The behind-the-scenes story of DeepSeek's first major capital raise reveals a frenzy among investors, with brokers peddling dubious allocations and founders personally vetting backers.
DeepSeek, the artificial intelligence foundation model company, has concluded talks for its first funding round, with agreements signed last month, according to a person involved in the transaction. The reported financing exceeds 50 billion yuan, or approximately $7.4 billion. Founder Liang Wenfeng contributed 20 billion yuan personally—the largest single investment in this round. Tencent invested 10 billion yuan, Contemporary Amperex Technology Co., Limited (CATL) invested 5 billion yuan, while NetEase, JD.com, and IDG Capital each invested 3 billion yuan. The National Artificial Intelligence Industry Investment Fund contributed 1 billion yuan.
The person added that DeepSeek's pre-money valuation was approximately 350 billion yuan following the round. However, after adding a 5% ESOP (employee stock ownership plan) issuance, the actual pre-money valuation is around 367.5 billion yuan, or about $54.3 billion. The final list of investors shows significant changes from earlier market rumors, with several major corporate strategic investors and market-oriented investment institutions joining. Notably, the state-owned capital allocation ended up being lower than many expected.
Since April, soaring valuations for DeepSeek have driven sentiment in the primary market. Its valuation jumped from $20 billion in April to approximately $54 billion currently. Upon completion, this round will make DeepSeek the company with the largest first-round financing in the history of China's AI large model sector.
This marks the first time the three-year-old company has opened to external funding. Previously, founder Liang Wenfeng stated DeepSeek would "not raise funds, not go public, and not commercialize." Market interest in who would ultimately get a piece was unprecedented, leading investors to try every method to secure an allocation. One investor reportedly paid a 5 million yuan "meeting fee," traveled to Hangzhou three times, and waited outside DeepSeek's offices but still failed to meet Liang personally. Another investor from a state-backed fund met with eight self-proclaimed financial advisors claiming to have allocations, with almost no success.
Even though DeepSeek's first round is complete, allocations are still being peddled today. Capital enthusiasm in China's AI sector remains high, with valuations for listed companies like Zhipu AI and MiniMax continuing to rise post-Hong Kong IPO, and companies like Moonshot AI and StepFun securing multi-billion-dollar financings. DeepSeek's funding process has become a concentrated display of the capital market's "fear of missing out."
The World of Brokers and Questionable Allocations
Numerous investment institutions have been searching for channels, leading to the emergence of various brokers and intermediaries of unknown origin. Investor Li Jing, also tasked with finding an allocation, traveled to Hangzhou three times after the financing news broke, hoping to meet Liang Wenfeng, but returned empty-handed each time. In late May, Li encountered an introducer claiming to be close to Liang, who demanded a 5 million yuan "meeting fee" to arrange an introduction. Li arrived in Hangzhou two days early, staking out near DeepSeek's headquarters. Upon hearing the financing situation had changed, he canceled the meeting and retrieved his 5 million yuan.
"Being able to pay the money is a capability in itself," Li said. For most LPs, the most challenging issue isn't how much to invest, but finding the right person to give the money to. With opaque channels, people are already profiting at every intermediate step.
DeepSeek is undoubtedly one of the hottest AI targets in China. Despite the investor list being finalized, social media is still flooded with offers to sell DeepSeek allocations.
The global AI investment boom continues, with valuations for hot targets rising. After U.S. model startup Anthropic completed its latest round at a $65 billion valuation, it reached $96.5 billion, surpassing OpenAI as the world's highest-valued AI unicorn. For such sought-after targets, investing through an SPV (special purpose vehicle) was a common industry practice, where GPs package assets into tradable securities to bring in new LPs, who often lack the reputation for direct investment.
Pass-through investment via special funds incurs additional management fees, a method less favored by top-tier funds. However, many investors stated they were willing to pay extra channel fees to invest,普遍抱着 the belief that DeepSeek's valuation will continue to rise and profits can still be made upon a future listing.
One channel contact, when queried by a reporter posing as an LP, claimed the current channel secured a 500 million yuan allocation from the "National Integrated Circuit Fund," limited to individual participation. Investors would enter as direct LPs in the state-owned shareholder structure, required to submit sufficient proof of assets within two days, with subscription potentially completed within the week.
