DeepSeek Completes Nearly 100 Billion Yuan Financing at a Pre-Money Valuation of 500 Billion Yuan

[[**DeepSeek Completes Nearly 100 Billion Yuan Financing at a Pre-Money Valuation of 500 Billion Yuan; Founder Liang Wenfeng Personally Invests 20 Billion Yuan as a Major Backer, with Industrial Capital Leading the Entry. Funds Are Mainly Bet on Breakthroughs in Domestic Computing Power and API Commercialization; the Company Plans to List on the STAR Market in 2027, but Overall Profitability and the Debugging of Domestic Ten-Thousand-Card Clusters Remain Two Major Tests.**

**Key Points**

**1. Financing Scale Doubles While Valuation Stays Unchanged**

This round's fundraising target was raised from 50 billion yuan to nearly 100 billion yuan, but the pre-money valuation remains at 500 billion yuan. The company chose to increase capital without raising the price and to release more equity, prioritizing the introduction of industrial capital from upstream and downstream sectors such as computing power, chips, and complete vehicles, rather than purely financial investors.

**2. Founder Backs with 20 Billion Yuan, Industrial Capital Enters Collectively**

Liang Wenfeng personally contributed about 20 billion yuan, making him the largest contributor. CATL's cumulative contribution may reach 15 billion yuan, making it the largest external investor; Tencent contributed 10 billion yuan; top financial VCs such as Sequoia and Hillhouse were absent throughout, reflecting a strategy of prioritizing industrial allies and downplaying financial capital.

**3. API Gross Margin at 82.9%, but Overall Still a Large Loss**

Revenue in the first seven months of 2026 was 475 million yuan, 10 times that of the full year 2025, and the API business's gross margin of 82.9% leads overseas top manufacturers. But the net loss in the same period was 715 million yuan, investment in computing infrastructure was as high as 11 billion yuan, and pre-training R&D and cluster construction continued to eat into profits.

**4. Tens of Billions in Funds Bet Big on Domestic Computing Power**

The largest investment is the GW-level intelligent computing center in Ulanqab, which plans to deploy 160,000 Huawei Ascend 950DT chips, with procurement costs of 17 billion to 40 billion yuan. There is still no mature case of training large models on ten-thousand-card-level domestic chips, and the supply pace and delivery cycle further amplify project risks.

**5. IPO Path Finalized, Two Major Tests Remain**

The company plans to submit a prospectus at the end of 2026 and list on the STAR Market in 2027. Whether it can deliver on the valuation depends on whether the A-share issuance window can cover the 500 billion valuation, and whether it can complete debugging of the domestic ten-thousand-card cluster and produce implementable results before listing. 2026-10-20 12:34 In-Depth | DeepSeek Raises Nearly 100 Billion: Who Is Betting on Chinese AI, and What Are They Betting On? Sikai Has Something to Say Quick Overview

This article is from the WeChat public account: Sikai Suisui Kan, author: Sikai, original title: "In-Depth | DeepSeek Raises Nearly 100 Billion: Who Is Betting on Chinese AI, and What Are They Betting On?"

In October 2026, DeepSeek's new round of financing is about to close.

This round has set a guaranteed fundraising amount of 80 billion yuan, with the final scale expected to approach 100 billion yuan, directly doubling the initial fundraising target of 50 billion yuan. The company's pre-money valuation is locked at 500 billion yuan.

Against the backdrop of the entire AI industry generally continuing to burn money and with unclear commercialization prospects, a large-model company securing nearly 100 billion yuan in massive financing has a capital logic worth unpacking: where the funds come from, how the valuation is priced, what the core of the capital bet is, and how this 100-billion-yuan financing will change the industry landscape?

Financing Scale Doubles: Capital Increase Without Price Increase, Industrial Capital Rushes for Shares

This round's fundraising amount was raised from 50 billion yuan to nearly 100 billion yuan, driven mainly by both supply and demand.

Demand side: with the launch of the V4.1-Flash model, API call volume has continued to rise. In the first seven months of 2026, the company recognized revenue of 475 million yuan, 10 times the full-year revenue of 2025. Revenue is growing rapidly, but investment in computing power procurement, model R&D, and business expansion is growing even faster, and fundraising demand has swelled accordingly.

Supply side: new and old shareholders split the allocation equally, old shareholders firmly follow on, industrial capital lines up to enter, and high-quality large-model assets are scarce, further driving up subscription enthusiasm.

It is worth noting that the financing amount doubled, but the pre-money valuation remains unchanged at 500 billion. The company chose to increase capital without raising the price, preferring to release more equity rather than blindly inflate the paper valuation, and prioritizing the introduction of industrial parties across the upstream and downstream of computing power, chips, complete vehicles, and cloud computing, rather than simply absorbing financial capital.

But the pressure of the high valuation is also prominent: a pre-money valuation of 500 billion corresponds to revenue of 475 million yuan and a net loss of 715 million yuan in the first seven months. To exit smoothly through an IPO, it must present a sufficiently hardcore long-term growth story.

Investor Lineup: Founder's Heavy Bet Endorsement, Industrial Capital Enters Collectively

In this round's capital list, the largest contributor is not Tencent or CATL, but founder Liang Wenfeng himself, with a personal contribution of about 20 billion yuan.

Among external institutions, Tencent contributed 10 billion yuan; CATL invested 5 billion yuan in the first round and added 10 billion yuan in this round, for a cumulative contribution of up to 15 billion yuan, becoming the largest external investor; NetEase, JD.com, Lisi Capital, and IDG Capital each invested 3 billion yuan.

