虎嗅

DeepSeek reportedly hired a post-90s partner from Hillhouse as its CFO

原文:DeepSeek被曝聘高瓴90后合伙人为CFO

From Hermit to Star: The Capital Changes and Survival Logic Behind DeepSeek’s IPO Pursuit

Hello everyone, I’m your financial journalist. Today, we’re talking about one of the most mysterious and unexpected players in the AI world—DeepSeek.

If you follow tech news, you might remember DeepSeek as a low-profile, hardcore company that refused to raise funds or become commercialized, like a “technological ascetic” living in the mountains. But recently, things have changed. Reuters reported that DeepSeek is hiring a seasoned investor from the 90s to serve as its CFO (Chief Financial Officer) and is preparing to go public on the Science and Technology Innovation Board (STAR Market).

This is more than just a change in position; it’s a signal that DeepSeek, which once shunned capital in favor of focusing on technology, is accelerating its integration into the mainstream business world.

To help you understand the behind-the-scenes reasons, I’ve broken down this news into five key points in plain language.

---

1. Why the sudden need for a CFO? It’s more than just managing finances

When people hear “CFO,” they often think of someone who oversees the company’s finances. However, for a company preparing for an IPO, the CFO’s role is much more crucial.

Think of the CFO as a “translator” to the capital market and a “chief designer.” Before, DeepSeek was a research lab that communicated in code and parameters. Now, it needs to explain its financial strength and potential profits to thousands of investors and institutions.

Key point: DeepSeek used to rely on funding from its founder, Liang Wenfeng’s company, Huanfang Quantitative (a quantitative hedge fund), which was private funding. Going public means following public financial rules, and hiring Yan Wentao, a partner at Hillhouse Capital with experience investing in companies like Tencent, ByteDance, and MiniMax, shows DeepSeek’s commitment to becoming more transparent and professional.

---

2. From “no funding” to “hiring like crazy”: A major shift in DeepSeek’s business model

One of DeepSeek’s most admirable qualities was its initial refusal to raise funds or commercialize. This was unique in the AI industry, where companies were spending heavily on expansion. But why the change? Because AI competition has become more competitive, and the game is now about money.

  • Previously: The focus was on clever algorithms, which could be developed by a few talented programmers.
  • Now: The competition revolves around computing power (GPU chips), data, and infrastructure—all of which are extremely expensive.
  • Financial highlights: In the first seven months of this year, DeepSeek invested about 11 billion yuan in AI infrastructure. If it wants to develop its own chips and build data centers, the cost could be in the tens of billions.
  • First round of financing: Approximately 50 billion yuan was raised in June, with a valuation of over 350 billion yuan.
  • Second round of financing: Launched in August, aiming to raise another 50 billion yuan, with a valuation of 500 billion yuan.
  • IPO: The main goal is to raise even more funds through the public market.

In simple terms: DeepSeek used to be a small, self-sufficient operation, but now it needs massive amounts of money to buy chips, rent servers, and hire a large team. Without funding or a public listing, it couldn’t compete with companies like OpenAI, Google, and Alibaba. So, by seeking capital, DeepSeek aims to strengthen its technical capabilities.

---

3. The talent war: Going public to retain top talent

What’s in high demand in the AI industry? Top talent. DeepSeek has accelerated its recruitment this year, even creating 150 new positions for core technologies. Why? Global tech giants are competing for these experts, and their salaries are rising.

  • Before going public: DeepSeek retained talent with passion, technical challenges, and financial support from Huanfang Quantitative. But this won’t be sustainable in the long run.
  • After going public: DeepSeek can offer stock options to its employees.
  • Example: Employees might think, “Working at DeepSeek is great, but what if the company runs out of money? I don’t own any shares, so my earnings are limited.”
  • After going public: Employees can own shares and benefit from the company’s success, making them both employees and shareholders.

Key point: Hiring a CFO and going public helps establish a standardized incentive system to retain top talent.

---

4. The dilemma of commercialization: Making money, but not in a way that attracts investors

Many think AI companies are losing money, but DeepSeek is already generating revenue—although it’s a bit complicated.

  • Revenue: According to the Financial Times and The Information, DeepSeek’s annual recurring revenue (ARR) has reached 500 million US dollars (about 3.5 billion yuan), with revenue in the first seven months being 10 times that of last year.
  • Losses: Despite increasing revenue, the company still had a net loss of 715 million yuan in the first seven months.
  • Reason: AI is a costly industry. Every additional user and model requires more electricity, servers, and chips.
  • Strategies: DeepSeek has adjusted its pricing: it raised API fees in August to improve profitability and then lowered the prices of its flash models shortly after.

In simple terms: DeepSeek is balancing its pricing to attract developers while ensuring profitability for investors. It needs to prove to the market that it can be both technologically advanced and financially viable.

---

5. The battle for control: How does Liang Wenfeng retain control while raising funds?

Liang Wenfeng, DeepSeek’s founder, is a tech enthusiast with a low-profile. During the private funding phase, he used a clever structure:

  • Limited partnerships: Investors from Tencent, JD.com, and CATL invested through limited partnerships led by Liang Wenfeng as a general partner.
  • Impact: Investors had only the right to dividends, no voting rights, and their shares were locked up for five years.
  • **This meant Liang Wenfeng maintained control, keeping investors as “financial backers” without interfering with company decisions.
  • Challenges with going public: A public company requires a board of directors,信息披露, and regulatory compliance. This structure no longer works.
  • Future plans: Liang Wenfeng may adopt a “different share structure” (similar to Xiaomi or JD.com) to ensure he still has significant voting rights, even if his shares are diluted. He also needs to attract diverse investors (such as state funds and industry capital) while protecting the company’s technical independence.

In simple terms: Liang Wenfeng faces a dilemma: he needs to raise funds to sustain the company but doesn’t want to lose control. He must find a legal and commercial balance that satisfies both investors and maintains his control over the company’s direction.

---

Conclusion: What does DeepSeek’s next step mean?

DeepSeek’s transformation reflects the broader trends in China’s AI industry:

1. From idealism to realism: Early AI startups relied on passion and talent; now, competition relies on capital and computing power. DeepSeek’s transition marks a new phase driven by capital.

2. The maturity of China’s AI industry: With companies like DeepSeek going public, investors can see the real financial performance of AI companies, not just hearsay.

3. Global competition: DeepSeek’s IPO is about gaining a foothold in the global AI landscape, securing the funds needed to compete with American tech giants and develop its own technology.

To conclude: DeepSeek’s path to an IPO will undoubtedly be challenging, with issues like profit models, control, and compliance to overcome. Nevertheless, DeepSeek is no longer a hidden lab; it’s stepping into the spotlight, facing global scrutiny and evaluation.

For consumers, this means AI services may become more stable and affordable due to economies of scale. For investors, it’s an opportunity to assess the true value of China’s AI industry.

DeepSeek’s story is just beginning.