U.S.-China AI race turns on capital, chips and commercial deployment

The U.S.-China AI race cannot be reduced to American spending or China’s lower operating costs, CNBC’s Aug. 17 China Connection newsletter argued. It cited BMI, a Fitch Solutions unit, estimating that private-sector AI investment in the United States is about 23 times mainland China’s, while reporting that China is pursuing domestic computing capacity and industry deployment.

The newsletter’s framework puts three constraints together: capital pays for infrastructure, chips determine how much computing capacity can be deployed, and commercial adoption tests whether those inputs create economic value. CNBC presented the final element as unresolved rather than declaring either country ahead.

What will decide the U.S.-China AI race?

CNBC said Beijing’s strategy is directed at AI self-sufficiency. Bruce Liu, chief executive of Esoterica Capital, told the network that China’s objective is to avoid dependence on the US, rather than necessarily produce the world’s leading AI system. The newsletter pointed to policy support, local subsidies, domestic chip development, lower-priced models and planned computing infrastructure as parts of that approach.

China released a three-year plan in June for infrastructure supporting faster computing power, CNBC reported. It subsequently said that computing-power networks could attract 4 trillion yuan in capital through 2030. That is an estimate of potential capital attracted, rather than an amount already committed or spent.

The US capital advantage has also drawn fresh backing from financial institutions. On Aug. 10, Nvidia announced memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to create independent compute-financing platforms intended to mobilize more than $500 billion of third-party capital over time. Nvidia said final agreements had yet to be executed.

Such platforms are designed to provide pools of financing for customers building AI infrastructure. Nvidia describes computing capacity as an investable asset, but its assertions about the economics and longevity of its equipment are company claims.

Why do chips limit China’s AI spending?

More financing does not by itself provide the advanced processors needed to train and run AI systems. Clifford Kurz, a director at S&P Global Ratings, told CNBC that support has limited value if companies lack chips to buy.

Kurz estimated that Huawei has roughly one-eighth of Nvidia’s computing capacity, mostly outside China. He said Huawei’s Ascend 950 has about 13% of the computing power of Nvidia’s GB300 chip, and projected Huawei would produce 1.35 million advanced AI chips this year, compared with a conservative estimate of 6 million for Nvidia. CNBC reported that Huawei has sought to compensate by combining more chips, while Chinese advanced chips still trail Nvidia’s.

CNBC also reported that Chinese companies have released models with similar capabilities at lower prices, an assessment it presented alongside the continuing constraint on computing hardware. A Brookings commentary separately described China’s effort as a full-stack approach spanning chips, infrastructure, models and applications, while noting constraints on compute and capital.

Access to funding remains a structural issue. Alexander Kheder of BMI told CNBC that US leadership would retain a durable advantage unless Beijing makes it easier for Chinese AI companies to secure external, non-state capital. Zhu He of the CF40 Institute told CNBC he had not observed significant plans for large-scale debt issuance by leading Chinese companies for AI projects; he said firms generally rely on equity finance and internal funds.

For CNBC, the remaining question is commercialization. US companies have spent heavily on advanced models, while China is seeking integration across industries. Winston Ma, an adjunct professor of law at New York University, told the network that the competition concerns applications built across the full AI stack. Whether either approach produces broader, durable returns remains unproven.

This story draws on original reporting from CNBC.