He also mentioned that the front-end management fees for the FA and GP in this transaction were 8% and 9% respectively, totaling a hefty 17%. For a 100 million yuan investment, 17 million yuan would be deducted upfront as fees—a rarity in standard private equity deals where GPs typically charge around 2% annually based on fund size. Intermediaries peddling DeepSeek allocations often propose collecting management fees for three to five years in one lump sum.
These FAs also create an atmosphere of extreme scarcity. Another FA claiming to have secured a 2 billion yuan DeepSeek allocation from a "local state-owned entity" said the channel had several options, with "a 2 billion yuan fund basically snapped up in two or three days." However, when asked for the GP's specific name, the person became evasive, only stating there was a "long-term cooperative relationship." Half a month later, upon follow-up, the same intermediary indicated that the same channel still had available额度.
Despite the dubious origins of these channels, some are willing to take the risk. Investor Chen Duo, serving at a state-backed fund managing tens of billions, negotiated with eight different FAs. Most claimed allocations came from "a local state-owned entity," but in the vast majority of cases, after Chen sent a POF (proof of funds), communication ceased. Chen categorized any situation where he couldn't witness the underlying investment agreement or confirmation letter on-site as "being strung along."
Through these interactions, Chen gradually understood the fundraising logic: many institutions first raise a fund to demonstrate they have money on their books, then use that to seek an allocation from DeepSeek. FAs then craft a seemingly reliable channel using rhetoric like connections to senior leaders, urging investors to quickly provide proof of funds. The goal is to make Liang Wenfeng believe they already have committed capital to secure an allocation. This creates a chicken-and-egg cycle, as even with sufficient raised funds, obtaining a DeepSeek allocation is not guaranteed.
An investment manager at a local state-owned fund expressed frustration that FAs still approach him to introduce LPs. "DeepSeek simply doesn't lack LPs and doesn't need FAs," he said, believing that most people wouldn't trust these unclear channels unless they were very small institutions or individual LPs who might be persuaded.
An institutional source familiar with DeepSeek revealed that all capital entering this round came from main funds, with no allocations through any special purpose vehicles. According to media reports, Liang Wenfeng and his team required verification of all LP identities participating in this round to prevent allocations from falling into the hands of unknown investors. The source stated that FAs still trying to raise funds ultimately won't succeed, questioning their motives.
The Selection Process for Investors
Tang Yu, a partner at a private equity fund, has been following DeepSeek since its viral fame in February 2025. After news of the open funding round emerged, Tang, privately commissioned by a state-backed fund with a 3 billion yuan investment intent, searched everywhere for a channel. Through personal connections, he contacted a senior executive at High-Flyer Quant, a quantitative hedge fund controlled by Liang Wenfeng that has long funded DeepSeek's R&D costs.
Tang and the executive had known each other for years. He originally thought the 3 billion yuan commitment could open the door, but the executive merely provided a company email address, instructing him to send institutional introductions, letters of intent, and cooperation proposals via email, stating "a responsible person will review them uniformly." He never received a positive response thereafter.
Tang noted that anyone with intent and direct access to this executive likely received a similar reply. On May 28th, he inquired again with the High-Flyer executive about financing developments. The response was that participating in this round was "definitely too late," and the timing of the next round remained unknown.
Multiple investors involved in the transaction stated that investors in this round were primarily personally selected by Liang Wenfeng, who倾向于选择 institutions that wished for his success or had good relations with him. Contact was mostly initiated by Liang's invitation; those not invited found it difficult to establish truly effective connections, even after multiple attempts to meet.
Another investor, Hu Jie, also noted that funds unfamiliar with Liang or lacking reputation基本上 couldn't get in. Hu works at a market-oriented investment institution where AI is a key focus area. However, at the start of DeepSeek's open funding, his interest was limited because initial understanding suggested the round might be state-capital-led, and the deal size was already large, making the potential investment回报 not necessarily high. Several institutional investors shared this view, initially believing it was a state-capital-dominated deal.
However, after this year's May Day holiday, Hu observed several investment institutions, including Monolith Capital and Hillhouse Capital, engaging intensively with DeepSeek. His institution subsequently joined the negotiations.
