The DeepSeek Financing Story

链捕手Published on 2026-06-18Last updated on 2026-06-18

Abstract

DeepSeek's financing round, totaling approximately 3 billion USD, concluded recently, revealing details about the process and key investors. The round was initiated around April with strict initial terms: a minimum commitment of 5 billion RMB, no syndication, and a pure RMB structure. These were later relaxed, with the minimum ticket size reduced to 1.5 billion RMB. A pivotal four-hour online investor meeting in mid-May served as the primary interaction for many backers with DeepSeek's founder, Liang Wenfeng. Despite not being a naturally eloquent speaker, Liang's philosophy deeply resonated. He consistently emphasized the company's singular focus on AGI (Artificial General Intelligence), the principle of "less is more," extreme caution in spending, and the paramount importance of team stability. His notable quotes included describing the team as "ordinary people doing extraordinary things" and stating that "AGI is a big enough thing; everything else is just process." The final investor list featured 10 entities, but underlying fund structures indicate participation from nearly 100 institutions and individuals. Notable lead investors include Monolith Capital (increasing its commitment from 1.5 to 3 billion RMB), Zhenxingu Capital, IDG Capital, and state-affiliated investors like Guozhitou. Conspicuously absent were major firms like Sequoia China and Hillhouse Capital, despite earlier speculation about their involvement. A core condition set by Liang Wenfeng for all investo...

Author:@Liu Jing, elsewhere

Information about DeepSeek's financing is already everywhere.

Known information will not be repeated by "elsewhere." Below are some untold stories or details we have learned about.

The Legendary Four-Hour Meeting

First, let's talk about that investor meeting, the one whispered about as the "four-hour meeting."

In mid-May, DeepSeek organized this online meeting using the Tencent Meeting system. At that time, the investors were largely finalized, with each institution having two spots to attend.

Liang Wenfeng spoke first, followed by a Q&A session where he addressed investors' questions.

Legend has it that one CVC investor spent ten minutes just asking questions: first a lengthy self-introduction, then three lengthy questions. The atmosphere became awkward for a moment—a scene easy to imagine. Some people whispered among themselves, asking who this person was.

But Liang was unfazed and answered each question seriously. Although after answering the second question, he momentarily forgot the third one and had the investor repeat it.

Prior to this, most of these investors who eventually got in had never met Liang Wenfeng offline, so this was essentially their first real interaction with him. After the meeting, many were left with a lingering impression, a feeling that the essence of the discussion remained with them.

Liang Wenfeng's memorable quotes and spirit began circulating in the market.

Restraint and AGI

Liang Wenfeng is not an exceptionally articulate speaker, and his pace of speech isn't fast, but several quotes from that meeting are often mentioned (paraphrased, the exact wording may vary slightly)—

"We have no fame or influence; we are a group of very ordinary people."

"The narrative I like is that a group of ordinary people accomplished something extraordinary. For DeepSeek, our core is to ensure team stability—this is more important than money or resources. This is the biggest risk and also the greatest challenge."

"Assuming AGI will account for 20% of GDP, those who want 5% of it will lose to those who want 1%; those who want 1% will lose to those who want 0.1%."

"We will focus on things related to enhancing intelligence; we won't do anything else."

"AGI is a significant enough goal; everything else is just a process."

Throughout the four hours, the themes Liang Wenfeng repeatedly emphasized were consistent: focusing solely on AGI; 'less is more'; restraint, and so on.

From 5 Billion to 1.5 Billion

Formal fundraising began around April this year. There was reportedly a so-called "fund white list" consisting of institutions with strong investment capabilities and good reputations.

Initially, DeepSeek required a minimum ticket size of 5 billion RMB, no syndication of shares, and a pure RMB structure.

Against this backdrop, a prominent investor who was initially rumored to definitely participate but ultimately didn't, is widely believed in the industry to have been excluded because they had reportedly marketed shares to Middle Eastern LPs.

However, in China, there are very few market-oriented funds capable of a single investment of 3 billion RMB. Therefore, some VCs repeatedly negotiated whether it could be split among multiple fund entities.

So, looking at the result, the minimum ticket size for a fund was lowered to 1.5 billion RMB, and the fund-raising method became more flexible.

There was also a rumor that participating institutions should preferably not have invested in other large language models. This was probably unrealistic for a simple reason—finding a top-tier fund that hasn't invested in large models is contradictory.

During the investor meeting, Liang Wenfeng also expressed that DeepSeek is willing to help other companies—even including other LLM companies. Of course, the premise remains that they don't poach his people.

Monolith

Monolith Capital appeared early on the investor list. The change was that initially it was 1.5 billion, but it ended up being 3 billion.

The change in numbers is said to be related to the adjustment in Guozhitou's allocation, freeing up some shares—the latter's figure of 980 million RMB suggests it was kept below the 1 billion integer.