The new shareholders are uniformly industrial capital: Shixi Capital co-founded by GigaDevice, SMIC's Zhongxin Juyuan, Legend Capital, CPE Yuanfeng, and Geely Group; the National Artificial Intelligence Industry Investment Fund also participated. Top financial VCs such as Sequoia and Hillhouse were absent throughout, fully in line with Liang Wenfeng's consistent financing strategy: prioritizing industrial allies and downplaying purely financial investors.

The investment goals of each industrial capital are very clear:

CATL is betting on computing-power-electricity synergy, laying out energy storage and power supply infrastructure for intelligent computing centers;

Geely is targeting in-vehicle large models to implement intelligent driving business;

Zhongxin Juyuan and Shixi Capital hope to use DeepSeek's ten-thousand-card-level training cluster to validate domestic chip performance on a large scale.

Commercialization Status: API Gross Margin Leads the World, Overall Still Burning Cash on a Large Scale

Against the backdrop of losses across the entire industry, DeepSeek has delivered a highly instructive commercialization report card.

Data from foreign media outlet The Information shows that in the first seven months of 2026, the company's overall gross margin was 44.6%, with the gross margin of its API interface business reaching as high as 82.9%.

Compared with industry levels: gross margins at mainstream SaaS companies generally range from 70% to 80%; OpenAI's gross margin in the first quarter was only 39%, and Anthropic's full-year expected gross margin is 63%. DeepSeek's API business profitability has already pulled ahead of leading overseas vendors, and its low-cost technology route has been validated by the market.

The other side of the impressive gross margin is massive capital expenditure. In the first seven months, the company posted a net loss of 715 million yuan; investment in computing infrastructure alone reached 11 billion yuan, whereas for the whole of 2025 this investment was only 1.2 billion yuan.

To put it simply: the standalone API business can already generate stable profits, but pre-training R&D, the construction of computing clusters, and team expansion are still continuously eating into profits. The 82.9% high gross margin is only an interim result, and there are still three hurdles to clear before overall profitability.

Where the funds are going: a high-stakes bet on domestic computing power, building an independent training foundation

The vast majority of the 100 billion yuan in financing will be invested in computing infrastructure, with the largest single investment being the GW-level intelligent computing center project in Ulanqab, Inner Mongolia.

The project plans to deploy 160,000 Huawei Ascend 950DT chips, with chip procurement costs ranging from 17 billion to 40 billion yuan. This is also the core reason why outsiders define this investment as a "high-stakes gamble": large-scale training of large models on ten-thousand-card-level domestic chips has yet to see a mature success case in the industry.

Once the project runs successfully, DeepSeek will completely break free from the constraints of Nvidia's computing power and gain autonomy over underlying hardware; if the engineering implementation falls short of expectations, the tens of billions of yuan invested in chips will directly become sunk costs.

In addition, the supply schedule for Huawei Ascend 950DT chips is uncertain, high-end components are in short supply, and delivery cycles exceed one year, further amplifying the project's risks.

The IPO path is set: listing on the STAR Market in 2027, with two major tests ahead

This round of financing also finalized the listing timetable:

Yan Wentao, a former senior executive at Hillhouse, joined in September as CFO; CITIC Securities has already stepped in to sort out the shareholder structure and adjust fund filings to meet the STAR Market's shareholder count requirements; the company plans to submit its prospectus by the end of 2026 and list on the STAR Market in 2027.

To successfully realize the 100 billion yuan valuation, two thresholds must be crossed:

First, whether the A-share IPO window can open, and whether the issue price can cover the 500 billion yuan pre-money valuation plus capital costs;

Second, whether the domestic ten-thousand-card computing cluster can be debugged before the listing and deliver implementable large-model training results. Neither matter carries absolute certainty.

Behind the 100 billion yuan financing, what are investors really betting on?

Taking into account the capital structure, use of funds, and commercialization pace, this round of 100 billion yuan financing sends four major industry signals:

Investment leadership has completely shifted: AI investment has moved from being led by financial VCs to being led by industrial capital; the founder has put 20 billion yuan of his own money into a heavy bet, firmly holding the company's direction of development and refusing to be held hostage by capital.

Commercialization has achieved a partial breakthrough: the API business has already produced a high-gross-margin profit model, proving that selling large models to businesses is commercially viable; but at the company-wide level, it will still need to keep burning cash on infrastructure in the short term.

Domestic computing power has entered the practical combat stage: betting tens of billions of yuan on Ascend chip clusters is essentially a technological breakthrough effort, with the goal of breaking the overseas monopoly on computing power and building an autonomous and controllable training foundation.

The ultimate question investors are betting on: whether AI companies can rely on their own API revenue to sustainably fund model R&D and no longer be "choked" by overseas chips.

The 82.9% API gross margin is the first positive signal from this high-stakes gamble. But to achieve overall profitability and deliver on the IPO valuation, DeepSeek still has a long road of infrastructure and technological climbing to complete.

This nearly 100 billion yuan AI gamble is essentially a two-way contest between breakthroughs in domestic computing power and commercialization implementation. The AI industry has bid farewell to the era of stories and entered a hardcore elimination tournament of capital, technology, and production capacity. AI Venture Capital Daily Channel: Frontier Technology. This content is published with the author's authorization; the views represent only the author and do not represent Huxiu's position. If you have 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[ ]