To the surprise of several participants, the final state-owned capital allocation was lower than预期. However, according to media reports, only the National Artificial Intelligence Industry Investment Fund invested directly in DeepSeek in this round,不受 lock-up restrictions and享有 company voting rights. Other institutions invested in a limited partnership managed by Liang Wenfeng, rather than directly in DeepSeek itself, with shares locked up for five years during which they cannot be sold.
Public information shows the National Artificial Intelligence Industry Investment Fund Partnership was established in early 2025 by Guozhitou Private Fund Management Co., Ltd. and the National Integrated Circuit Industry Investment Fund Phase III Co., Ltd. Guozhitou, founded in 2024, has concentrated its recent investments in AI, chips, and robotics, including companies like Moonshot AI, MetaX, and CloudWalk Technology.
In this funding round, several state-backed funds actively sought allocations. Chen Duo, serving a state-owned fund, approached eight FAs before getting接近 the deal's periphery. Another investment manager at a local state-owned fund actively接触 AI companies told reporters he started paying attention to DeepSeek during the 2025 Spring Festival. Other promising targets he看好 included star model companies like Zhipu AI and Moonshot AI, but he failed to invest in any,甚至 lacking access to them.
"Objectively speaking, state capital has its limitations," he said. "Large companies要求快速出资, and单笔规模 are very large. For state capital, this意味着 risk."
The investor list also includes several tech giants. Among them, Tencent has the largest allocation, followed by Contemporary Amperex Technology Co., Limited (CATL). Media previously reported that Alibaba was also in negotiations with DeepSeek, though Alibaba has publicly denied this, expressing limited interest.
Multiple AI investors analyzed that between Tencent and Alibaba, Tencent has stronger motivation to participate, as its foundational model capabilities have long lagged behind the first tier, creating a stronger desire for strategic cooperation with DeepSeek. Alibaba has built its own ecosystem from models to chips, making the strategic need to invest in DeepSeek less compelling. A transaction participant believed the cooperation foundation between DeepSeek and Alibaba was weaker than with Tencent.
The entry of Contemporary Amperex Technology Co., Limited (CATL) as a new energy giant indicates, to some extent, that AI model vendors are strategically指向 the energy variable behind computing infrastructure. A transaction source revealed that DeepSeek is already building its own data center, planning to operate computing power directly. Public information shows that in late April, DeepSeek posted several computing power-related job openings, including senior data center operations engineers and senior delivery managers, requiring candidates to manage the full process from project initiation, construction, to operation, with the location in Ulanqab, Inner Mongolia.
DeepSeek's Valuation Rationale
According to several investors in the AI field, DeepSeek's decision to open funding stems from at least two reasons. First, DeepSeek's model iteration requires continuous investment in large-scale training compute. Sufficient capital is a crucial source to sustain this long-cycle投入, as the自有资产 previously generated by High-Flyer Quant may no longer meet future R&D needs.
Chen Shi, a partner at Ruifeng Capital, analyzed that top large model companies like Anthropic operate the industry's "premier business model" but also face the "greatest pressure model" because the Scaling Law remains in effect. Each generation's model training cost exceeds the previous, requiring continuous investment in compute and data. Falling behind means being forgotten by the market. As an AI startup, today's primary market can hardly满足 the funds needed for model iteration; the next step inevitably leads to the secondary market.
The second reason is to provide market validation for employee stock options to retain talent. This round included a 5% employee option pool issuance, worth approximately 17.5 billion yuan. However, previous analysis indicated that, as of the V4 model release, DeepSeek's core research team had not experienced significant talent流失.
Hu Jie stated that DeepSeek employees also希望 the company to go public. An AI practitioner familiar with DeepSeek's R&D staff noted it's unsurprising employees期待 a listing. "People aren't saints; those坚守 DeepSeek can't rely on ideals alone."
Since 2025, several prominent Chinese AI large model companies have密集融资 and pursued listings, accelerating capitalization. Besides Zhipu AI and MiniMax, which listed on the Hong Kong stock exchange earlier this year, StepFun reportedly raised about $2.5 billion in its latest round, with an expected cornerstone valuation of around $10 billion. Corporate information platform Qichacha shows Moonshot AI has累计融资 about $4.4 billion, with a latest pre-money valuation of $20 billion. Moonshot AI正在拆除 its VIE and红筹 structure in preparation for listing, while StepFun aims to file as soon as the first half of this year.
