US AI Leaders Generate Ten Times More Revenue Than Chinese Models
According to a recent report by Rhodium Group published in September 2026, OpenAI and Anthropic, two leading US artificial intelligence companies, earn approximately ten times the revenue of all Chinese AI models combined. OpenAI alone generated $40 billion in annual recurring revenue (ARR), and Anthropic reported $65 billion, dwarfing the combined Chinese figures.
This significant revenue gap is calculated using ARR, an industry standard metric that annualizes recent monthly revenue data to capture fast-growing businesses’ financial performance. Despite rapid user adoption in China, Chinese AI firms’ revenues remain substantially lower, emphasizing the disparity between usage and monetization.
Chinese AI Model Revenues Lag Behind Despite Market Growth
Among Chinese AI companies, Z.ai leads with an ARR of $1.8 billion as of mid-2026, followed by ByteDance at $4 billion and Alibaba at $2.4 billion. Other firms like Moonshot and MiniMax report ARR figures of approximately $1 billion and $800 million respectively, while DeepSeek trails at $500 million.
Although Z.ai forecasts its ARR could reach $3 billion by the end of 2026, these numbers remain far below the US leaders. The slower revenue growth despite widespread AI adoption raises questions about the sustainability and valuation of Chinese AI startups amid intense competition and evolving business models.
Valuations Suggest Overvaluation of Chinese AI Startups
Rhodium’s analysis highlights that Chinese AI firms such as Moonshot and DeepSeek exhibit valuation-to-revenue ratios of 50x and 163x respectively, signaling potentially inflated valuations compared to actual revenue. In contrast, OpenAI and Anthropic have more moderate ratios of 34x and 21x.
These discrepancies indicate investor optimism in China may be disconnected from the underlying financial performance of these companies. Both Moonshot and DeepSeek are reportedly preparing for initial public offerings—Moonshot confidentially in Hong Kong and DeepSeek possibly elsewhere—where market scrutiny could test these lofty valuations.
Revenue Models and Market Dynamics Differ Sharply Between US and China
One key factor behind the revenue divide is the business model: US AI companies mostly operate closed models with higher costs per AI task, enabling more direct monetization. In contrast, many Chinese AI models are open-source, allowing third parties to run the software independently, which dilutes direct revenue capture.
Rhodium notes that Chinese AI labs are exploring ways to increase revenue shares from third-party access but face challenges scaling sustainably. Logan Wright, a Rhodium partner, points to heavy reliance on favorable equity markets and government support for hardware investments, while direct funding for frontier AI development remains limited in China.
Investment and Market Performance Reflect Uneven Prospects
State-affiliated sources provide over 60% of equity investments in Chinese AI hardware, underscoring government involvement in the tech ecosystem. However, public market performance of Chinese AI stocks has been volatile. Z.ai shares recently rebounded 5% after declines, while Minimax shares have struggled to maintain IPO gains.
Meanwhile, US AI stocks have faced pressure after industry leaders cautioned about risks from rapid AI development. Chinese AI firms have yet to respond publicly to these concerns, leaving market watchers to assess how regulatory and competitive dynamics will shape the sector’s future.
Takeaway: Despite rapid adoption, Chinese AI models generate only a fraction of the revenue of US leaders OpenAI and Anthropic, highlighting fundamental differences in monetization and market maturity.