In a landmark first funding round exceeding 50 billion yuan, DeepSeek concurrently allocated a 5% employee stock ownership plan (ESOP). Valued at approximately 175 billion yuan based on a pre-money valuation of 3500 billion yuan, this move is strategically aimed at retaining core talent.
The artificial intelligence foundation model company DeepSeek has concluded negotiations for its inaugural funding round, with agreements finalized in late May. Reports indicate this round raised over 50 billion yuan, equivalent to about $7.4 billion. Key investors include Liang Wenfeng, who personally contributed 20 billion yuan, Tencent with 10 billion yuan, CATL with 5 billion yuan, NetEase, JD.com, and IDG Capital each investing 3 billion yuan, and the National Artificial Intelligence Industry Investment Fund contributing 1 billion yuan.
Following the round's closure, DeepSeek's pre-money valuation is approximately 3500 billion yuan. However, with the issuance of the 5% ESOP, the adjusted pre-money valuation reaches about 3675 billion yuan, or roughly $54.3 billion.
Frenzied Investor Interest and Market Speculation
The finalized investor list shows significant changes from earlier market speculation, with several leading strategic investors and market-oriented institutions joining. Notably, the share allocated to state-owned capital was surprisingly the smallest, contrary to initial expectations.
Since April, DeepSeek's soaring valuation has captivated the primary market, climbing from $20 billion to approximately $54 billion. This funding round establishes DeepSeek as the Chinese AI model company with the largest-ever first-round financing.
This marks the company's first external fundraising in its three-year history, a shift from founder Liang Wenfeng's previous stance of "no fundraising, no IPO, no commercialization." The intense market interest in securing an allocation led to extraordinary efforts by investors, including reports of one paying a 5 million yuan "meeting fee" and making three trips to Hangzhou without success, and a state fund investor meeting eight self-proclaimed intermediaries with little result.
Even after the round's completion, offers for DeepSeek shares continue to circulate online, reflecting the intense "fear of missing out" (FOMO) sentiment in China's AI capital markets, where other leading startups have also secured massive funding or seen their market values surge post-listing.
Intermediaries and Questionable Offers
Numerous investment firms scrambled to find channels, leading to a proliferation of brokers and middlemen of dubious origin. One investor, after being quoted a 5 million yuan fee for a meeting, ultimately retrieved his money after hearing the deal terms had changed. The opaque process created opportunities for intermediaries to profit at various stages.
Despite the investor list being set, social media remains flooded with offers to sell DeepSeek shares. The global AI investment frenzy has driven valuations of hot targets sky-high, with practices like using Special Purpose Vehicles (SPVs) becoming common for investors lacking direct access.
Some intermediaries claimed to have access through channels like the "National Integrated Circuit Fund," offering allocations to individual investors with exorbitant upfront management fees as high as 17%, a rate considered highly unusual in standard private equity transactions. These fees often represented several years' worth of management charges collected in advance.
These intermediaries also created an atmosphere of scarcity. One FA claiming to have a 2 billion yuan allocation from a "local state-owned entity" said their fund was typically subscribed within days, yet weeks later, the same channel reportedly still had available quota.
Some investors, like one from a state fund, engaged with multiple FAs, only to find communications ceased after submitting proof of funds. The process revealed a circular logic where some institutions would raise a fund first to demonstrate available capital and then seek an allocation from DeepSeek, creating a chicken-and-egg dilemma.
An investment manager from a local state fund expressed frustration at still being approached by FAs, stating DeepSeek neither lacks limited partners nor needs financial advisors. He believed only very small institutions or individual LPs might be persuaded by such unverified channels.
A source close to DeepSeek clarified that all funding in this round came from primary funds, with no SPV participation. It was reported that Liang Wenfeng and his team required verification of all participating LPs' identities to prevent shares from going to unknown investors. This source suggested that FAs still fundraising would ultimately fail.
The Selection of Investors
Access to the round was tightly controlled. One private equity partner, seeking a channel for a 3 billion yuan investment on behalf of a state-backed fund, contacted an executive at Liang Wenfeng's quantitative hedge fund, which had long funded DeepSeek's R&D. Despite their existing relationship, the response was merely a company email address for submitting materials, with no further engagement.
