Interview with Ma Yue of OSChina: DeepSeek is not a once-in-a-generation innovation; young people are.

Text | Deng Yongyi

Editor | Su Jianxun

If the discussion about DeepSeek has become overly prevalent, then open source might still be a topic worthy of discussion at present.

For a long time, when talking about "open source" in China, one will inevitably fall into an awkward context.

It is, of course, idealistic. The freedom and openness behind "open source" are generally regarded as the externalization of the Internet spirit - the source code is open and shared with the public, and users are allowed to freely use, modify, and redistribute the software under the premise of following specific license terms.

The most well - known open - source project, "Linux", is the kernel of the operating system, which has spawned tens of millions of open - source software. It is the foundation of the Internet world.

But there is often a question behind it: Why open source? How to consider commercialization after open - sourcing? Even now when DeepSeek has become extremely popular, it's still difficult for anyone to give a perfect answer.

Ma Yue, the chairman of OSChina, is one of the people most qualified to talk about the history of open source in China. He has been on this path for 18 years.

In 2008, Ma Yue returned to China from Silicon Valley to start a business. First, he founded "Hengtuo Open Source" - using open - source software to help enterprises break free from the constraints of large - scale software such as databases and ERP.

But soon he found that this solution was difficult to get rid of the heavy investment of the To B project system and was likely to turn the company into an outsourcing company.

Subsequently, Ma Yue chose to acquire the "OSChina" community and started a tortuous entrepreneurial journey. OSChina has undergone several transformations, expanding from an open - source community to code hosting and code toolchains. During the exploration of commercialization, it was separated from its parent company and developed independently. In 2019, it was strategically controlled by Baidu. Finally, in the wave of Sino - US competition and domestic substitution, it decided to develop independently again and seek listing.

Running an open - source community requires a large amount of resources and capital investment. At the most difficult time of OSChina, Ma Yue's personal debt reached a maximum of 180 million yuan.

Ma Yue, born in 1972, has an elder - brother - like frankness. He has absolutely no intention of complaining about hardships, but it's easy for you to feel the humor after experiencing these setbacks from his narration. He said that doing To B business in China is like "encircling the countryside from the cities", and enterprise software is the solidification of managers' wisdom. When Chinese enterprises are still in the initial stage of development, "it's hard to ask people to engage in open source when they haven't had enough to eat and drink."

But those difficult times are over. OSChina has also explored a more suitable Chinese - style open - source path for itself.

Now, OSChina has become the second - largest code hosting platform in the world, bringing together more than 18 million developers. Its self - developed DevOps toolchain has achieved an 80% market penetration rate in key fields such as finance and the military industry. In 2024, OSChina's revenue exceeded 200 million yuan.

According to "Intelligent Emergence", OSChina recently officially completed a Series C financing of hundreds of millions of yuan, led by the Beijing Information Industry Development Investment Fund (Beijing Information Industry Fund), followed by the Shenbao Yiben Equity Investment Fund (Shenbao Yiben) and the Beijing Shanghe Momentum Private Equity Fund (Shanghe Momentum).

So far, OSChina has received more than 1.6 billion yuan in strategic investment.

△ Ma Yue, the chairman of OSChina. Image source: OSChina

Ma Yue believes that even on a global scale, open source is not an easy thing. He took GitHub as an example: From its establishment in 2008 until it was acquired by Microsoft, and after the emergence of ChatGPT in 2022, it launched Copilot, which officially proved its commercialization potential.

"Open source is a game for the strong and the rich." He said that the previous generation grew up in an era with more material shortages - so did the business society. Enterprises need to make enough money first before they have the luxury to consider whether to open source and do some good deeds of "everyone for me, and I for everyone." "Only when you are full can you have the energy to talk about open source."

It's not hard to understand that even though the popularity of DeepSeek has given a boost to the whole of China, Ma Yue's view remains calm. He believes that DeepSeek is difficult to fundamentally change the problems in the domestic software ecosystem, which is a limitation of the era.

And if you want to achieve something in the open - source route, it requires a new generation of developers to go global from Day 1 and compete in the global market like DeepSeek.

If DeepSeek has changed anything, it's mostly in terms of culture and values. "Ten years ago, people generally didn't understand open source and thought it was something done by a group of grass - roots people. Now the whole society has realized that open source equals innovation." Ma Yue said.

