發生了什麼事
榮鼎集團發布了一份報告，分析了中國七家主要人工智慧開發商的財務業績，包括DeepSeek、Moonshot AI、Z.ai、MiniMax、阿里巴巴、字節跳動和快手。报告估计，这些公司在 3 月至 8 月期间的年度经常性收入 (ARR) 合计约为 107 亿美元。相較之下，OpenAI 和 Anthropic 同期的合併 ARR 超過 1,000 億美元。該報告強調了估值與收入之間的巨大差異，指出 DeepSeek 和 Moonshot 等中國公司的 ARR 估值倍數分別為 163 倍和 50 倍，而 OpenAI 的估值倍數為 34 倍，Anthropic 的估值倍數為 21 倍數為 21 倍數。儘管收入快速增長，Z.ai 報告同比增長 400%，但該報告表明，開放權重策略使中國開發商的變現工作變得更加複雜。
美國研究公司榮鼎集團週四發布了一份報告，詳細介紹了中國七家主要人工智慧開發商的財務指標。研究計算得出，包括DeepSeek、Moonshot AI、Z.ai、MiniMax、阿里巴巴集團控股、位元組跳動和快手科技在內的這些公司3月至8月的年度經常性收入（ARR）合計約為107億美元。
這一數字約佔 OpenAI 和 Anthropic 報告的合併 ARR 的 10%，同期 ARR 超過 1,000 億美元。具体而言，OpenAI 8 月份的 ARR 达到 400 亿美元，而 Anthropic 7 月份的 ARR 为 650 亿美元。在中國開發商中，位元組跳動截至 7 月的 ARR 為 40 億美元，位居榜首，阿里巴巴緊隨其後，8 月為 24 億美元。
The report highlights a stark contrast in valuation relative to revenue. Rhodium noted that the valuation-to-ARR multiples for DeepSeek and Moonshot stood at 163 times and 50 times, respectively. In comparison, OpenAI’s multiple was 34 times and Anthropic’s was 21 times. The report described these Chinese valuations as 'exorbitant' relative to their current revenue streams, even though they remain substantially lower in absolute terms than US counterparts.
Despite the revenue gap, Chinese AI companies are experiencing rapid growth. Z.ai reported a 400% year-on-year increase in first-half revenue to 953.9 million yuan, while MiniMax’s revenue surged 283% to US$116.6 million. However, Macquarie Group’s head of Asia internet and software research, Ellie Jiang, told the South China Morning Post that Z.ai and MiniMax could remain loss-making through 2030 due to the high costs of computing power required to maintain technological competitiveness.
為什麼這很重要
該報告為美國和中國人工智慧產業的競爭格局提供了具體的財務基準。报告显示，尽管投资积极且估值较高，但中国人工智能公司目前的收入仅为美国同行的一小部分。这种差距引发了人们对当前估值的可持续性以及开放权重策略在产生直接收入方面的有效性的质疑。對於投資者和政策制定者來說，這些數據更清晰地描繪了技術競賽背後的經濟現實，突顯出雖然中國模式正在縮小績效差距，但支持它們的商業模式仍處於成熟階段，與 OpenAI 和 Anthropic 的閉源、訂閱密集型模式相比，面臨著明顯的貨幣化挑戰。
The report underscores a central challenge for China’s AI industry: converting technological parity into sustainable business models. While Chinese developers have rapidly narrowed the performance gap with US counterparts, their revenue generation lags significantly behind. This discrepancy suggests that the current investment cycle in Chinese AI is driven more by strategic positioning and market share acquisition than by immediate profitability.
A key factor complicating monetization is the widespread adoption of open-source and open- models by Chinese developers. Rhodium noted that this strategy makes it difficult for original developers to capture revenue, as third-party cloud providers can deploy these models and sell access without paying the creators. This stands in contrast to the closed-source, subscription-based models of OpenAI and Anthropic, which allow for more direct revenue capture.
The financial data provides a critical for investors evaluating the risk profile of Chinese AI stocks. The high valuation-to-revenue multiples indicate that investors are pricing in significant future growth or strategic value that is not yet reflected in current earnings. This creates a potential vulnerability if growth slows or if the cost of maintaining technological leadership continues to outpace revenue growth.
The report also highlights the evolving nature of AI business models, particularly regarding open- strategies. As Chinese firms like Moonshot and Alibaba push for revenue-sharing agreements with major users, the industry may see a shift in how open-source AI is monetized. This could have broader implications for the global AI market, potentially influencing how other developers approach open-weight releases.
互動機制：它實際上是如何運作的
以互動方式探索這項發展背後的基礎技術。
Which component of an AI application is the machine-learning model itself?
接下來看什麼
正如荣鼎所讨论的那样，投资者应关注中国人工智能公司是否能够成功实施其开放权重模型的收入分享协议。此外，Moonshot AI 即将进行的 IPO 以及其他主要参与者的潜在上市将考验市场对这些高倍数估值的兴趣。這些公司在不產生不可持續的計算成本的情況下將快速用戶採用轉化為可持續盈利能力的能力將是中國人工智慧行業長期生存能力的關鍵指標。
The implementation of revenue-sharing agreements for open- models will be a key indicator of whether Chinese AI companies can improve their monetization efficiency. Rhodium noted that Moonshot is in discussions with Microsoft, Amazon, and Google, with potential revenue shares reaching up to 30%. The success of these negotiations could serve as a model for other developers.
Upcoming initial public offerings (IPOs) for Chinese AI firms, such as Moonshot AI’s confidential filing for a Hong Kong IPO and DeepSeek’s pre-IPO funding round, will test market confidence in these high valuations. The reception of these listings will provide insight into whether investors are willing to sustain high multiples in the face of limited current revenue.
The ability of Chinese AI companies to manage computing costs while maintaining technological competitiveness will be crucial. As noted by Macquarie Group, the high cost of computing power could keep companies like Z.ai and MiniMax in the red through 2030. Monitoring their cost structures and efficiency improvements will be essential for assessing long-term viability.
The competitive dynamics between US and Chinese AI developers will continue to evolve, with both sides competing on price, performance, and open-source adoption. The report’s findings suggest that while the performance gap is narrowing, the economic gap remains significant. Future developments in pricing strategies and model releases will be key to understanding how this dynamic plays out.