DeepSeek Unmanned Entrepreneurship

Entrepreneurs who came out of DeepSeek are almost never seen.

However, a large number of young AI talents with top academic backgrounds are gathered here. Even if a small number of employees left before, most of them chose to join other large tech companies instead of starting their own businesses. In sharp contrast to this quiet situation, this year is almost the peak year for AI talents to leave large AI companies, and those who quit to start their own businesses can easily get a valuation of tens of billions of yuan.

"Do not be tempted by praise, do not be intimidated by slander, act in accordance with the Tao, and maintain a proper and upright state of self." DeepSeek quoted this saying when it released V4 in April this year to illustrate its values.

After that, along with DeepSeek's first financing, there were rarely any news of employee turnover coming out.

Now such cases will only become rarer. After the second round of financing, as a company with a valuation of more than 500 billion yuan, DeepSeek only has hundreds of employees, who can obtain "huge returns" through their stock options.

After the launch of financing, employees get DeepSeek stock options

Financing has always been one of the ways for startups to stabilize the morale of their team, and DeepSeek is no exception.

In the past six months or more, there has been very little news of core personnel leaving DeepSeek. But if we look back half a year before the launch of the first round of financing, the outside world witnessed successive changes in core researchers of this company.

These names are familiar: Luo Fuli joined Xiaomi in November 2025 as the head of Xiaomi Mimo Lab; Wang Bingxuan joined Tencent's Hunyuan team at the end of 2025; Ruan Chong joined autonomous driving company WeRide as chief scientist in January 2026; Guo Daya joined ByteDance's Seed Lab in April 2026 as one of the heads of Agent, and so on.

As we previously reported, based on the list of contributors to DeepSeek's papers, the turnover rate of its core researchers is actually not that high. Further observation shows that even if members like those mentioned above leave the company, most of them choose to switch jobs or return to universities rather than start their own businesses.

This should be related to DeepSeek's unique temperament itself.

Its internal organizational management is more centered on technological innovation. The research team generally only has two levels: Liang Wenfeng and researchers. Graduates from top universities are gathered here, and internal information flows at a high speed. In addition, there is no requirement for overtime work, no clock-in, and no clear performance appraisal. Liang Wenfeng's logic is: it is difficult for a person to work with high quality for more than 6 to 8 hours a day. The muddleheaded judgments made under the fatigue of overtime work will instead waste precious computing power resources, which is not worth the loss.

Under such circumstances, the research atmosphere at DeepSeek is more pure. It can allocate resources from top to bottom for a good idea put forward by young people, thus gathering many geeks with AGI ideals.

On the other hand, since the organizational hierarchy is simple enough, this company does not produce many executives or people in charge. Naturally, the outside world can hardly see any news of people in charge related to DeepSeek starting their own businesses.

Now after two rounds of super financing, the mobility of core employees from the DeepSeek system has become even rarer.

At present, looking through DeepSeek's public equity structure, although no clear employee stock ownership platform has been shown, there has been an internal employee stock option plan, and important and senior employees have obtained a considerable number and amount of stock options. "For a company with a scale of hundreds of people and a valuation of over 500 billion yuan, the core employees have undoubtedly already obtained huge returns." someone said with envy.

Liang Wenfeng asked not to poach people

There are more reasons that are tacitly understood by all parties.

I remember that at an investor meeting in the first half of the year, Liang Wenfeng clearly mentioned: On the road to AGI, maintaining team stability is the biggest risk and the core interest that cannot be compromised.

From the perspective of investors, facing DeepSeek, the protagonist of financing that makes the entire investment circle salivate, actively poaching people or instigating them to start their own businesses would inevitably seem unkind.

At the same time, after the two rounds of financing, a large number of well-known investors have gathered behind DeepSeek. As stakeholders on the same boat, they have no reason to undermine the internal stability of DeepSeek.

In particular, the investors participating in the second round of financing are more prominent, with the financing scale exceeding 100 billion yuan. We learned that up to now, the second round of financing of DeepSeek has not been completely finished, some GPs are still in the final stage, and there are occasional financing quotas circulating in the market.

Looking back at this bustling half year, in addition to supplementing a large amount of capital ammunition through financing, what impresses the outside world about DeepSeek is the team expansion strength that surpasses any previous year.

In May, DeepSeek recruited Cui Tianyi to form the Harness team, stepping onto the global Agent competition table. In June, DeepSeek officially released recruitment information, stating that it was working to expand the scale of all departments by at least double, and publicly released 33 job categories, which can be called the largest recruitment since its establishment.

In September, Cui Tianyi, head of the Harness team, publicly posted recruitment information on social platforms, releasing about 150 vacancies for server-side development engineers and Agent elastic computing R&D engineers. In the past October, DeepSeek integrated the Harness team, as well as the Web, App and Platform teams into the "Product Department", and urgently recruited product managers, designers and senior-level product managers.

The most sensational thing among them is that in late September, DeepSeek filled the long-vacant position of CFO, and Yan Wentao, the post-90s partner from GL Ventures, officially joined the company. Subsequently, DeepSeek began to make moves in the primary market, and in October, news came out that it invested two embodied startup companies.

In half a year, DeepSeek has achieved comprehensive acceleration in financing, organization and business. The next step is the IPO process — it is reported that DeepSeek plans to list on the Sci-Tech Innovation Board, and complete the listing as soon as 2027.

The wind is coming, those who leave, those who stay

Some are calm, while others are restless.

"Now is probably the best time in history to start a business." This sentence thrown out by Sam Altman in a recent interview is quite tempting. Facing the AI boom and opportunities, we have seen a year when talents from various large tech companies leave to start their own businesses.

This year there is already a long list:

Lin Junyang, former head of the Qwen large model, founded the AI company Pragmatic Tech in Shanghai; Zhang Qizhi, former head of Jianying China, founded OJO, an AI product prototype design workspace; Hu Han, former distinguished scientist of Tencent Hunyuan, founded Winverse AI, a company related to embodied intelligence; in addition, Huang Wenhao, former head of ByteDance Seed model evaluation, Zeng Xinxun, former head of AI search of Moonshot AI, Shi Lingxiang, former head of ByteDance Volcano Engine innovation incubation, and many others, have all left their jobs to start their own businesses.

Such figures are emerging in an endless stream. They often leave their jobs quickly, then start their own businesses quickly, and then quickly gather a group of investors.

As a group of people get high valuations, a number of people who have not left their jobs are also getting ready to make a move. "When some senior employees of large tech companies have the idea of starting a business, they will not resign directly. Instead, they will first find channels to indirectly inquire about prices from VCs, to understand how much they can probably get in the market. If it meets their expectations, they will resign to start a business; if it does not meet their expectations, they will continue to work in large tech companies with high salaries with peace of mind." a headhunter friend said so.

After all, the opportunity window will not last forever. The industrial tide and life opportunities are too tempting not to give it a try.

But the iron law that nine out of ten startups fail has never been invalid even in this booming era. AI has come to a delicate crossroads, from Agent, world model, AI4S, to RSI, Personal Agent, the next scene is evolving day and night. The survival factors that are real, specific, paid for by people and can continuously deliver results are ancient and cruel.

Not all those who venture out into the waves will reach the shore, and not all those who stay calm on the shore will miss the tide. No matter how bustling the era is, taking full responsibility for one's own choices is the most honest portrayal of entrepreneurship.

This article is from WeChat Official Account "Investment" (ID: pedaily2012), author: Feng Yuchen, published with authorization from 36Kr.