In comparison, DeepSeek started fundraising the latest but achieved approximately $7.4 billion in its first round, with a valuation of around $54.3 billion, far surpassing its competitors.
Both internal and external factors刺激 DeepSeek's valuation不断攀升 above other model companies. Several transaction participants consider the current range reasonable because, in terms of foundational model capability, DeepSeek's推理效率 is公认 at the industry forefront.
Since its breakout R1 model last year, DeepSeek has maintained low cost and high performance. Taking the V4-Pro released in April as an example,官方披露 that in a 1 million Token long-context scenario,单Token推理 FLOPs were only 27% of the previous generation V3.2. This efficiency advantage directly体现在 API pricing, with V4-Pro at 0.025 yuan per million Tokens, among the lowest globally. Extreme affordability has made DeepSeek widely popular among global developers. On the AI model aggregation platform OpenRouter's latest榜单, DeepSeek V4 Flash ranked first in Token消耗量 this month.
A dollar fund practitioner familiar with the AI领域 stated that on the OpenRouter platform, Chinese model调用量占比 has jumped from 1% three years ago to 60%. Chinese model companies have successfully extended the advantage of极致降本 in the supply chain to the AI field, gradually turning large models into a standardized commodity, with DeepSeek achieving this most极致.
Furthermore, several investors and industry experts分析 that another reason for DeepSeek's high valuation is its上升为 a national-level AI strategic platform.
The external environment also plays a role. The二级市场 performance of Zhipu AI and MiniMax has一定程度上刺激了投资人 pricing for later entrants. Taking Zhipu AI as an example, its IPO market cap was 57.9 billion HKD, while its latest市值 has risen to 740.1 billion HKD. A state-backed fund person familiar with AI venture capital分析 that after this year's Spring Festival,对标 model companies including Zhipu AI and MiniMax experienced rapid surges, bringing极强的资本市场赚钱效应.
This fund person also believes the current moment is a关键节点 for discussing large model commercialization because, after three to four years of development, technology can now支撑成熟产品落地, and the market迫切验证 AI's商业价值. Against this backdrop, more primary market capital is涌入大模型 companies. In the past, these model startups were普遍被视为 "lacking certainty" by state-backed funds.
In contrast to the overall market's "fear of missing out (FOMO)" sentiment, some investors who never had接触意愿 from the start consider DeepSeek's current valuation too high, "unfathomable." Participating in negotiations consumes大量时间和金钱成本, with a very low likelihood of最终投进去. Considering投资回报, the deal doesn't seem划算. Hu Jie views DeepSeek as a "white horse" stock—referring to those with high market capitalization and relatively stable performance.
Facing the capital market's考验, DeepSeek cannot avoid the problem all tech startups face: how to balance技术理想与商业回报 long-term. How to commercialize is an important拷问悬在 DeepSeek's high valuation之上. According to previous reports, in March this year, a DeepSeek高管透露 at a meeting that外界关注 DeepSeek's商业模式落地和技术进程, and DeepSeek一直在做很多努力和尝试, also初步验证找到了一些路径.
Compared to other model companies and internet giants, DeepSeek's产品端进展 have been slow, with previous重心一直放在模型训练上. Regarding whether DeepSeek should develop products, Hu Jie stated that Liang Wenfeng大概 realized the重要性 of productization around late last year, but the team一直没招到合适的负责人 and struggled with the product路线. "The product matter requires Liang Wenfeng's own思考." Recently,公开招聘网站显示 DeepSeek正在开放不少 product-related岗位, including model strategy product managers and Agent data strategy engineers.
However, this doesn't必然意味着 DeepSeek has undergone a路线变化. According to a May 22nd media report, Liang Wen锋表态 at a meeting面向投资人 that the company will继续推进开源 AI models, aiming to achieve Artificial General Intelligence (AGI). In his表述,技术突破 is DeepSeek's核心使命 to突破技术边界, rather than追求盈利.
Hu Jie stated that融资只是资本操作, and DeepSeek's日常运作和技术路线 are也不必然被资本市场绑定.
An investment person on the transaction list stated that for DeepSeek, he believes短期商业化 is not that important; achieving AGI is paramount, and China needs its own AGI.
Comments