Rhodium: OpenAI and Anthropic Generate Ten Times the Revenue of All Chinese AI Models Combined — Valuation Questions Follow

Table of Contents

• You might want to know


• Main Topic


• Key Insights Table


• Afterwards...

You might want to know

• Why do OpenAI and Anthropic report vastly higher revenues than the combined total of Chinese AI models?

• How do revenue figures relate to current valuations and the future scalability of Chinese AI startups?

Main Topic

Recent estimates from U.S.-based research firm Rhodium Group indicate a substantial disparity in reported revenue between leading American AI companies and the collection of Chinese AI model developers. Using an industry metric known as annual recurring revenue (ARR) — which projects yearly income by annualizing a recent monthly revenue number — Rhodium found that all Chinese AI models together generate roughly 10% of the revenue attributed to OpenAI and Anthropic combined. This gap highlights a divergence between rapid model adoption and direct monetization in China.

Rhodium's report lists ARR figures for several major Chinese AI companies. DeepSeek was estimated at about $500 million in ARR, MiniMax at $800 million, and Moonshot at $1 billion. Z.ai reported an ARR of $1.8 billion in a recent investor update, and larger internet firms such as ByteDance and Alibaba were estimated at $4 billion and $2.4 billion respectively. By contrast, Rhodium estimated OpenAI's ARR at $40 billion and Anthropic's at $65 billion, underscoring the scale difference between the U.S. leaders and their Chinese counterparts.

The disparity raises questions about how investors are pricing Chinese AI startups. Rhodium highlighted that valuation-to-revenue ratios appear particularly elevated for some Chinese firms: Moonshot and DeepSeek were estimated at roughly 50x and 163x revenue, respectively. Those multiples exceed the roughly 34x multiple Rhodium attributed to OpenAI and the 21x multiple for Anthropic. Such figures suggest that market optimism is pricing in future growth that current revenue levels do not yet justify.

Several Chinese startups have reportedly taken steps toward public listings. Anthropic is widely reported to be preparing a U.S. listing in the near term, while OpenAI has delayed its IPO plans until next year. In China, Moonshot has reportedly filed confidentially for a Hong Kong IPO, and DeepSeek is also said to be preparing a listing. Company representatives have either declined to comment or not responded to inquiries about those filings, and one firm stated it does not comment on market speculation.

It is important to note limitations in the Rhodium analysis. The report drew on figures available as of this past summer, and usage of Chinese AI models has accelerated substantially from earlier in the year. For example, Z.ai revised its expected ARR upward, saying it now targets $3 billion by year-end, up from a prior $2.4 billion projection. Nevertheless, the report flagged structural differences that could keep revenue growth constrained: many Chinese models are distributed in open-source form, which allows third parties with sufficient hardware to download and run them independently. That reduces the ability of original developers to capture downstream revenue compared with closed U.S. models.

Cost dynamics diverge as well. AI-comparison firm Artificial Analysis has noted that U.S. models tend to be closed and generally incur a higher cost per task compared with leading Chinese models. This pricing differential helps explain the larger reported revenues for U.S. firms despite differences in global usage patterns.

Rhodium's analysts emphasized that the revenue shortfall relative to valuations could make sustainable scaling more challenging for Chinese frontier AI labs. Logan Wright, a partner at Rhodium, observed that these firms will be sensitive to conditions in equity markets — a historically uncertain prospect in China — and that government support has largely favored hardware buildout rather than direct operating subsidies for frontier AI labs. The report estimated that more than 60% of equity investment in Chinese AI chips and servers came from state-affiliated sources, signaling strong public-sector involvement in infrastructure, if not in operating budgets for model developers.

Market reaction has been mixed and volatile. Shares of some Chinese AI companies that have listed experienced sharp moves after fundraising news and changes in investor sentiment. For instance, Z.ai's shares recovered after declines linked to recent fundraises, while Minimax has seen difficulty sustaining IPO-day gains. Meanwhile, U.S. tech stocks slipped after executives of major American AI firms warned about risks associated with rapid development of the technology. Leaders of leading Chinese labs have not offered similar public commentary.

In sum, Rhodium's snapshot suggests that, despite fast adoption and active investment, Chinese AI models as a whole still trail U.S. peers significantly in revenue generation. That gap matters for how investors value companies and for the practical ability of labs to scale computing, talent, and commercial ecosystems without continued favorable financing conditions.

Key Insights Table

Afterwards...

Looking forward, critical areas for exploration include better mechanisms for monetizing model deployment, governance around open-source distribution, and scalable funding models that balance public infrastructure support with private operating capital. Policymakers and industry leaders should also consider standards for cost transparency, model pricing, and revenue-sharing arrangements when models are deployed by third parties. Continued investment in compute infrastructure is essential, but equally important are strategies to convert usage into sustainable revenue streams while maintaining competitive pricing.

Additionally, cross-border comparisons of ARR and valuation metrics would benefit from standardized reporting practices. Improved transparency around revenue recognition, customer concentration, and per-task pricing would help investors and regulators assess growth trajectories more accurately. Finally, research into cost-efficient inference, differentiated product offerings, and partnerships that align incentives across model creators and distributors could materially influence how the market values emerging AI leaders.

These directions — monetization design, transparent metrics, and sustainable funding — are likely to shape which labs can scale commercial AI responsibly and profitably in the coming years.