Fernando Capeto illustration for Forbes; Ruby / Noun Project icons
The fraternal rise of Deepseek highlighted the artificial intelligence of China. Here are the other buzzing Chinese AI companies to watch.
InJanuary, A little -known Chinese AI laboratory called Deepseek rocked the world when he published an advanced open source model that rivaled with those of us, apparently using a fraction of their resources. A who’s who du Juggernauts d’Ai, from Openai to Anthropic, praised the achievements of the company while defending its own progress and methods. President Donald Trump described this as “alarm clock”.
But beyond this viral moment, turned by the own cycle of sparkling media threshing and a frenzy of geopolitical concerns, the emergence of Deepseek had a deeper impact: he put the spotlight on Chinese AI and gave him a face on an international scene. Deepseek and other Chinese companies have not been included on the IA 50 list, which honors the most promising private companies in artificial intelligence, because their finances and their business practices are opaque. But they are worth emphasizing that many of them have significant impacts beyond China, thanks to the emphasis on open source models made freely available so that anyone can use.
Forbes
Forbes 2025 AI 50 List
The list of artificial intelligence of forbes 50 of 2025 highlights promising companies focused on AI. See the leaders at the origin of the future of the industry.
Show the full list
Many Chinese AI models winning the field are manufactured by country technology giants. There is the application of generation of Hanyuan videos, owned by the Tencent technological conglomerate, the manufacturer of $ 92 billion (2024 income) of WeChat. The company claims that its recent “reasoning” models, which can answer complex questions by dividing them into smaller sub which surpass the flagship models of Deepseek. Then there is Doubao, an application focused on the consumer of Tiktok Parent Bytedance, who built space models that analyze physical environments and generate 3D landscapes. There is also Qwen, a family of large-language models from the Alibaba electronic commerce giant, which has accumulated more than 90,000 business users on the company’s cloud platform.
“Alibaba is in a way the main champion of Big Tech AI in China, comparable to what Google or Meta represents in the United States,” said Rob Toews, partner of Radical Ventures, said Forbes. During the publication, the models of Alibaba and Deepseek were among the 5 best trend models on the face of cuddles, a center widely used for open source AI models and data sets.
“In the field of open source versions, there is no barrier. There is no big firewall.”
The meteoric rise of Deepseek opened the valves to other Chinese startups. In March, the Butterfly Startup effect, based in Wuhan, launched an AI system called manus, which, according to him, can independently browse the web and do things like looking for apartments, analyze actions and design websites. The tool has some traps, ranging from the creation of incorrect hypotheses on the task to be performed to block while treating large amounts of text. But his release was praised as an emerging rival of the OpenAi service, the operator. Food by tastes of the co -founder of Twitter Jack Dorsey, buzz around the company has attracted the interests of investors: the butterfly effect would be in talks to raise American investor funds to an evaluation of $ 500 million, according to Information.
While Deepseek has built its models in finetuning other Meta and Alibaba models, Manus managed to use the Claude of Anthopic models out of the shelf. “It is essentially a much better version of what Optai tries to do with the operator,” said Toews.
China has also made significant jumps in humanoid robotics. Agibot, founded in 2023 by the former Huawei recruit “Genius Youth” Peng Zhihi, says that he has already made more than 1,000 bipedal robots fueled by the AI and plans to have planned to pass this number to 5000 robots by the end of the year, in order to match Elon Musk’s plans for the Tesla general robot. Earlier this month, the company hired Luo Jianlan, who previously worked at Google X, the Moonshot factory in Alphabet, to direct its research efforts: “I hear about it when I speak with my robotics companies and they talk about competition they see in China,” said Aaron Jacobson, partner of the Capital Capital Company Nea. “This one presents itself several times now.”
The eminent investor and expert in AI, Kai-Fu Lee, who helped Google and Microsoft to create their outposts in China, founded an AI startup called 01.AI in 2022. He recently pivoted his own open source models to the use of the Deepseek AI to build business applications in fields such as the game, law and finances Deepseek. The company has raised around $ 200 million to an evaluation of $ 1 billion, according to Pitchbook. It is one of China’s “six tigers”, an elite group of better AI companies in the country which includes the multimodal developer IA Minimax AI and Model Maker Moshot Ai, who both attracted Alibaba’s investments.
The progress of China AI comes as political relations with the United States are becoming heavier. In 2022, the president of the time, Biden published export controls on semiconductor companies, including manufacturers of crucial fleas like NVIDIA and AMD, restoring sales of their most powerful equipment in the country, aimed at strangling the growth of China AI. Now, a trade war has broken out between the two nations created by the prices of President Trump’s “Liberation Day”, the United States imposing 125% tasks on exports.
“It is the power of a centralized state that can say:” We will try to go in this direction. “”
The Gains of China in AI are partly motivated by an emphasis on academic research and open source publication in universities, Russell Wald, Executive Director of the Institute of Artificial Intelligence Centered on Humanity of the University of Stanford (HAI), said Forbes. In 2018, China said it wanted to be the AI leader by 2030 and gathered important academic resources for this objective. Now, the country produces most world research on AI: from 2023, it represented around 70% of all patents granted and produced 23% of IA publications and quotes. “It is the power of a centralized state that can say:” We will try to go in this direction “,” said Wald. (The disadvantage, however, is that censorship in AI models by the Chinese government could repel Western users, he added.)
Until now, the strategy works. Earlier this week, Hai published its annual AI index, which shows the AI breed between the narrowing of the United States and China. The United States still produces most of the world’s cutting-edge AI, American companies releasing 40 “notable models”, defined as “particularly influential models in the IA / Machine Learning ecosystem”. China, in second place, published 15 of these models. And the country fills the gap in the performance of the model: two years ago, the United States led by two-digit points on various reference tests. Last year, China reached “near parity”, according to Hai.
The country’s open source approach, which allows anyone to download the model and create applications with it, has made Chinese companies easier to have a global impact. For decades, the United States and China have had very different technological ecosystems, with products and services only available in one region or the other. But Deepseek’s success has shown the AI community that Chinese laboratories could get around these restrictions with an open publication. “In the field of open source outings, there is not such a barrier,” said Jeff Boudier, responsible for product and growth in Hugging Face, told Jeff Forbes. “There is no big firewall.”
Meanwhile, given the strong culture of the country of AI talents in its universities and support for R&D, Wald sees a lot of promises not only in Deepseek and its current cohort, but what follows them. “What I think is more interesting is the company you don’t know at the moment, and you might hear about a year or two from now on.”
More Forbes
