Chinese AI startups raised 24× less than US rivals as of mid-2026

Why can't Chinese AI companies raise as much funding as their US counterparts? The answer lies in a combination of export controls, market dynamics, and regulatory uncertainty that has created a 24× funding gap. As of July 2026, the three leading US AI startups—Anthropic, OpenAI, and xAI—had raised an average of $116 billion in equity funding, while four leading Chinese companies—DeepSeek, MiniMax, Moonshot, and Z.ai—had raised only $4.9 billion on average.

The first major constraint is compute access. US export controls have prevented Chinese companies from acquiring advanced AI chips, leaving them with far fewer training resources. Chinese companies owned 1.8 million H100-equivalents (a standard measure of chip capability) in early 2026, compared to at least 16 million for US companies—an order of magnitude difference. As a result, Chinese AI models typically use more than 10× less training compute than US counterparts. This gap has forced Chinese startups to operate at smaller scales; DeepSeek, for example, had to delay its R1 release in 2025 due to repeated failures during training on inferior domestic chips. While Chinese companies can legally rent compute through foreign cloud providers, the scale of this access remains unclear, and at least one company, Z.ai, has been added to the US Entity List, significantly limiting its ability to work with US firms. From an investor's perspective, if a company cannot access more chips to scale, additional capital offers limited returns, making large funding rounds less attractive.

A second major factor is the weakness of the Chinese market for AI products. The global AI market has been dominated by US companies: OpenAI released ChatGPT in 2022 and reached 100 million users within two months, while Anthropic has succeeded in the enterprise AI market globally. Chinese companies, meanwhile, have struggled to capture international share; only DeepSeek achieved a brief surge in early 2025. Within China itself, despite being the world's second-largest economy, Chinese AI companies have not generated significant revenue. At the end of 2025, their combined annualized revenue was less than 3% of OpenAI's. By July 2026, Z.ai may have reached $1 billion in annualized revenue, while Anthropic reached approximately $60 billion. The core problem: Chinese consumers don't want to pay for AI. In May 2026, Tencent President Martin Lau stated, "The penetration rate of paid users is currently only in the single digits, making it difficult to replicate the large-scale subscription-based development path seen overseas." ByteDance's Doubao lost 6.1 million monthly active users after introducing a paid model in May 2026. Lower labor costs in China also reduce the incentive to automate: median software engineer pay in Shanghai was $86,000 in 2022, compared to $234,000 in San Francisco, meaning Chinese firms can cheaply hire engineers to adapt open-source models rather than investing in proprietary AI solutions.

Third, the Chinese venture capital ecosystem is both smaller and more cautious about AI. In 2025, US VC funding reached $322 billion, compared to just $54 billion in China. More strikingly, only 5% of Chinese VC investment went to AI model companies, versus 28% in the US. Chinese venture investors have instead favored semiconductors and equipment, reflecting their focus on closing the compute gap left by export controls. Many Chinese VC funds are state-backed and thus aligned with government industrial priorities rather than independent investment returns. Chinese AI companies also lack access to sovereign wealth from the Gulf and Southeast Asia that has backed US firms: Abu Dhabi's MGX closed a $49 billion AI fund in July 2026 to invest in US companies including OpenAI, Anthropic, and xAI, while Qatar's QIA and Singapore's Temasek have joined recent US funding rounds. Under US–UAE trade agreements, state-linked Emirati firms must avoid Chinese AI companies flagged by Washington. Some foreign investment remains open to Chinese firms—Saudi Aramco's Prosperity7 joined Z.ai's $400 million round in 2024, and Temasek raised its China exposure by $8 billion in fiscal year 2026—but the pool is far narrower.

Regulatory risk compounds these barriers. In 2025, some DeepSeek employees faced travel restrictions after the R1 release, suggesting that commercial success invites state scrutiny. The Chinese government reportedly also pushed DeepSeek to train new models on Huawei rather than NVIDIA GPUs, a move that caused repeated training failures. Chinese authorities barred Manus co-founders from leaving the country amid scrutiny of the company's acquisition by Meta. These actions echo a precedent set in 2020: the Chinese Communist Party suddenly blocked Ant Group's IPO shortly after founder Jack Ma criticized financial regulators, then subjected both Ant and its majority owner Alibaba to years of pressure, imposing a $2.8 billion fine on Alibaba for antitrust violations and a three-year "rectification" period, while fining Ant $984 million and forcing it to restructure as a financial holding company. This pattern signals to investors—both domestic and foreign—that state intervention can disrupt companies' activities, acquisitions, and exits, reducing expected returns. While the US government has also intervened in AI (designating Anthropic a supply chain risk in March 2026 and forcing it to withdraw Fable in June 2026), these actions have not substantially curtailed Anthropic's capital access; Anthropic raised $65 billion in May 2026 and is reportedly pursuing an IPO. By contrast, state intervention in China has been more frequent and disruptive, discouraging large-scale private investment.

The combined effect is that Chinese AI startups face both a weaker business case and a more challenging investment environment than their US peers. Despite this gap, Chinese companies remain the only non-US actors competing near the AI capability frontier. However, evidence suggests they may be relying heavily on distillation of superior US models rather than fully homegrown innovation, likely keeping them 4–6 months behind and unable to capture significant market share from leading US competitors. The Chinese government could theoretically subsidize its AI companies with tens of billions of dollars per year to fund frontier-scale training, but subsidies may prove only a temporary strategy given that the cost of training the largest models is more than doubling every year. Moreover, securing large-scale compute remains difficult as long as US export controls remain in force. Nonetheless, the Chinese market itself is likely large enough to sustain a robust domestic AI ecosystem; Z.ai's revenue growth by deploying models to Chinese companies and government organizations suggests that path is viable, even if global competition remains out of reach.