China Builds an AI Industry at Uncommon Speed
China’s artificial-intelligence industry is expanding at a pace that would ordinarily be expected to lift technology shares, attract speculative capital and reinforce Beijing’s claim that the country can compete with the U.S. in a defining industry.
Instead, the boom is colliding with two powerful forces: a widening American campaign to exclude Chinese technology from critical infrastructure, and a sharp retreat by investors who have grown less willing to finance the AI race at almost any price.
China’s core AI industry exceeded 1.2 trillion yuan, equivalent to roughly $175 billion, in 2025, according to the China Academy of Information and Communications Technology, a research institute affiliated with the Ministry of Industry and Information Technology. The institute estimated annual growth of about 40%, while the number of Chinese AI companies rose above 6,000.
The figures illustrate how quickly AI is moving beyond Chinese laboratories and internet companies. More than 30% of the country’s larger manufacturing enterprises had adopted AI technology by the end of 2025, the government said. Beijing is promoting its “AI Plus” strategy, which seeks to embed machine intelligence in factories, healthcare, transportation, finance and consumer products.
Applications accounted for more than half of China’s AI industry chain last year, according to the academy. The fastest percentage growth, however, came from the layer containing models and software frameworks—a sign of the intense competition among developers seeking to turn large language models into commercial services.
China’s large domestic market offers developers a vast testing ground. Local companies can refine models for Chinese-language services, industrial automation and government-regulated sectors, while state policy directs capital toward computing clusters, semiconductors and other infrastructure.
But the industry’s scale also makes it more exposed to geopolitical pressure and investor scrutiny. Building competitive models requires expensive chips, networking equipment, electricity and data-center capacity. Revenue growth can be rapid while profits remain distant.
Washington Targets the Machinery Behind AI
The U.S. is broadening its technology confrontation with China from individual chips and telecommunications products to the physical infrastructure supporting AI.
The Trump administration is drafting restrictions on Chinese-made devices used in American data centers, including components that move information at high speeds between servers, according to reports published this week. The Federal Communications Commission is playing a central role in the effort, which officials are considering on national-security grounds.
The proposed action differs from export controls designed to prevent advanced American chips from reaching China. It would seek to remove Chinese equipment from the U.S. side of the AI supply chain, reflecting concern that foreign-made components inside critical facilities could be used for surveillance, disruption or unauthorized access.
Data centers have become the engine rooms of the AI economy. They contain not only graphics processors and servers but optical transceivers, networking systems, cooling equipment and power-management technology. Chinese manufacturers have established strong positions in several of those markets, often competing aggressively on cost.
A ban could deprive them of a major export market while forcing American data-center operators to find alternative suppliers. It could also encourage U.S. allies to adopt similar standards, further dividing a technology supply chain that was built over decades as a global network.
For Beijing, the pressure reinforces the case for self-sufficiency. U.S. semiconductor controls have already encouraged Chinese companies to develop domestic chips, optimize models to run on less-powerful processors and embrace open-source systems.
That response creates a paradox for Washington: Controls can slow access to leading technology, yet they also give Chinese companies and policymakers a stronger incentive to replace it.
Investors Question the Price of the Race
China’s stock market is now revealing the financial tension beneath the industrial advance.
Mainland technology shares suffered a severe selloff in July, contributing to the CSI 300’s worst monthly performance in roughly a decade. The benchmark fell 8.6% during the month, while an index of AI-related companies dropped sharply as investors pulled back from businesses tied to data centers, chips and optical equipment.
The retreat hit some of the companies positioned closest to China’s AI infrastructure push. Cambricon Technologies and optical-component makers including Zhongji Innolight and Eoptolink experienced heavy selling, erasing billions of dollars in market value.
The weakness also reached the market’s plumbing. The Shanghai Stock Exchange recorded 2.66 million new investor accounts in July, down 7% from June, while new margin-trading accounts fell more than 22%, according to exchange and securities-industry data cited in market reports.
Margin balances, a measure of money borrowed to buy stocks, also contracted as traders reduced leveraged technology positions. The decline indicates that investors who had amplified the rally were no longer confident enough to keep borrowing against expectations of further gains.
The market’s skepticism doesn’t negate China’s AI progress. It shows that technological achievement and shareholder returns are separate propositions.
Chinese AI companies are producing more models, applications and infrastructure. Washington is imposing barriers that make doing so more expensive. Investors, meanwhile, are asking whether revenue can grow quickly enough—and margins remain strong enough—to justify enormous spending on chips and computing capacity.
China’s AI industry has proved that it can scale. Its next test is whether it can withstand strategic containment abroad and financial discipline at home.
