents while continuing to compete in technology and infrastructure.

The summit will also provide an early indication of how U.S.-led international standards and China’s domestic AI infrastructure will coexist as technology restrictions tighten. The durability of cross-border capital will remain another important variable.

The Trump-Xi summit is not a meeting to end the AI rivalry. It is a test of whether the two countries can compete in technology and infrastructure while establishing minimum rules for managing shared risks.

Its outcome could provide an early policy signal of how the U.S.-China AI contest develops into a long-term competition encompassing international rules, semiconductors, data centers, power and capital.

Korean version available here