DeepSeek's CFO has finally arrived.

DeepSeek, which has been advancing at full speed, has finally filled a critical missing piece on the capital side.

On September 14, multiple independent sources confirmed that Yan Wentao, Partner of GL Ventures, is about to join DeepSeek as Chief Financial Officer. He has already started the resignation process inside GL Ventures and is expected to officially take office in the near future.

The news did not cause a huge stir, but instead made many insiders feel that "it finally happened". The CFO position of this company has been vacant for far too long. Public recruitment information shows that DeepSeek opened the CFO recruitment as early as the beginning of 2025. Over the following more than a year, the recruitment notices for the finance team were repeatedly updated, but the core responsible person was never finalized. This vacancy period is exactly the stage when DeepSeek expanded from a technical team to a full-stack AI company at the fastest pace.

The person in charge of finances selected by DeepSeek naturally has an impressive resume.

Yan Wentao is a well-known post-90s investor in China's venture capital circle. Born in 1991, he graduated from Fudan University. Since 2013, he has worked at Tencent Investment and H Capital successively. After joining GL Ventures in 2020, he gradually grew into a representative of the new generation of partners, and was selected into the Top 50 Nova Rising Star Investors list by Zero2IPO in 2025.

During his tenure at GL Ventures, Yan Wentao has long been deeply engaged in the cutting-edge technology track. He began to track the large model technology route as early as 2020, and laid out the embodied intelligence direction in advance from 2023, leading or deeply participating in investments in dozens of leading enterprises such as ByteDance, Zhipu AI, MiniMax, Xiaohongshu, J&T Express, Webull Securities and others.

Different from CFOs with traditional financial backgrounds, Yan Wentao's core competencies focus on primary market financing, valuation of science and innovation enterprises, IPO equity structure design and industrial resource docking. This kind of investor background is exactly the trait that DeepSeek values most at the current stage.

The node when the candidate is finalized just falls at the critical intersection of capitalization.

Just five days before the CFO candidate was finalized, media just reported that DeepSeek had selected CITIC Securities as the sponsor and started the preparatory work for IPO on the Sci-Tech Innovation Board. The two pieces of news echo each other, clearly revealing the capitalization rhythm of DeepSeek.

In June this year, DeepSeek completed the first round of external financing since its establishment, with total fundraising exceeding 500 billion RMB and a post-investment valuation close to 4 trillion RMB, setting a record for the scale of a single round of financing in China's AI industry. This round of financing was led by founder Liang Wenfeng with a personal contribution of about 200 billion RMB, jointly participated by many industrial and financial capitals such as Tencent, CATL, NetEase, JD.com, IDG Capital, and followed strategically by the National Artificial Intelligence Industry Investment Fund.

What is quite special is that except for the National Artificial Intelligence Industry Investment Fund, which directly holds shares and enjoys voting rights, all other external investors hold shares indirectly through limited partnerships controlled by the founder, accepting a five-year lock-up period and no voting rights. This extremely strong founder's right to speak is extremely rare in the global primary market.

While the capital side is pressing ahead with intense efforts, the business side has not slowed down.

In April this year, DeepSeek released the V4 series of large models. With its self-developed mixed attention architecture, it reduced the computing power per token of the Pro version to 27% of that of the V3.2 version. Then it announced that the V4-Pro API would be permanently reduced to 25% of the original price, greatly lowering the threshold for using large models.

On September 10, DeepSeek officially launched the V4.1-Flash model, adding native multi-modal visual understanding capabilities, and refreshed performance in more than ten international benchmark tests such as code programming, mathematical reasoning, and cybersecurity. At the same time, DeepSeek is also laying out computing power and terminal ecosystems, promoting the construction of a 1GW-level intelligent computing center in Ulanqab, and cutting into the embodied intelligence track by taking shares in Unitree Robotics, extending from basic models to downstream applications and hardware.

In retrospect, it is almost the industry's default consensus that DeepSeek needs a full-time CFO.

Over the past year and more, DeepSeek has been advancing rapidly in technology, products and financing, but there has always been a lack of full-time responsible persons for the top-level design of finance and capital. As the financing scale jumps to the 100-billion level and the IPO is put on the agenda, the value of the CFO is far more than just managing accounts, but undertaking the core functions of capital operation, shareholder communication, and listing compliance.

Yan Wentao's arrival fills a key short board on DeepSeek's path to large-scale development. For DeepSeek, this CFO really should have come long ago.