The following is an edited dialogue between "Intelligent Emergence" and Ma Yue, the chairman of OSChina:

DeepSeek is not a product of national fortune; young people are

"Intelligent Emergence": OSChina is now the second - largest code hosting platform and the largest open - source community in China. When did the direct impact of the DeepSeek craze on you start?

Ma Yue: It was when it topped the App Store. First, V3 was released, and then R1, and it suddenly became popular. We worked overtime during the Spring Festival to make DeepSeek run on Chinese - made GPUs first, which required a lot of work. We were the first to deploy it on the Chips from Moore Threads.

We were all joking that there were only two things during the Spring Festival: DeepSeek and Nezha. DeepSeek is like Nezha emerging in the open - source circle.

"Intelligent Emergence": Now there is a view that DeepSeek is a product of national fortune. But in your recent public statements, you don't seem to agree with this.

Ma Yue: First of all, in the field of large models, you can't do without NVIDIA, right? Secondly, you can't do without the Transformer architecture; thirdly, you adopted the idea of distillation, and none of these are original in China.

DeepSeek is essentially the best on the existing route and has achieved a curve - overtaking, which is worthy of respect.

But the fact that DeepSeek can achieve global top - level results with technology alone without relying on external financial support and without doing any PR - the rise of young people represented by Liang Wenfeng is a phenomenon of national fortune.

"Intelligent Emergence": Then what can be regarded as a product of national fortune?

Ma Yue: Completely original technological innovation. Who says Transformer is the end - game of algorithms? If someone uses a non - Transformer solution to achieve results ten times stronger than DeepSeek, that would be a real breakthrough and a human - level progress.

"Intelligent Emergence": What's the biggest inspiration that DeepSeek can give to the open - source ecosystem?

Ma Yue: It makes the whole society realize that open source equals innovation.

What's most regrettable about DeepSeek is that it was unknown in China for two years and was not as well - known as many large - model companies that spent a lot on advertising. It wasn't until 2024 that it was super - focused on by Americans because of its technology and open source. Although some people strongly support it and some strongly oppose it, this kind of attention has instead forced the formation of a patriotic sentiment in China.

"Intelligent Emergence": Didn't people believe this view before?

Ma Yue: Before, many people thought that open source was just a bunch of grass - roots people and a mob, and it was hard to compare with regular troops like large companies.

Actually, I've been saying these things twenty years ago: Open source is approximately equal to innovation ability, and innovation ability is mapped to national strength. It's precisely because we are rich and prosperous that enterprises like DeepSeek have emerged.

No one listened before, but now someone does.

The second important point is to have awe of young people. It's not just respect, but also fear and trust. Each generation has its own mission and limitations of the era.

"Intelligent Emergence": Why can this group of young people, or the new generation of open - source contributors, grow up?

Ma Yue: This qualitative change is based on sufficient quantitative changes before.

We should thank the leading Internet giants in the past decade. In fact, the main open - source forces in China are concentrated in these powerful enterprises. Including the open - source projects organized by Baidu, Alibaba, Tencent, etc., and Huawei's HarmonyOS, Euler, etc. Their employees, who are paid salaries, are engaged in these open - source works, not purely based on interest.

"Intelligent Emergence": The most crucial thing that DeepSeek proves is that through breakthroughs in underlying technology, it can attract a large number of users and gain respect.

Ma Yue: Now our country should do two things: One is to take the lead in jointly developing China's CUDA; the other is to make all domestic GPUs quickly support these models.

When it comes to the ecosystem, an ecosystem requires more people to participate, and everyone should have a high - level consensus. The biggest problem now is not the bottleneck in chips, but the restriction of the CUDA ecosystem. China can completely develop a system similar to CUDA, just like we have our own GPUs.

Open source is a game for the rich and the strong

"Intelligent Emergence": After the popularity of DeepSeek, how many people came to discuss with you? What topics are they most concerned about?

Ma Yue: Some people asked me if DeepSeek would bring new vitality to the Chinese To B market. It's impossible that fast.

The per - person - day price of IT outsourcing has increased less than that of massage therapists in the past 20 years. Now, an average per - person - day price of 1,000 yuan for outsourcing is considered high, and there are also those at 500 - 600 yuan. When you go for a massage, it costs 100 - 200 yuan per hour now. Ten years ago, the hourly wage of IT outsourcing was already lower than that of massage, and now the gap is even bigger because the price of massage has risen rapidly.

