DeepSeek Confirms New Chief Financial Officer.

Hillhouse Venture Capital’s Post-90s Talent Lands at DeepSeek.

On September 14, multiple media sources reported that Yan Wentao, a partner at Hillhouse Venture Capital, is set to assume the role of CFO at DeepSeek.

It is understood that Yan has already started the resignation process at Hillhouse and, barring any unforeseen issues, is expected to officially join DeepSeek shortly. If confirmed, this will represent a significant non-technical executive appointment for DeepSeek since its inception.

The news has garnered market attention because, just five days prior, Reuters disclosed that DeepSeek had chosen CITIC Securities to facilitate its STAR Market IPO.

With the CFO position now confirmed and the lead underwriter secured, it is evident that DeepSeek, a company that has revolutionized the global AI landscape with its model capabilities, is systematically completing the necessary capital arrangements for listing as it transitions from a technology laboratory.

The background of the incoming CFO is noteworthy. Yan Wentao does not hail from a conventional finance background; his career has been deeply entrenched in the investment sector. How did a young partner with an investor background secure this critical role?

A Formidable Post-90s VC

Yan Wentao has already established himself as a prominent figure in the venture capital arena.

Born in 1991 and a graduate of Fudan University, Yan worked at Tencent Investments and H Capital from 2013 to 2020 before joining Hillhouse Venture Capital in 2020. He has since emerged as a representative of Hillhouse's new generation of investors and was recognized among the Top 50 Nova Investors by China Venture in 2025.

Yan's investment portfolio is impressive, including ByteDance, Xiaohongshu, MiniMax, Zhipu AI, J&T Express, and Webull. This year in July, Yan revealed at the 14th Venture Capital Conference that he had been studying large language models since 2020 and continued to track the embodied AI sector in 2023, positioning him as one of the few investors in China with a deep understanding of both large models and hard tech.

Yan's industry insights, cultivated over years in venture capital, are precisely the attributes DeepSeek sought in a CFO.

In fact, DeepSeek had been seeking a CFO for some time. As early as February 2025, the company began an external recruitment drive for the position; as of June this year, the CFO role remained unfilled on its official website.

Also in June, DeepSeek secured its first external funding round, raising approximately 50 billion yuan, with Liang Wenfeng personally contributing around 20 billion yuan. Subsequently, in August, during its second funding round, DeepSeek's valuation soared to 500 billion yuan.

While capitalization efforts surged ahead, the key executive responsible for capital operations was notably absent. Compared to domestic counterparts, DeepSeek was relatively late in recruiting such high-level talent.

For instance, Hillhouse-affiliated investor Xue Zizhao joined MiniMax in 2023 to oversee investment, financing, and capital market operations; former GSR Ventures Managing Partner Zhang Yutong joined Yuezhiyanmian in April 2024 as President, overseeing financing and commercialization.

Most peers had secured such capital-related executives early in their funding cycles, whereas DeepSeek waited until the IPO sprint stage to appoint a CFO with an investor background. What considerations underpinned this choice?

The answer lies in DeepSeek's current developmental needs. Today's DeepSeek requires more than just a financial leader proficient in accounting, internal controls, and financial reporting—it needs a versatile talent familiar with financing, shareholder relations, capital market regulations, and capable of understanding the strategic resource allocation of a rapidly growing tech enterprise.

Cases of investors transitioning to corporate executive roles are not uncommon. In 2024, former Capital Today partner Dai Lidan joined Xiaohongshu as CSO, leading a new strategic investment team focused on hard tech and AI applications; in September this year, ByteDance recruited Jiang Kai, a senior executive from overseas firm Coatue, to head its Hong Kong financial investment team.

These personnel shifts reflect a broader industry trend: as AI becomes a core industrial variable, investors are no longer merely external capital providers but are also stepping in to participate deeply in company operations.

While the market debates the CFO appointment, DeepSeek's talent mobility over the past year has also become a focal point of external attention.

Behind the Rumors of Talent Exodus at DeepSeek

Over the past year, rumors of a mass exodus of core talent from DeepSeek have persisted, but the reality is less dire.

Since 2025, DeepSeek has experienced the departures of key personnel such as Luo Fuli, Wang Bingxuan, Wei Haoran, and Ruan Chong. Many of these individuals have since assumed core business leadership roles at their new employers:

Luo Fuli joined Xiaomi to lead the MiMo large model; Wang Bingxuan joined Tencent's Hunyuan team; Guo Daya joined ByteDance's Seed team to lead Agent-related R&D; Ruan Chong became Chief Scientist at Yuanrong Qixing; the whereabouts of Wei Haoran, a core member in the OCR direction, have not been publicly disclosed.

