AI News

DeepSeek Completes Two Rounds of Financing Totaling About 150 Billion Yuan in Half a Year, and Prepares for a STAR Market IPO

HuxiuUpdated Nov 13 22:00

AI summary

According to a Bloomberg report, DeepSeek's second financing round was completed on October 22, with a scale of about 100 billion yuan and a pre-money valuation of 500 billion yuan; together with the first round of about 51 billion yuan settled in June, it raised a total of about 150 billion yuan within half a year. Tencent and CATL are the external shareholders that contributed the most money in the two rounds. Liang Wenfeng's combined shareholding remains above 75%, and most external institutions hold shares indirectly through Hangzhou Chengli and are locked in for five years. The funds are mainly invested in computing power. The company plans to build a gigawatt-level data center in Inner Mongolia and deploy at least 160,000 Huawei Ascend 950DT chips. It has also approached investment banks to prepare for a STAR Market IPO, planning to submit its listing application as early as the end of this year and to list in 2027.

Why read this

SelectedAI score 83

The original article breaks down the funding structure of the two financing rounds, the control arrangements, and the computing power procurement constraints, making clear the turning point for this company from refusing financing to preparing for an IPO.

Topics

Tags

Body · AI translation

[[**DeepSeek completes about 150 billion yuan in two funding rounds within six months; the second round raises 100 billion yuan at a pre-money valuation of 500 billion yuan, with Tencent and CATL continuing to hold heavy positions; the funds are mainly invested in computing power, and a STAR Market IPO is being prepared.**

**Key points**

**1. Two funding rounds total about 150 billion yuan, with control concentrated**

The first round was about 51 billion yuan and the second about 100 billion yuan, lifting the pre-money valuation to 500 billion yuan; Liang Wenfeng's combined stake remains above 75%, and most external institutions hold indirectly with a five-year lock-up.

**2. Talent attrition and computing-power investment force fundraising**

Ten of the authors of the V4 report have already left, with Wang Bingxuan, Guo Daya and others poached by big tech; the lack of a fair price for options, plus self-built machine rooms and a gigawatt-level data center, prompted Liang Wenfeng to raise funds.

**3. Computing-power expansion is constrained by chip supply**

There are plans to build a gigawatt-level data center in Inner Mongolia deploying at least 160,000 Ascend 950DT chips, but Huawei can initially supply only about 16,000 and has raised prices; according to reports, the two sides are already deeply bound, with pre-training to run as early as the end of the year.

**4. STAR Market IPO advances, with long investor lock-ups**

This article is from the WeChat public account: Dongcha Beating, author: Sleepy

On October 22, according to Bloomberg, DeepSeek's second funding round was officially completed, at a scale of about 100 billion yuan and a pre-money valuation of 500 billion yuan.

Counting the first round of about 51 billion yuan settled in June, DeepSeek raised about 150 billion yuan in just half a year, and across the two rounds the largest external shareholders were Tencent and CATL.

From July to October, the second round was suspended once, with the fundraising target changed from an initial at least 10 billion yuan all the way to 50 billion yuan and 80 billion yuan, finally settling at 100 billion yuan.

Half a year ago, this would have been almost unimaginable. In its first three years, DeepSeek took not a cent from outside, with the bills for R&D and computing power entirely carried by Liang Wenfeng's quantitative private fund High-Flyer. In a 2024 interview, Liang Wenfeng also said DeepSeek had no short-term fundraising plans.

This spring, DeepSeek ran into talent problems, which made Liang Wenfeng change his mind.

In the V4 technical report released at the end of April, 10 of the nearly 300 authors had asterisks after their names, indicating they had left. Wang Bingxuan, a core author of the first-generation large model, went to Tencent; Guo Daya, a core researcher on R1, left in March and was reported in April to have joined ByteDance Seed; Luo Fuli, who left earlier, went to Xiaomi to lead the MiMo large model.

It is said that big tech has been poaching DeepSeek's people at high prices. At that time, the options in DeepSeek employees' hands still had no fair market price. A piece of paper without a price tag cannot hold back real money.

Liang Wenfeng later leveled with investors, saying that keeping the team stable is the company's only non-negotiable interest. On the other side, DeepSeek was already building its own machine rooms and intended to manage its own computing power, and relying on High-Flyer's resources alone would make it hard to sustain consumption of that magnitude going forward.

