Foreign investors once dismissed China’s AI push as a constrained, second-tier effort. Yet now, strategists argue that the country may be better positioned for AI, thanks to cheaper power, growing capital spending, and a swarm of open-source developers. Those same analysts are also wondering whether the U.S. AI boom, after years of sky-high valuations and data center spending, is running out of steam.

“We’ve actually reduced our exposure to U.S. tech,” Mohit Kumar, Jefferies’ global macro strategist, told Fortune at the investment bank’s Asia Forum in Hong Kong last week. “We believe that China is the big winner in this tech war for a number of reasons: valuation, wider adoption of AI, an advantage in power generation.”

“China basically has unlimited access to cheap energy, whereas the U.S. has this massive energy bottleneck,” Jefferies global head of equity research Chris Wood told Fortune. 

The country will have roughly 400 gigawatts of spare power capacity by 2030, equal to three times what the world will need to meet data center demands, according to Goldman Sachs. In contrast, the U.S. is struggling with aging infrastructure and a lack of generating capacity, leading to spiking power prices in data center states like Virginia. 

In China’s western provinces like Ningxia and Gansu, electricity can cost as little as five cents per kilowatt-hour, versus 25 cents in Beijing or Shanghai, or 40 cents in some parts of the U.S, according to Baidu’s chief financial officer, Henry He, who noted to attendees that power can make up about 35% of inference costs.

China’s manufacturing sector also helps its AI sector, particularly in applications that interact with the physical world. For example, He noted that Baidu’s Apollo robotaxis can charge 1 yuan ($0.15) per mile and still be operating at breakeven in the city of Wuhan.

In the U.S., robotaxi manufacturers have had to choose between lidar sensors or visual cameras owing to cost. But He argued that, in China, “we don’t need to make that difficult choice,” as both are affordable thanks to Chinese manufacturing. 

The same logic extends to autonomous aviation. EHang, for example, relies on a domestic supply chain that can provide affordable batteries and electronic components to build its aerial vehicles. “We have an extremely competitive component cost, and we can turn it into a very competitive selling price,” CFO Conor Yang explained to Fortune. 

A peaking U.S. AI cycle?

Wood’s optimism in China is contrasted by his questions about the U.S. The Jefferies strategist thinks that investors are starting to ask questions about the amount of money being spent in the U.S. on AI infrastructure, and expects U.S. capital spending to peak this year.

In an interview with Fortune, Wood described the U.S. private equity market as suffering from “a giant case of financial constipation,” with tens of thousands of portfolio companies waiting to IPO. In addition, private equity has recently poured money into the software sector, right as AI now threatens to erode their business models.

Private credit has been tested since the bankruptcy of auto parts supplier First Brands Group last year, which spooked retail investors and pushed some asset managers, like Blue Owl, to restrict withdrawals. This year’s “AI scare trade” has only made matters worse as retail investors try to pull their capital from fund managers, leading other fund managers, like BlackRock and Morgan Stanley, to also cap withdrawals. 

Monetization is still difficult

Still, China’s AI companies may struggle to charge much for their products in the country’s hypercompetitive environment, where labs compete on both price and performance.

“It is very difficult for a single company to always [have] a top model globally,” He, from Baidu, said. (Baidu, an early mover in the space with its ERNIE model, has since lost ground to Chinese rivals like Alibaba and DeepSeek.)

Startup MiniMax, for example, generated $79 million in revenue last year, yet reported a $1.8 billion net loss. (Investors don’t seem to care, driving shares up by more than sixfold since the startup’s IPO in early January.)

The problem for AI labs, whether in the U.S. or in China, is that any lead in performance could disappear in a matter of months, as other labs catch up by offering models with near-frontier-level performance, often on an open-source basis.

Wood, from Jefferies, thinks that large language model providers will end up like utilities: capital-intensive, commoditized, and unlikely to earn sustainable returns. Instead, he argued that China’s AI boom “is going to be in applications—cheap applications—made possible by open‑source models and very cheap power.”

Agentic AI could be the next battleground. Chinese companies large and small are quickly rolling out their own OpenClaw frameworks, and local governments are offering subsidies to “one-person companies” building their own AI agent startups. 

Chinese tech companies already have strong consumer platforms to put AI agents in front of users. Tencent’s WeChat, for example, has over 1.3 billion monthly active users and hosts millions of mini programs spanning commerce, transport, lifestyle, and finance. 

“We’re going to see some interesting innovations in China. This is the only area where the U.S. and the West are going to be behind,” Michael Bruck, founding partner at 71 Capital, told Fortune. 

“Is it a huge stretch of imagination to think the next version of a mini program is going to be an agent built into WeChat?”