The combined annual recurring revenue (ARR) of the seven largest Chinese AI developers has reached $10.7 billion, which is merely around 10% of the figures reported by OpenAI and Anthropic, according to a report from the Rhodium Group.
Experts noted that the revenue estimates of certain Chinese firms are now comparable to or even exceed those of their American counterparts. The analysis focused on the funding of China's AI industry, examining data from Alibaba, ByteDance, Kuaishou, DeepSeek, Moonshot AI, Z.ai, and MiniMax, comparing their performance with U.S. competitors.
Revenue Lags Behind Valuations
According to the Rhodium Group, the total ARR of Chinese AI models was approximately $10.7 billion from March to August 2026. In contrast, OpenAI and Anthropic reported over $100 billion in combined revenue.
ByteDance emerged as the frontrunner among Chinese companies, achieving around $4 billion in ARR by July, followed by Alibaba with $2.4 billion in August. The revenue figures for other developers were significantly lower.
Source: Rhodium Group.Despite the rapid growth in revenue, profitability remains a challenge. For instance, Z.ai's ARR surged from $74 million in January to $1.6 billion in August, marking an increase of nearly 20 times. MiniMax's revenue also quadrupled during the same period.
China's AI Investments Reach Hundreds of Billions
The Rhodium Group estimates that Chinese companies will invest 932 billion yuan ($139 billion) in AI infrastructure by 2026, more than double the previous year's level.
In 2027, this figure could exceed 1.2 trillion yuan ($193 billion). For comparison, U.S. investments in data centers are projected to be around $800 billion in 2026.
This indicates that China is rapidly expanding its infrastructure, even as the revenue from AI models grows at a slower pace. The combined free cash flow of the three largest Chinese hyperscalers—Alibaba, Tencent, and Baidu—turned negative in the first half of 2026, amounting to a loss of 16 billion yuan, down from a positive 170 billion yuan the previous year.
Source: Rhodium Group.Chinese AI Firms Rely on Investors
The financial model of China's AI sector contrasts with that of the U.S. In the United States, major tech companies are increasingly utilizing debt financing: the net inflow from the issuance of debt instruments among the five largest American hyperscalers rose from $90 billion in 2025 to $163 billion in the first half of 2026.
Conversely, Chinese companies rely more on equity financing and bank loans.
By August 21, 2026, capital investments in China's AI industry had reached 282 billion yuan. Notable growth in private and venture capital funding has been observed for leading AI labs like Z.ai, MiniMax, DeepSeek, and Moonshot. While their combined investments from private equity and venture capital were around 9 billion yuan in 2025, they surged to 179 billion yuan in the first eight months of 2026, including IPO and private placement proceeds.
Source: Rhodium Group.The government plays a significant role in this landscape. Approximately 25% of direct investments in China's AI sector in 2026 are attributed to state funds and related entities. In the chip and server segment, the share of state-affiliated investors reaches 47%, with another 14% coming from banks, predominantly state-owned. Consequently, over 60% of investments in this segment have governmental or quasi-governmental origins.
High Valuations with Low Revenues
Another finding from the report highlights the valuation of Chinese AI companies. The Rhodium Group compared their market capitalization with ARR and discovered that some Chinese developers are trading at higher multiples than OpenAI and Anthropic.
The valuation-to-ARR ratio for OpenAI and Anthropic is estimated at around 34x and 21x, respectively. In comparison, Z.ai's ratio is approximately 46x, Moonshot's is about 50x, and DeepSeek's reaches 163x.
Moonshot's example is particularly striking: its valuation jumped from around $4 billion at the end of 2025 to $50 billion by August 2026.
However, the profitability of Chinese developers remains lower than that of their American rivals. For instance, DeepSeek's gross margin was estimated at 45% in July, compared to approximately 65% for Anthropic. Z.ai and MiniMax recorded margins of about 26% and 18%, respectively, in the first half of 2026.
One reason for this discrepancy, according to the Rhodium Group, is the lower pricing of Chinese AI services. Even after recent price increases, most Chinese frontier models cost between $0.04 and $0.50 per task, while top-tier LLMs from Claude range from $2 to $4, and the most expensive GPT models cost between $1 and $2.
The AI Race Hinges on Financing
The authors of the study conclude that the Chinese AI sector faces fundamental challenges similar to those in the U.S.: companies are simultaneously increasing capital expenditures and computational capacities while not generating sufficient cash flow.
Moreover, the financing model in China has its unique characteristics. To sustain investments, companies will need to maintain access to equity markets and bank credit. If the valuations of Chinese AI firms continue to rise faster than their revenues, any downturn in the stock market could restrict their ability to attract new capital.
The Rhodium Group anticipates further growth in AI investments in China during 2026–2027, but identifies financing as a key constraint for scaling the industry, alongside the availability of modern chips.
Notably, Morgan Stanley estimates that by 2030, monetizable revenue from consumer AI in China could reach 294 billion yuan (~$44 billion).