In early 2026, as news of Guo Daya's move to ByteDance's Seed team spread, discussions about "DeepSeek's brain drain" reached a fever pitch. Coupled with the delayed release of the V4 model at the time, many worried about a potential gap in DeepSeek's R&D capabilities.

However, public data tells a different story. In late April this year, DeepSeek-V4 was officially released, accompanied by a 58-page technical report that included a list of nearly 300 researchers acknowledged for their contributions, with 10 marked as having left the company. This translates to a core R&D team attrition rate of less than 4%.

For comparison, OpenAI has lost over 25% of its core research talent in the past two years, with many moving to competitors like Meta or starting their own ventures.

Extending the timeframe and analyzing 27 technical papers published by DeepSeek from January 2024 to April 2026 reveals key insights.

The team has steadily expanded: when DeepSeek-LLM was released in January 2024, the papers listed only 86 authors; by January 2025, DeepSeek-R1 had grown to 194 authors; and by April 2026, with the V4 release, the author count reached 317. Over 27 months, the team grew from under 100 to over 300 members.

Core personnel retention has also been notable. Among the 15 most frequent core researchers in the 27 papers, 13 remain employed; of the initial 86 team members, 71 still appear on the V4 author list, representing an early team retention rate of 82.6%, consistently maintained at 70-80%.

Across all 27 papers, 391 authors have been listed, with only 25 departures, approximately 6.4%.

Comparing domestically, ByteDance's Seed papers have a cumulative total of 967 authors, Alibaba's Qwen has 352, and Yuezhiyanmian has 387. DeepSeek's 391 authors, while far behind ByteDance, place it in the same league as Alibaba and Yuezhiyanmian, hardly indicative of a talent drain.

This data demonstrates that, despite the departure of a few top technical talents to major firms, DeepSeek's core R&D foundation remains robust. Liang Wenfeng's unique approach to talent management has successfully attracted and retained most top performers.

Critically, the large-scale external funding and IPO progress have provided employees with clear market valuations and realization pathways for their equity options, stabilizing team morale and resolving the earlier challenge of quantifying option value without external funding.

DeepSeek's Commercialization Transformation

After finalizing the CFO appointment and stabilizing its talent base, DeepSeek now faces the challenge of comprehensively upgrading its organizational capabilities.

In its early days, DeepSeek resembled an idealistic tech geek team. Relying on Liang Wenfeng's technical judgment and financial support from High-Flyer Quantitative Trading, the team could temporarily set aside short-term commercial gains to focus on exploring cutting-edge technologies.

This approach produced industry-shaking achievements like DeepSeek-R1 but also revealed clear shortcomings: the early team was small, the organizational structure flat, and commercialization progressed slowly.

Facing a valuation in the hundreds of billions and its impending status as a publicly listed company, DeepSeek is now strengthening its organizational capabilities across the board.

Business and team structures have been the first to expand. In June this year, DeepSeek announced plans to at least double the size of each department, recruiting across research areas such as supercomputing clusters, model alignment, code and mathematical reasoning, and multimodality. It also formed an Agent Harness team modeled after Claude Code, added software engineering roles in server-side, front-end, back-end, and testing, and built a reserve pool of AI cross-disciplinary technical talent.

Simultaneously, backend functions such as HR, legal, finance, and administration were established, propelling the overall team toward a thousand-member scale.

Commercialization has also advanced alongside technical R&D. Data shows that DeepSeek generated approximately 475 million yuan in revenue in the first seven months of this year, nearly tenfold growth compared to all of 2025; its comprehensive gross margin reached 44.6%, with the API business margin as high as 82.9%, while net losses narrowed to 715 million yuan, indicating gradually emerging commercial viability.

Global open-source community OpenRouter data reveals that DeepSeek's model token usage has consistently accounted for 5-20% over the past year, ranking among the top six globally with strong developer stickiness.

From capital operations and organizational expansion to commercial realization, DeepSeek is evolving from a world-class AI lab into a fully integrated tech enterprise.

With the CFO in place and the STAR Market IPO process advancing, Liang Wenfeng and DeepSeek are writing a new chapter in China's large model industry.

References:

"Liang Wenfeng Recruits a CFO," China Money Network;

"DeepSeek Welcomes a Post-90s CFO," Zhidx;

"Liang Wenfeng Retains 97% of Employees," China Money Network;

"Is DeepSeek Really Losing Its Core Talent? Answers Lie in 27 Papers," Caijing;

"DeepSeek Thanks Nearly 300 Researchers, Including 10 Former Key Figures," National Business Daily.

THE END