DeepSeek Initiates IPO Preparation

Author: Shen Xiao

DeepSeek is going public.

On July 14, 2026, Bloomberg reported that Hangzhou DeepSeek Artificial Intelligence Fundamental Technology Research Co., Ltd. (hereinafter referred to as "DeepSeek") has begun preparations for an IPO, aiming for the STAR Market of the Shanghai Stock Exchange. The application could be submitted as early as late 2026 or early 2027, with the goal of completing the listing in 2027.

After completing the first round of financing, Liang Wenfeng's net worth soared to $36 billion (approximately 244 billion yuan), making him the world's richest AI model founder.

According to sources familiar with the matter, DeepSeek is in preliminary contact with new investors to discuss a second round of financing, with a pre-money valuation target of approximately $71 billion (approximately 480 billion yuan), representing a roughly 37% increase from the first round's valuation.

From a low-profile company that adhered to a "no financing, no commercialization, no roadshows" policy to one now engaged in intensive financing and Sprint (sprinting toward) an IPO, DeepSeek is entering a new phase.

From No Financing to 50 Billion: DeepSeek Opens the Capital Gate

Over the past two years, nearly all Chinese large model companies have been focused on financing, expansion, and market capture. However, DeepSeek has maintained a nearly "anti-trend" restraint, adhering to a "no financing, no commercialization, no roadshows" policy.

Years ago, Chen Tianqiao, China's former richest man, had a four-hour conversation with Liang Wenfeng but was ultimately politely declined for investment. In early 2024, Legend Capital also approached DeepSeek but to no avail. Even after DeepSeek R1 gained significant attention following its release in early 2025, Liang Wenfeng did not rush to introduce external capital.

At the time, the outside world generally believed that DeepSeek was more like a research institution than a startup eager to monetize.

This restraint persisted until this year. From late May to June 2026, DeepSeek completed its first external financing round since its inception, raising approximately 50 billion yuan with a post-money valuation of about $52 billion (approximately 350 billion yuan), setting a record for the first round of financing among Chinese tech startups.

Founder Liang Wenfeng personally participated, contributing about 20 billion yuan of his own funds as the largest single investor in this round. Tencent invested approximately 10 billion yuan; the CATL ecosystem invested about 5 billion yuan; NetEase, JD.com, IDG Capital, and Monolith Capital each invested about 3 billion yuan; Zhenxin Capital and Shixiang Technology each invested about 1.5 billion yuan; while the National Artificial Intelligence Industry Investment Fund directly injected 1 billion yuan into DeepSeek.

The uniqueness of this financing round lies in the fact that investors did not entirely enter the company through traditional means. Except for the National Artificial Intelligence Industry Investment Fund, which took a direct stake, most external funds first entered a limited partnership enterprise where Liang Wenfeng serves as the general partner, which then increased its capital in DeepSeek. Investors are also subject to a long lock-up period and do not have voting rights or board seats.

In other words, capital has entered, but Liang Wenfeng still firmly maintains control of the company.

While the afterglow of the first round of financing has not yet faded, DeepSeek launched a second round of financing in mid-July, with significantly relaxed terms, allowing investors to directly inject capital into DeepSeek's main entity. The company is also considering introducing overseas funds, including those from the Middle East.

This time, DeepSeek aims for a pre-money valuation of at least 480 billion yuan and plans to raise at least 10 billion yuan.

In just one month, DeepSeek's valuation has reached a new milestone. But more worthy of inquiry than the financing amount is: Why does DeepSeek suddenly need so much money?

First is computing power. According to sources familiar with the matter, Liang Wenfeng had been relatively resistant to external funds but realized in April this year, after seeing Anthropic's Mythos model achieve a breakthrough, that to keep up with this level of competition, the company's self-accumulated computing resources were no longer sufficient.

Thus, DeepSeek began building its own computing infrastructure. In June 2026, the company launched a recruitment plan for an IDC data center team to create a GW-level computing park. According to a Reuters report on July 7, DeepSeek has also initiated work on its own AI inference chips and is in contact with chip design companies and wafer foundries.

Beyond computing power, talent is also an increasingly expensive cost. From late 2025 to early 2026, several core R&D personnel left DeepSeek. The competition among major companies for top AI researchers has driven talent costs to new heights.

For a company that does not raise funds, it is difficult to provide core employees with clear wealth expectations. However, a company with a rising valuation and approaching the capital market can retain talent through equity and options.

From this perspective, DeepSeek's financing may not represent an abandonment of its previous technological ideals but rather a push into a new phase where computing power, talent, and organizational costs must be recalculated.

The first return on this capital operation has been realized by Liang Wenfeng himself.

With a Net Worth of $36 Billion, Liang Wenfeng Becomes the World's New Richest AI Founder

The surge in DeepSeek's valuation has directly rewritten Liang Wenfeng's wealth ranking.

According to the Bloomberg Billionaires Index, after the first round of financing, Liang Wenfeng's personal net worth has reached $36 billion. This figure surpasses those of Anthropic co-founder Dario Amodei and OpenAI co-founder Greg Brockman, placing him at the top of the global large model founder wealth rankings and among the top ten richest individuals in China.

However, viewing Liang Wenfeng solely as an entrepreneur who quickly profited from the AI boom would greatly underestimate him. In fact, before DeepSeek, Liang Wenfeng had already achieved a mature entrepreneurial success.

Born in 1985 in Wuchuan, Zhanjiang, Guangdong, Liang Wenfeng enrolled in the Electronic and Information Engineering program at Zhejiang University at age 17 in 2002. He later pursued a graduate degree in Information and Communication Engineering, specializing in machine vision.