According to CNBC, it was precisely after the gradual loss of core talent that DeepSeek made up its mind to open up to fundraising. Setting a sufficiently high fair value for options in the primary market became Liang Wenfeng's lavish bid to attract talent.

Tickets

Once the news got out, the door was immediately crowded with people wanting to hand over money.

To meet Liang Wenfeng once, one investor spent 5 million yuan on a meeting fee and enlisted an intermediary who claimed to be close to Liang Wenfeng to make an introduction, making three trips to Hangzhou without ever meeting him, and finally got the money back. A fund with state-owned capital backing sent someone with a 3 billion yuan intention to find High-Flyer executives, only to receive a company email address and hear nothing more. On social media, intermediaries even peddled quotas of unknown origin, daring to charge 17% in upfront fees.

The final list of those who got tickets was chosen by Liang Wenfeng himself, with most institutions invited by him on his own initiative. Funds that neither knew him nor had enough brand clout could not even get in the door.

The agreement was signed at the end of May and settled on June 16, with the first round set at about 51 billion yuan.

Of this money, the largest portion came from Liang Wenfeng personally, contributing about 20 billion yuan. Among external shareholders, Tencent put in about 10 billion yuan and the CATL system about 5 billion yuan; NetEase, JD.com, IDG Capital and Monolith each took about 3 billion yuan; Zhengxin Valley and Shixiang Capital each about 1.5 billion yuan, and the National Artificial Intelligence Industry Investment Fund about 1 billion yuan.

In just two months, DeepSeek's value more than doubled. In late April, the market valuation was still around 20 billion US dollars; by the first-round settlement in June, the pre-money valuation was set directly at about 350 billion yuan. On that basis, the company carved out a new 5% employee option pool, which, converted at the first-round price, instantly had a book value of about 17.5 billion yuan. Counting this option pool, the actual pre-money valuation was about 367.5 billion yuan, or about 54.3 billion US dollars, and close to 400 billion yuan post-money.

However, although external capital came in, the control Liang Wenfeng gave away was minimal. The vast majority of external investors could only invest in Hangzhou Chengli, a limited partnership managed by Liang Wenfeng, getting on board through indirect shareholding, with shares locked up for five years. The only external institution to directly enter the shareholder register and obtain voting rights was the National Artificial Intelligence Industry Investment Fund.

The industrial and commercial change records in July show that Hangzhou Chengli held 8.52%, with Tencent accounting for more than 30% of that; the National Artificial Intelligence Industry Investment Fund held 0.28%. Liang Wenfeng's direct stake was diluted from 34% to about 31%, and counting the indirect stake held through Ningbo Cheng'en, his combined holding remained above 75%.

More volume at the same price

Just as the first round closed, the second round of negotiations started in mid-July. The initial word was that the target was at least 10 billion yuan, with a pre-money valuation of no less than 480 billion yuan.

But by the end of July, DeepSeek suddenly verbally notified several investors who were preparing to sign that the agreements originally scheduled to be signed within a few days were being paused. It was said that the leak of the contents of the first round's closed-door meeting with investors had caused too much of a stir in the market, and Liang Wenfeng wanted to keep the second round more low-key.

In early August, the second round quietly restarted beneath the surface.

This time it proceeded very quietly, and several institutions that had been at the negotiating table in July did not even receive notice of the restart. The fundraising target was directly raised to 50 billion yuan, with a pre-money valuation set at about 500 billion yuan.

Although the quota was opened up, screening of the funds' backgrounds was not relaxed at all. Some investors tried to enter by setting up SPVs, but the source of funds and the ultimate upstream contributors behind them were scrutinized layer by layer, and they were ultimately shut out.

Milestone, Shixiang, Zhengxin Valley and Shixi, a mix of new and existing DeepSeek shareholders, have all been intensively raising money for new funds over the past two months, with a string of listed companies subscribing alongside them. Zhihu planned to contribute 1.5 billion yuan, Dahua planned 500 million yuan, Century Huatong's subsidiary subscribed 650 million yuan, and even Zhongwang Fabric, which makes furniture fabrics, put in 50 million yuan.

