MarTech & AI

Chinese AI Companies Growing by Hundreds of Percent, Yet Earning 10% of OpenAI and Anthropic's Revenue

Seven Chinese AI companies earned ~$10.7 billion in ARR — about 10% of OpenAI and Anthropic's revenue. ByteDance reached $4 billion, and Z.ai grew 400% in half a year.

2 min read

Китайские ИИ-компании растут на сотни процентов, но зарабатывают 10% от выручки OpenAI и Anthropic
Китайские ИИ-компании растут на сотни процентов, но зарабатывают 10% от выручки OpenAI и AnthropicUK Prime Minister / Openverse

The seven largest Chinese AI developers collectively earned about $10.7 billion in annual recurring revenue (ARR) between March and August — just 10% of the figures for OpenAI and Anthropic, which together brought in more than $100 billion. These figures come from SCMP, citing a study by the Rhodium Group.

The leader among Chinese companies was ByteDance: $4 billion in ARR as of July. Alibaba took second place with $2.4 billion in August. For comparison: OpenAI reached $40 billion in ARR in August, while Anthropic hit $65 billion in July.

The worst figures among the tracked companies belonged to DeepSeek and Kling: $500 million each as of June and March, respectively. MiniMax reached $800 million, and Moonshot reached $1 billion (both figures as of August). Z.ai told investors it had $1.8 billion in ARR, of which $1.6 billion came in August.

Growth Rates Measured in Hundreds of Percent

The modest absolute revenue is offset by the speed of growth. Z.ai grew its annual revenue by 400% in the first half of the year, year over year — reaching $6.3 billion. MiniMax showed growth of 283%, to $116.6 million.

At the same time, Macquarie analysts allow that Z.ai and MiniMax may remain unprofitable until 2030 due to spending on computing power. Investors are so far willing to tolerate the losses: AI adoption is at an early stage, and the market is pricing in growth for future periods.

China's Leading Neural Networks

Medium ranks DeepSeek, Qwen, Kimi, Doubao, GLM, and ERNIE among the strongest Chinese AI models as of July 2026.

Positions in the rankings depend on the type of task: general logical reasoning, coding, working with long context, knowledge of the Chinese language, or multimodal scenarios. In OpenCompass' weekly ranking from mid-July, the leaders in processing large volumes of data included DeepSeek-V4-Pro, Doubao-Seed-2.0-Pro, Kimi-K2.6, and Qwen3.6-Max-Preview. In the multimodal segment, Alibaba's Qwen stood out, along with Kimi, Doubao, SenseNova, GLM, and Step.

The key takeaway is not about one ranking's dominance. Several Chinese companies have come very close to one another: first place can change depending on the task, the model release cycle, or pricing.

Source: AdIndex; Source: scmp.com

Share this article:

More from Артём Пасечник

Related articles

« Previous articleNext article »