Rhodium Finds OpenAI and Anthropic Generate Roughly Ten Times the Revenue of All Chinese AI Models Combined

Table of Contents

• You might want to know


• Main Topic


• Key Insights Table


• Afterwards...

You might want to know

How do revenue figures for Chinese AI model developers compare with those of leading U.S. firms?

What implications do revenue differences have for startup valuations, investment risk, and future scaling?

Main Topic

Recent estimates published by the U.S.-based research firm Rhodium Group indicate a striking gap between the revenues generated by leading American AI companies and those produced by the full set of Chinese AI model makers. Using an industry metric called annual recurring revenue (ARR), which annualizes a recent monthly revenue figure to reflect rapid growth, Rhodium found that all Chinese AI models combined account for roughly one-tenth of the ARR reported for OpenAI and Anthropic. This gap highlights a divergence between broad user adoption and monetization success in different markets.

Rhodium's analysis evaluated reported and estimated ARRs for several prominent Chinese AI firms. The report placed DeepSeek at an ARR of $500 million, MiniMax at $800 million, and Moonshot at $1 billion. Z.ai reported an ARR of $1.8 billion in a recent investor communication, while larger technology groups such as ByteDance and Alibaba contributed approximately $4 billion and $2.4 billion respectively. In contrast, OpenAI's ARR was estimated at around $40 billion and Anthropic's at roughly $65 billion. Taken together, the two U.S. firms dominate the revenue landscape, despite the rapid uptake of Chinese models in domestic and regional markets.

These figures are notable not only for the absolute differences but also for how they relate to firm valuations. Rhodium emphasizes that several Chinese startups are being valued at levels that appear disconnected from their current revenue streams. Specifically, the report highlights estimated price-to-revenue ratios of about 50x for Moonshot and 163x for DeepSeek. By comparison, the report cites multiples near 34x for OpenAI and 21x for Anthropic. These disparities imply that investors are pricing in stronger future growth for some Chinese startups than current revenues seem to justify.

Part of the explanation for lower reported revenue in China lies in differences in product models and distribution. Many Chinese AI labs publish or distribute models in ways that make them more accessible to third parties, including open-source releases or licensing approaches that permit customers to deploy models on their own infrastructure. That open approach can reduce direct revenue capture because third parties with sufficient hardware can run the models independently of the developer. In contrast, leading U.S. models are largely closed and often accessed via provider-hosted APIs, where the platform retains control and captures a larger share of usage fees. According to AI-comparison firm Artificial Analysis, the cost per task for top U.S. models is substantially higher than for comparable Chinese models, supporting higher revenue per unit of usage for the U.S. providers.

Rhodium also notes that its comparison relies on the latest available figures, some of which reflect summer data. Usage of Chinese models has been rising quickly from earlier lows, and several companies have updated their forward-looking revenue expectations. For example, Z.ai reportedly revised its year-end ARR projection upward to $3 billion from an earlier $2.4 billion target. As such, the revenue gap may narrow over time if growth continues and firms convert usage into recurring customer contracts and monetized services.

Still, the financing landscape presents a structural challenge. The Rhodium team argues that Chinese frontier AI labs may face difficulty achieving sustainable scale without continued favorable conditions in public and private capital markets. Historically, those markets for high-growth technology in China can be volatile. While the state has been an important source of funding—particularly for hardware such as chips and servers—direct government funding for frontier AI research and private model developers is likely to be more constrained. Rhodium estimates that more than 60% of equity investment in Chinese AI chips and servers has been linked to state-affiliated sources, underlining a reliance on public support for the infrastructure buildout.

Valuation and market dynamics also intersect with the IPO plans and investor sentiment surrounding several firms. Anthropic has been reported to be preparing for a U.S. listing, and OpenAI has postponed its IPO until next year. On the Chinese side, Moonshot has reportedly filed confidentially for a Hong Kong initial public offering, and DeepSeek is also said to be preparing for a potential listing. Public markets' reception of these listings will play an important role in validating or challenging the high valuation multiples currently observed.

Market volatility has been evident. Chinese AI stocks have moved sharply at times: Z.ai's shares rose over 5% in one session after earlier declines, while other companies such as MiniMax have struggled to sustain IPO-day gains. Broader sentiment for tech names also shifted after executives at leading American AI firms warned about the risks of rapid development, triggering sell-offs in U.S. tech stocks. Leaders of major Chinese AI labs have largely remained silent on those broader risk discussions.

Analysts caution that capturing a higher share of revenue will require Chinese labs to either adopt monetization models that more directly capture usage value or build proprietary, closed offerings that lock in customers to paid services. Otherwise, the open distribution and lower cost-per-task characteristics that aid rapid adoption could continue to compress direct revenue. As Logan Wright, a Rhodium partner and co-author of the report, observed, the combination of heavy hardware investment support and limited direct funding for frontier labs makes long-term scaling dependent on favorable equity markets—a condition that is historically uncertain in China.

In sum, the Rhodium report highlights a clear tension: strong adoption and technological progress do not automatically translate into proportional revenue capture. The structure of distribution, product packaging, and market-channel strategies matter greatly. Until Chinese AI firms convert usage into robust recurring revenue streams at scale, their aggregate monetization will remain well below that of the largest U.S. players—despite rising prominence and fast growth in usage.

Key Insights Table

Afterwards...

Looking ahead, several areas merit close attention. First, the evolution of monetization strategies by Chinese labs will determine whether revenue growth can close the gap with U.S. competitors. Companies can pursue API-based, hosted services, enterprise contracts, or tiered licensing models to capture more recurring revenue. Second, developments in hardware and compute economics—particularly efforts to localize supply chains and reduce unit costs—will shape competitive positioning. Governments and state-affiliated investors are likely to continue supporting infrastructure, but the role of public markets in validating valuations remains critical.

Third, governance and distribution choices—open versus closed models—will have long-term implications for both innovation diffusion and commercial returns. Open approaches can accelerate adoption and ecosystem development but may depress direct revenues; closed approaches increase capture but can limit reach and integration. Finally, international dynamics, regulatory frameworks, and investor sentiment will all interact to influence which firms can sustain scaling and justify high valuations.

In the near term, monitoring updated ARR disclosures, IPO outcomes, and product-go-to-market shifts will provide the clearest signals about whether Chinese AI developers can convert momentum into sustained, large-scale revenue. The interplay between adoption, monetization design, and capital markets will largely determine which companies become long-term leaders in the global AI economy. Strategic choices about pricing, distribution, and product control are likely to be decisive.