Multiple investors involved stated that Liang Wenfeng personally selected participants, favoring those who wished for his success or had good relations with him. Contact was typically initiated by Liang himself; unsolicited approaches, even with repeated attempts, rarely led to cooperation. As one investor noted, funds lacking name recognition or a connection to Liang had little chance of participating.
Some market-oriented investors were initially less interested, assuming state capital would dominate the round and that the large scale might limit returns. This perception shifted after the May holiday when they observed other prominent funds actively engaging with DeepSeek.
Contrary to expectations, the final state-owned capital share was lower than anticipated. Reports indicate only the National Artificial Intelligence Industry Investment Fund invested directly in DeepSeek, enjoying no lock-up period and voting rights. Other institutional investors funded a limited partnership managed by Liang Wenfeng, not the company directly, with a five-year lock-up on their shares.
The National AI fund, established in early 2025, has recently invested in other AI, chip, and robotics companies. Several state-backed funds actively sought DeepSeek allocations, with one investor contacting eight FAs just to get near the deal. Another state fund manager noted the difficulty for state capital to meet the requirements of large, fast-moving investments due to associated risks.
Among corporate investors, Tencent secured the largest share, followed by CATL. While there were earlier reports of talks with Alibaba, the company has publicly denied significant interest. Analysts suggest Tencent, whose foundational model capabilities have lagged behind the top tier, had stronger motivation for a strategic partnership with DeepSeek than Alibaba, which has built its own ecosystem from models to chips.
CATL's participation highlights the strategic importance of energy infrastructure for AI model companies requiring massive computing power. Reports indicate DeepSeek is building its own data centers, with recent job postings for roles related to data center operations in Ulanqab.
DeepSeek's Valuation Rationale
According to AI investors, DeepSeek opened to funding for at least two key reasons. First, continuous large-scale investment in computing power for model training is essential, and internal funding from its associated hedge fund may no longer suffice for future R&D needs. As one partner noted, leading model companies operate under a high-pressure "top-tier" business model, requiring constant investment in compute and data to avoid falling behind, a need that eventually outgrows the private markets.
Second, the move validates employee stock options in the market to retain talent. The newly issued 5% ESOP, worth about 17.5 billion yuan, serves this purpose, although the core research team has reportedly remained stable thus far. Employees, like those at any startup, naturally anticipate a potential future liquidity event.
DeepSeek entered the funding scene later than its domestic peers but secured a first round of about $7.4 billion at a $54.3 billion valuation, far surpassing its rivals. This high valuation is attributed to both internal strengths and external factors.
Internally, DeepSeek is recognized for leading inference efficiency. Its models, starting with the breakthrough R1, have maintained high performance at low cost. For instance, the V4-Pro model demonstrated significantly lower computational cost per token than its predecessor, translating to an API price among the lowest globally. This cost advantage has made it popular with developers, leading it to top usage charts on platforms like OpenRouter, where Chinese models now account for a significant portion of traffic.
Externally, the strong secondary market performance of comparable listed Chinese AI companies has influenced pricing for later-stage players. The rapid rise in valuations post-Chinese New Year created a powerful wealth effect, attracting more private capital into a sector once viewed by some state funds as lacking certainty.
Some investors, however, consider the current valuation too high and the process of securing a stake too costly and uncertain to justify the potential return. They view DeepSeek as a high-quality but expensive "blue-chip" startup.
The Commercialization Question
A critical challenge for DeepSeek's high valuation is the question of commercialization, a test all tech startups face in balancing technical ideals with financial returns. Compared to other model companies and internet giants, DeepSeek's progress on product development has been slower, with its focus historically on model training.
Insiders suggest Liang Wenfeng recognized the importance of productization around late last year but struggled to find suitable leadership, leading to internal debate on the product direction. Recent job postings for product-related roles indicate a shift, though this does not necessarily signal a complete change in strategy. Liang has reportedly reiterated the company's commitment to open-source models and the pursuit of Artificial General Intelligence (AGI), with technological breakthrough as the core mission over profit-seeking.
As one investor on the final list noted, for DeepSeek, achieving AGI—and China having its own AGI—is the paramount short-term goal, with commercialization being of secondary importance for now. The recent fundraising is seen as a capital operation that does not necessarily bind the company's daily operations or technical roadmap to market pressures.
Comments