During the 2008 global financial crisis, Liang Wenfeng began collecting financial market data with classmates and experimented with automated quantitative trading using machine learning techniques. In 2013, Liang Wenfeng and his Zhejiang University alumnus Xu Jin founded Jacobi Investment Management in Hangzhou. In 2015, they co-founded Hangzhou High-Flyer Technology, later known in the industry as High-Flyer Quant.

Unlike traditional fund managers, Liang Wenfeng early on applied artificial intelligence and machine learning technologies to quantitative investing.

By 2021, High-Flyer Quant's assets under management had exceeded 100 billion yuan, making it one of the leading domestic quantitative private funds.

This accumulation also explains why DeepSeek could remain independent of external financing for an extended period in its early stages: While other large model companies were seeking capital, Liang Wenfeng had a stable cash flow from his own ventures to support the company.

In May 2023, at age 38, Liang Wenfeng announced his commitment to general artificial intelligence research. In July of the same year, he founded DeepSeek, with the first phase of R&D funding of approximately 3 billion yuan provided by High-Flyer.

At the time, OpenAI had about 1,200 researchers, and Anthropic had over 500, while Liang Wenfeng led a purely domestic team of just over a hundred people.

In May 2024, DeepSeek released the V2 model, reducing API prices to 1 yuan for input and 2 yuan for output per million tokens, about one-hundredth of GPT-4 Turbo's price, quickly triggering an industry price war and earning DeepSeek the nickname "the Pinduoduo of AI."

However, what truly brought DeepSeek into the global spotlight was the R1 model released in early 2025. It demonstrated reasoning abilities in a manner different from traditional large models, making the world recognize anew that top-tier AI models do not necessarily rely solely on massive funding and unlimited computing power.

This shock caused NVIDIA's stock to plummet 17% in a single day, wiping out nearly $600 billion in market value, an event later referred to as the "DeepSeek Moment" by the outside world.

Since then, DeepSeek has quickly become the focus of the global AI industry, while Liang Wenfeng himself has remained low-key, rarely giving interviews or publicly discussing his personal wealth. "Model openness, founder silence" has become his most contrasting personal label.

This low-key demeanor is, to some extent, also characteristic of DeepSeek as a company.

From Technological Ideals to the Capital Market: DeepSeek Faces New Challenges

DeepSeek was not initially a company focused on commercialization.

Liang Wenfeng had previously worried that external capital would interfere with the company's strategy, pushing for premature commercialization and thus affecting long-term technological research. Therefore, during the peak of financing activities, he chose to reject investments.

However, by 2026, the competitive environment DeepSeek faced had completely changed. The large model industry was transitioning from isolated technological breakthroughs to a phase of comprehensive competition involving computing power, talent, products, and ecosystems.

On one hand, competitors were rapidly raising funds and going public. Zhipu and MiniMax had already entered the capital market, while companies like Jieyue Xingchen and Yuezhiyanmian were advancing new rounds of financing. On the other hand, internet giants were fully entering the large model competition with their capital, computing power, and user access points.

The advantages DeepSeek had previously established through "low cost, high efficiency, and open-source models" were facing new tests.

In the past, it could create globally leading models with a relatively restrained R&D team and limited computing resources. However, in the next phase, computing power demands would continue to grow, and top talent would become increasingly expensive. This may be one of the reasons why Liang Wenfeng ultimately opened the door to capital.

DeepSeek's management has told investors that the company will continue to prioritize cutting-edge AI research over short-term commercialization. Liang Wenfeng has also committed to continuing the development of open-source models and pursuing artificial general intelligence as a long-term goal.

Nevertheless, DeepSeek has begun to explore extensions into the application layer. In May 2026, the company started forming the "Harness" code agent team, with job postings indicating its goal of transforming cutting-edge model capabilities into agent products.

This suggests that DeepSeek may be evolving from a company focused solely on foundational model research toward a more complete AI technology and product system.

For a company already valued in the hundreds of billions of yuan, focusing solely on research is no longer sufficient. It must answer several questions: How to retain core talent? How to secure sustained computing power? How to convert technological advantages into stable revenue? And how to find a balance between open-source models, commercialization, and the capital market?

DeepSeek's financing and IPO preparations are, to some extent, responses to these questions.

References:

"DeepSeek's First Round of Financing: Liang Wenfeng Invests Approximately 20 Billion Yuan," Cailian Press;

"Liang Wenfeng's 20 Billion Yuan Investment Makes Headlines: DeepSeek's First Round of Financing Is About More Than Just Money," Yicai;

"DeepSeek's Financing Nears Completion: Liang Wenfeng 20 Billion, Tencent 10 Billion, CATL 5 Billion," Zhidx;

"DeepSeek's Financing Rewrites Everyone's Valuations," Huxiu;

"Wait a Bit Longer for Liang Wenfeng," Investment Interface;

"DeepSeek's Latest Valuation Exposed!," China Fund News;

"DeepSeek's Annualized Revenue Nears $500 Million, IPO Expected as Early as 2027," Tencent Technology.

THE END

Follow Us for More Industry Insights

Copyright and Disclaimer

1. Content Copyright: Except for quoted public data, policies, and cases, all content is original. Professional data is sourced from authorized databases and government websites, while cases are compiled from real events.

2. Image Licensing: Some images in the article are proprietary or officially licensed, while others are AI-generated. For network images with unclear copyright, the rights belong to the original authors, and any infringement will be removed upon notification.

3. Republication Guidelines: Unauthorized republication is prohibited. Republished content must retain the complete source and author information.

4. Liability Disclaimer: This article is a commercial figure observation and industry commentary compiled by the author based on public information. The content is for reference only and does not constitute professional advice. Any risks arising from its use are borne by the user. Xinzhidongneng reserves the final interpretation rights for this article.