The money kept piling up. By October 6, the amount DeepSeek had locked in was already close to at least 80 billion yuan; on October 13, the final lineup surfaced, with the new shareholders confirmed as Shixi Capital, SMIC Juyuan, CPE Yuanfeng, Legend Capital and Geely; among existing shareholders, Tencent, CATL, IDG Capital, Milestone Capital and Shixiang Capital continued to follow on.

On October 22, the deal was officially completed. The same day, Perfect World issued two announcements: the two funds it had subscribed to, Milestone Xingshen and Shixiang Shangshi, had completed fundraising, at 7.52 billion yuan and 3.65 billion yuan respectively. Both Milestone and Shixiang were on this round's follow-on investment list.

The final closing size stopped at about 100 billion yuan, ten times the initial target in July. But the pre-money valuation remained locked at 500 billion yuan, more volume at the same price.

Inside and outside the game

If the first round of financing helped DeepSeek complete option pricing, then the second round's shareholder list reveals a strong industrial orientation.

Breaking down the backgrounds of the new shareholders, two come directly from the heart of the chip industry. Shixi Capital is a semiconductor CVC initiated by GigaDevice, and another identity of GigaDevice Chairman Zhu Yiming is none other than the founder of ChangXin Memory Technologies.

It is said that Liang Wenfeng set a check size threshold of 5 billion yuan for direct investment in this round, and the amount Shixi Capital planned to contribute in this round was exactly about 5 billion yuan. Another new entrant, SMIC Juyuan, has SMIC behind it. Add Legend Capital under Legend Holdings and Geely, the only automaker on the list, and the weight of industrial capital is quite obvious.

Among existing shareholders, Tencent put in about 10 billion yuan in the first round, and in this round it again ranked among the largest contributors together with CATL. Back in February 2025, when R1 went viral, Tencent Yuanbao and WeChat Search successively integrated DeepSeek; by the end of April this year, the news that Wang Bingxuan, a former core author at DeepSeek, had joined Tencent was also confirmed. Tencent both took in people who came out of DeepSeek and took on the largest single external share in DeepSeek's first round.

CATL is likewise heavily invested, putting in about 5 billion yuan in the first round and directly doubling that to about 10 billion yuan in this round. In the view of many analysts, CATL's increased bet reflects its focus on the enormous energy demand behind computing clusters.

Equally intriguing are the names not at the table.

Alibaba and ByteDance did not appear in either round; they each have Qwen and Doubao. Hillhouse is likewise not on the shareholder list. However, in September this year, DeepSeek brought in a new chief financial officer, Yan Wentao, a former partner at Hillhouse Ventures who had previously invested in ByteDance, Zhipu and MiniMax.

Ulanqab and buying cards

With 150 billion yuan coming in, DeepSeek has never publicly broken down exactly how this huge sum will be spent, but based on available information, the absolute bulk is still computing power.

As early as the end of April this year, DeepSeek posted job openings for data center operations and maintenance engineers and delivery managers, with the work location directly listed as the Ulanqab grasslands. By early September, more specific news emerged: DeepSeek planned to build a gigawatt-scale data center in Inner Mongolia, housing at least 160,000 Huawei Ascend 950DT chips.

But the problem is that having money does not necessarily mean being able to immediately buy enough cards.

Against that target of 160,000, Huawei could give DeepSeek capacity for only about 16,000 cards, while internet giants received quantities on the order of hundreds of thousands. Liang Wenfeng calculated that four Huawei cards are equivalent to one Nvidia card, while also lagging two generations behind. By September, affected by a shortage of high-bandwidth memory, Huawei raised the price of the 950DT to more than 250,000 yuan per card, an increase of fully 20% to 50% from two months earlier.

In the primary market, money is easy to raise, but cards are hard to buy.

To solve the card bottleneck, the two sides have become increasingly deeply bound together. In September, the prevailing market narrative was still that this batch of Huawei cards would only be used for inference, while core training would still have to rely on Nvidia chips obtained by every possible means. But by October, the word was that Huawei engineers had already been stationed long-term in DeepSeek's offices, writing the requirements DeepSeek encountered during training into the design of the next-generation chip; DeepSeek would begin running pre-training on the 950DT as early as the end of this year, while Huawei planned to reserve most of the first-batch capacity of the 960DT, to be launched in the first quarter of 2027, for it. Of course, DeepSeek has not stopped its efforts to find more Nvidia cards at the same time.