Regardless, being willing to place big bets is indeed Cao Xi's distinctive style. He has proven it once with Kimi.

Monolith is probably the youngest VC among the participating institutions in this round.

Zhengxingu

Zhengxingu might be the most surprising presence in the final list.

Many interpret it as Zhengxingu having not invested in any Chinese LLM company during this wave, thus being more eager to get on board. However, according to our knowledge, they did invest in Kimi, likely between $6 billion and $10 billion valuation. But compared to their fund size, the investment amount is relatively limited.

One market interpretation is that the founder of Zhengxingu comes from a secondary market background and seems to have more connections with quantitative funds like Huafang. But in reality, in the secondary market, quantitative investing and fundamental investing are two distinct circles, even with different values.

According to our understanding, this investment happened not only because of Lin Lijun and his partnerYe Chunyan, but also due to the efforts of a newly joined investor.

Zhengxingu was also one of the earliest VCs to start communicating with DeepSeek.

Will DeepSeek Be Sequoia and Hillhouse's Biggest Miss?

Now to the most surprising part of this deal.

In the initial rumors, Hillhouse would participate and invest 5 billion. Zhang Lei also reportedly mentioned to friends multiple times that since 2025, he and Liang Wenfeng had occasional exchanges, and when in Beijing, he would visit Liang's home.

But this story later evolved into: Hillhouse wouldn't participate, but Zhang Lei personally would invest. Reasons cited include Hillhouse's fundraising cycle and regulatory considerations.

Ultimately, neither Zhang Lei nor Hillhouse appeared on the list.

Sequoia China, on the other hand, was almost considered a non-participant from the beginning. The most widespread version cited concerns about overseas LPs. We asked several people close to Sequoia, and they expressed similar sentiments. Of course, other explanations are also circulating.

The bottom line is that Sequoia and Hillhouse, the two institutions that seemed almost certain to be involved, ultimately were not. A senior lawyer told us that for companies with an RMB structure, theoretically, investors cannot be hidden.

Among the established VCs, only IDG invested.

So someone asked us a question in return: Could DeepSeek potentially become Sequoia and Hillhouse's biggest miss?

It's hard to say. However, based on our understanding, DeepSeek might not just raise this one round.

100 Institutions and Individuals

Looking at the investor list, only 10 parties are listed as participating. But "elsewhere" did a simple look-through of the fund entities and found that the actual number of participants is quite high.

For example, the external CVC related to the CATL system—Puyuan Capital's backers include: state-owned assets from Xiamen, Ordos, and other regions; entities related to CATL; and the National Green Development Fund.

Several IDG Capital funds related to this transaction have strong insurance capital characteristics and account for a very high proportion of the 30 billion. Additionally, there are entities related to Nongfu Spring, Septwolves, etc.

Monolith's current LPs are mainly state-owned assets from Zhejiang and Shanghai, as well as listed companies like By-health and Yanghe Distillery.

Furthermore, iHealth Labs is also a participant. In fact, iHealth is also an LP behind many GPs.

A rough estimate, after a shallow look-through and simple deduplication, suggests that nearly 100 institutions or individuals participated in this round of DeepSeek's financing.

The Key Requirement: Don't Poach Talent

We won't reiterate DeepSeek's investment terms here. Just one addition: whether dealing with large corporations or VC funds, Liang Wenfeng's most important requirement is: do not poach DeepSeek's people or encourage them to start their own ventures.

"China's Largest Listed Company"

Returning to viewing this as an investment, what kind of odds does DeepSeek really have?

One participating investor once worriedly said, could this be an investment with too much consensus? According to the experience of the investment industry, consensus outcomes are often mediocre.

But it's hard to say that about DeepSeek. Or rather, we cannot simply understand it through a conventional investment lens, whether in terms of its significance for China's AGI industry or its spiritual meaning.

As one investor put it: "They are doing all this with immense goodwill. He seems to be telling us: this is how the world should be."

So while some hesitated, others believed, "there's no reason not to invest"—if it does IPO as some expect, could DeepSeek potentially become China's largest listed company by market cap on the A-share market?

Low Profile and Ambition

While following DeepSeek's financing progress, one prominent feeling stood out: for most people, DeepSeek was probably a completely unexpected answer. Even for the most competitive investors, they never imagined it would open up for fundraising one day.

So as mentioned, most of the well-known investment veterans only recently connected with the company. Objectively, Liang Wenfeng has indeed long avoided meeting with investors.

When news of the fundraising spread, many institutions didn't even attempt to compete for an investment opportunity. But looking at the result, some got their ticket based on synergy, some on insight, some on brand spirit, and some simply on persistence.

"elsewhere" tried to ask some investors, but most didn't want to discuss this investment much. The sensitivity of the deal is one reason, but more so, it seems to be a conscious desire to maintain this low profile, stemming from a sense of shared honor.