No one in the Chinese software industry is willing to spend money, which is the core reason for the under - development of the industry. It won't be a good time until this generation of young people become decision - makers.

"Intelligent Emergence": Essentially, it's still because Chinese enterprises are still in the early stage of development.

Ma Yue: Open source is essentially a game for the strong and the rich. It's precisely because we are well - fed that enterprises like DeepSeek have emerged. The previous generation of Internet users were generally reluctant to pay for software and knowledge. They would just reconnect when Tencent Meeting dropped, and were not willing to buy a membership.

But this generation of young people live in affluence, and you will change this situation.

"Intelligent Emergence": What inspiration will DeepSeek bring to the previous generation of To B entrepreneurs?

Ma Yue: I think it brings two important inspirations to entrepreneurs. First, have awe of money. The purpose of entrepreneurship is to make money. Talking about ideals and feelings doesn't make sense.

DeepSeek doesn't need to worry too much about commercialization because Magic Square has solved this problem.

The previous generation of software entrepreneurs had a fatal problem. They were bent on burning money and trying to capture the market with standardized products. This is not the operating logic of the Chinese market. The richest clients in China are large enterprises. In China, it's unrealistic not to do customized development if you want to make money.

The Chinese software industry is about "encircling the countryside from the cities", while in the United States, it's about "encircling the cities from the countryside", and there are many mid - sized enterprises.

"Intelligent Emergence": Will DeepSeek change people's views on commercialization? How to consider commercialization in open source is the "ultimate question" in this field.

Ma Yue: If you want to start an open - source business project, the technology must be rock - solid. Just like DeepSeek, go global from Day 1. Otherwise, it's too hard to make money in China, and the era is not mature enough.

People always give examples, such as the Red Hat model can be commercialized. But it's not the right time to use this way to build a listed company in China.

"Intelligent Emergence": You also went through a long period of commercialization exploration. When did you figure out what to do?

Ma Yue: 2020 was an important turning point. We decided to separate from Baidu and seek an IPO again that year. At that time, because the United States started to cut off supplies in many cutting - edge technologies, we wanted to seize this opportunity to truly become an independent open - source platform.

If you want to build a truly local open - source platform, it must be a completely neutral third - party. This is the core reason for choosing to develop independently again.

"Intelligent Emergence": After figuring it out, what did you do?

Ma Yue: Now we have developed three major product lines from the community.

The OSChina community has now completely evolved into an AI education platform. We are the largest open - source community in China, with more than 10 million users. Now our 24 - person team can generate about 50 million yuan in revenue and also make a profit, which is rare among community teams.

The second is the code hosting and R & D efficiency platform Gitee. Now the platform has 36 million code repositories and serves 360,000 enterprises. It mainly provides private code hosting warehouse services to ensure the code security of many small and medium - sized teams, with an average annual customer order of about 3,000 yuan.

From 2020 to now, we have been able to provide a full - life - cycle domestic substitution solution for DevOps. While meeting the needs of developers, we have also established a self - innovative, secure and reliable local open - source software tool and ecosystem.

The third is the AI large - model platform "Moli Ark", which provides services such as model experience, inference training, and application deployment.

"Intelligent Emergence": Why did you expand from the community to DevOps and the AI large - model base later?

Ma Yue: For an open - source company to succeed, relying on the community is not enough. We need to find a closed - loop business model, like the path of GitHub. The community and code hosting alone can't achieve this goal. GitHub also didn't boost its revenue until it launched Copilot after the wave of large models.

In the future, it will be difficult for companies without net profit to be listed in China. So I've always emphasized looking at gross profit margin and human efficiency. If these two indicators are high, there will naturally be net profit.

"Intelligent Emergence": Where does your main income come from now?

Ma Yue: Our main income comes from about 100 customers in the banking, securities, military, and manufacturing industries, all in the form of large - scale private deployments. Small and medium - sized customers mainly rely on SaaS services.

In 2024, our national orders exceeded 200 million yuan, which was a breakthrough. The year before last, it exceeded 100 million yuan, and in 2024, it doubled, and we also achieved break - even, which is not easy.

"Intelligent Emergence": If the main model relies on serving large enterprises, how can you avoid falling back into the old path of the project system?

Ma Yue: Our products are designed