A funding crisis is taking shape alongside China's chip constraints, according to reporting by The New York Times' Sui-Lee Wee and Li You published Friday. The country's leading technology companies plan to dramatically expand their AI infrastructure spending, yet their AI operations are generating only a fraction of what they need to sustain this growth.
Research from the Rhodium Group, released in September, surveyed 13 publicly listed Chinese firms and found their AI infrastructure capital expenditures will climb 103% this year to 932 billion yuan ($139 billion). The outlook extends further: Rhodium forecasts spending will exceed 1.2 trillion yuan ($193 billion) by 2027. However, revenue from AI labs at these companies covers less than 8% of this year's projected spending.
When you look at their numbers, it looks like they're running on a treadmill that's speeding up and getting harder to run
Kyle Chan, fellow at the Brookings Institution
Financing emerges as the deeper constraint
While much attention has focused on US restrictions limiting China's access to advanced semiconductors, the financial picture may pose an even greater obstacle. Logan Wright, the Rhodium partner who co-authored the report, told the Times that financing limitations "will be probably more decisive" in the long run than chip embargoes.
Rhodium estimates China's overall AI infrastructure development stands at only 15% to 20% of the US level. The combined annual recurring revenue across all Chinese AI models reached approximately $10.7 billion, roughly one-tenth of what OpenAI and Anthropic together report.
Major backers of China's leading AI initiatives face deteriorating finances. Alibaba, Baidu and Tencent—which support several of China's most prominent labs—are experiencing negative cash flows, according to the Times. Both Tencent and Alibaba issued warnings this week that fresh US restrictions will constrain their AI expansion.
Market sentiment has shifted sharply. China's STAR 50 index, the domestic equivalent to the Nasdaq, has declined 43% since late July, the Times reported.
Open-source strategy creates pricing pressure
China's emphasis on releasing open-source models has intensified competitive pressures and eroded margins. When Z.ai introduced GLM-5.3 Flash in August, it slashed prices for a two-week period to encourage users to run the model on its infrastructure. By November, approximately 20 competing providers were offering the same model at lower rates than Z.ai itself, according to OpenRouter data cited by the Times.
The scarcity and cost of computing power compounds the challenge. On Goofish, a Chinese secondhand marketplace, Nvidia's DGX B300 servers have been listed for nearly $3 million—roughly eight times their US price, the Times found.
DeepSeek's chief executive Liang Wenfeng acknowledged the limitations of alternative chip sources. He stated that Huawei's processors lag Nvidia's by approximately two years, and that DeepSeek would require four Huawei 950 chips to deliver equivalent performance to a single Nvidia GB300, according to a leaked investor call transcript cited by the Times.
We simply can't afford to train the largest models available right now
Liang Wenfeng, DeepSeek chief executive
DeepSeek declined to comment to the Times. The company has nonetheless secured more than $20 billion in funding, including a $15 billion round led by Tencent and CATL, positioning it well ahead of most competitors.
Limited capital sources
Private investment in AI across China reached $12.4 billion in 2025, compared with $285.9 billion in the US, according to Stanford's AI Index. The index acknowledges this figure likely understates China's actual investment, since state-directed funds channel additional capital into the sector. Foreign investment has contracted substantially, partly due to Washington's restrictions on US venture capital participation in certain Chinese technologies, the Times reported.
Despite these headwinds, China retains competitive advantages. The country's research community has repeatedly demonstrated capacity to train and operate models at lower cost. Han Lin, China country director at the Asia Group, told the Times that the nation possesses world-class engineering expertise and a government capable of mobilizing vast resources. What frontier AI development additionally requires, he noted, is venture capital willing to finance ventures that may fail.
Source: The Next Web