One participating investor said that throughout the process, he cherished every document, hoping to have them framed someday. Another investor repeated twice the 16-character phrase released with the DeepSeek V4 preview:"Not lured by praise, not frightened by slander; act according to the Way, upright and correct oneself."

Trending Cryptos

Related Questions

QWhat is the main topic of the article 'DeepSeek Financing Story'?

AThe main topic of the article is the details and background stories behind the latest financing round of the AI company DeepSeek, highlighting key events, investor meetings, and the company's philosophy.

QWho is Liang Wenfeng and what is his core message during the four-hour investor meeting?

ALiang Wenfeng is the founder of DeepSeek. During the four-hour meeting, his core message was to focus solely on AGI, emphasize team stability, and advocate for a 'less is more' and 'restrained' approach, stating that 'our core is to ensure team stability—this is more important than money and resources.'

QWhy were venture capital giants like Sequoia China and Hillhouse Capital absent from DeepSeek's investor list according to the article?

ASequoia China was reportedly absent due to concerns related to its overseas LPs. Regarding Hillhouse Capital and its founder Zhang Lei, various reasons are mentioned, including the firm's fundraising rhythm and regulatory factors, but ultimately neither participated in this round.

QWhat was the primary non-financial requirement DeepSeek had for its investors?

ADeepSeek's primary non-financial requirement was that investors, including large corporations and VC funds, must not poach DeepSeek's employees or encourage them to leave to start their own ventures.

QWhat is the article's perspective on the broader significance of DeepSeek beyond just an investment opportunity?

AThe article suggests DeepSeek's significance extends beyond a simple investment. It highlights its importance for China's AGI industry and its inspirational value, describing the team as acting with 'great goodwill' and potentially becoming 'the largest listed company in China's A-share market' if it IPOs.

Related Reads

Anchorage Unveils Plan to Protect Against Quantum Threats

Regulated crypto bank Anchorage Digital has presented a strategy for preparing institutional assets for the post-quantum era, including implementing hybrid protection for internal TLS connections. The bank highlights that for cryptocurrencies, the main quantum risk lies in digital signatures, as Shor's algorithm could theoretically derive a private key from a public one. To mitigate this, Anchorage uses Bitcoin addresses based on a hash of the public key, which keeps the key hidden until a transaction is sent, and practices minimal address reuse. Internally, the company has deployed post-quantum key encapsulation for employee devices on ChromeOS and is implementing post-quantum cryptography within its hardware security modules (HSMs). A key proposal is the "Post-Quantum Turnstile" mechanism, which uses STARK zero-knowledge proofs to allow users to migrate signing rights to a post-quantum key without exposing their old private or public key. This could help secure an estimated two-thirds of circulating coins that might otherwise become inaccessible if the network phases out classical signatures. Furthermore, Anchorage open-sourced its Rust implementation of the SQIsign digital signature scheme, which it claims offers signatures about seven times smaller than the Falcon algorithm. This comes amid warnings that around 6 million BTC in addresses with exposed public keys could be vulnerable to future quantum attacks.

cryptonews.ru13m ago

Anchorage Unveils Plan to Protect Against Quantum Threats

cryptonews.ru13m ago

Kimi K3, which used to require 16 B200s, now fits on just 8 AMD cards

This article highlights a key achievement for AMD in the AI hardware race. The company's MI355X GPUs, each with 288 GB of memory, successfully deployed the massive 2.8 trillion parameter Kimi K3 model on a single 8-GPU server. In contrast, the NVIDIA B200 (with 192 GB per card) required a two-server, 16-GPU setup to hold the model, leading to inter-node communication overhead. In performance tests for a 1024-input/400-output token task, the 8-card MI355X system achieved a total throughput of 952 tokens/s and a single-user generation speed of 118 tokens/s. This single-node throughput was approximately 3.8 times higher than the per-node average of the dual-node B200 setup (498 tokens/s total). While NVIDIA's B300 delivered higher absolute performance (1568 tokens/s on 8 cards), a cost-efficiency analysis based on assumed hourly rates showed MI355X offered better value per dollar. Notably, the deployment on AMD's ROCm software platform was relatively straightforward, requiring only minor fixes like patching a missing function for speculative decoding and a simple zero-padding workaround to optimize a prefill kernel for attention heads. This significantly reduced the Time-To-First-Token (TTFT). The article concludes that for extremely large models, memory capacity is becoming a critical differentiator. AMD's strategy of equipping cards with more HBM memory provides a tangible system advantage in single-node deployment efficiency and cost, posing a growing challenge to NVIDIA's CUDA ecosystem dominance.

marsbit14m ago

Kimi K3, which used to require 16 B200s, now fits on just 8 AMD cards

marsbit14m ago

Trading

Spot

Hot Articles

Discussions

Welcome to the HTX Community. Here, you can stay informed about the latest platform developments and gain access to professional market insights. Users' opinions on the price of S (S) are presented below.

活动图片