At the end of September, DeepSeek also open-sourced six low-level components for Ascend. On external investment, in August DeepSeek participated in the strategic placement of Unitree Robotics' IPO, subscribing about 141 million yuan, and also planned to lead the investment in Winverse AI, a robotics company founded by former Tencent AI scientist Hu Han.

2029

Beyond computing power and talent, another hidden thread already points to the secondary market, which is also an important reason why former Hillhouse partner Yan Wentao joined as CFO.

As early as late August, DeepSeek began contacting accounting firms and investment banks, busily preparing for a STAR Market IPO, planning to file as early as the end of this year and list in 2027. After Yan Wentao took office, the closing of the second round of financing and the preparations for listing were all matters he was responsible for.

Taking advantage of the window of financing and IPO preparation, the books of this mysterious company were glimpsed by the outside world for the first time. According to financial figures obtained by foreign media, DeepSeek's revenue from January to July this year was about 475 million yuan, 10 times that of the whole of last year, of which the gross margin of the API business reached 82.9%, while the net loss in the same period was about 715 million yuan. By September, the company's ARR had reached about 1 billion US dollars, equivalent to about 6.7 billion yuan.

What underpins this revenue curve is the model and infrastructure iteration that did not stop even during the financing period. On September 10, DeepSeek released V4.1 Flash; the 1017 version updated on October 17 scored 45 on Artificial Analysis' intelligence index, beating Haiku 5.5, which Anthropic had released just ten days earlier; then on October 20, DeepSeek open-sourced its internal inference framework DeepInfer, compatible with both Nvidia and Huawei Ascend 950.

However, for investors who have just squeezed onto the table, there is a more realistic timing issue beyond the books.

According to the current regulations of the China Securities Regulatory Commission, new shareholders who make last-minute investments within 12 months before filing must lock up their shares for 36 months from the date of acquisition. If DeepSeek really files at the end of this year as planned, the new and old institutions that poured 100 billion yuan into this round will not be able to cash out their chips until 2029 at the earliest. The portion that came in through limited partnerships in the first round is subject to a five-year lock-up, which will last until 2031.

What price the secondary market will give three years from now—there are currently two completely different reference points in the market.

One is Zhipu on the Hong Kong stock market, whose share price peaked at HK$2,980 on June 22 and had closed at HK$484.8 by October 20; the other is Changxin Technology on the STAR Market, which surged 465.82% on its first day of trading on July 27, pushing its market value to 3.28 trillion yuan and taking the top spot by market value on the A-share market in one stroke.

Two years ago, Liang Wenfeng said that DeepSeek had no financing plans in the short term, but that remark had already become a thing of the past by this spring. During the first round of financing, he also told investors that the original purpose of founding the company was not to go public. If all goes smoothly, by the end of this year at the earliest, that remark will probably have to be revised as well.

-END-

References

[1] DeepSeek completes ~100 billion yuan funding round, led by CATL and Tencent, Bloomberg

[2] DeepSeek said to tell backers of funding pause after viral posts, Bloomberg

[3] Uncovering DeepSeek: A More Ultimate Story of Chinese Technological Idealism, Anyong Waves

[4] DeepSeek thanks nearly 300 R&D staff, with 10 "departed" key figures prominently listed, National Business Daily

[5] Brokers, shares and tens of billions in funds: A factual record of DeepSeek's financing, Caijing

[6] DeepSeek's first financing round lands, with Tencent, CATL and the National Artificial Intelligence Industry Fund all taking stakes, STAR Market Daily

[7] Announcement on the progress of subscribing to fund shares (Lisi Xingshen), Perfect World

[8] Announcement on the progress of subscribing to fund shares (Shixiang Shangshi), Perfect World

[9] A VC prepared 5 billion yuan for DeepSeek, Chinaventure

[10] DeepSeek reportedly sprinting toward a 500 billion yuan valuation! Annualized revenue reaches 6.7 billion, Zhidongxi

[11] DeepSeek V4.1 Flash 1017 (Max) Intelligence, Performance & Price Analysis, Artificial Analysis

[12] deepseek-ai/DeepInfer, GitHub Channel: Finance and Economics. This content is published with the author's authorization; the views represent only the author and do not represent the position of Huxiu. If you have any objections or complaints about this manuscript, please contact tougao@huxiu.com. Everyone who is changing and wants to change the world is on the Huxiu APP [ ]

Source

Huxiu