Richer than Altman and Dalio combined: Liang Wenfeng, the "hidden richest man" in the global large model industry.

In the AI large model track, the world's most capital-intensive and most sought-after by investors, if you ask who is the most powerful founder, people will most likely think of Sam Altman of OpenAI or Dario Amodei of Anthropic.

But if the evaluation criterion is changed to "personal book wealth", the star founders in Silicon Valley may not be at the top of the list.

According to statistics from financial research institution BestBrokers based on Forbes' real-time rich list, the net worth of Liang Wenfeng, founder of DeepSeek, soared from about 1 billion US dollars to 39.5 billion US dollars between July 2025 and July 2026, ranking among the top 50 on the global rich list. In the same period, Forbes estimated Sam Altman's net worth in the range of about 3.2 billion to 4 billion US dollars, and his wealth mainly comes from venture capital layouts outside OpenAI, and he does not hold OpenAI's equity himself. There is no public estimate of Dario Amodei's personal net worth based on the same caliber; the total publicly disclosed compensation in 2025 is about 18 million US dollars, mainly from stock and option awards. Due to different statistical calibers, it is difficult to make an accurate arithmetic comparison of the "sum of net worth" of the three, but Liang Wenfeng's lead in the magnitude of book wealth is clear.

Behind this contrast lies the profound differences in the capital structure and governance logic of Chinese and American AI large model companies.

Differences in Equity Structure: Decentralized Dilution and High Concentration

The fact that Altman does not hold OpenAI's equity has been publicly confirmed by himself on many occasions. He controls this company valued at hundreds of billions of US dollars relying on the authorization of the board of directors and the governance structure, and his personal wealth comes from investments during the YC period as well as external shareholdings in Stripe, Reddit, Helion and other enterprises.

As co-founder and CEO of Anthropic, Dario holds the company's equity, but Anthropic has gone through multiple rounds of large capital injections from giants such as Amazon and Google. According to the disclosure of its IPO prospectus, Dario and co-founder Daniela Amodei promised to donate 80% of their respective personal equity to charity, and set up a public welfare fund accounting for 5% of the shares. The superposition of capital injection from giants and equity donation commitments has imposed multiple constraints on his personal shareholding ratio and realizable wealth.

Liang Wenfeng's equity path is different. In the first few years after its establishment, DeepSeek did not carry out external equity financing, and was supported by funds from High-Flyer Quant and Liang Wenfeng himself. But this state changed in 2026: in June 2026, DeepSeek completed the first round of financing since its establishment, with the media-reported financing amount at the level of about 500 billion RMB, and Liang Wenfeng himself is one of the largest investors in this round. Before the financing, Liang Wenfeng controlled a total of about 84.29% of the equity and nearly 100% of the voting rights through direct and indirect methods; after the completion of the financing, his shareholding ratio decreased to some extent, but he still maintained a leading position in the corporate governance. He is not an "absolute major shareholder who has never raised financing", but a founder who still maintains high control over the company after introducing external capital.

Where Does the Confidence Come From: Quantified Profits and Early Computing Power Investment

The foundation of Liang Wenfeng's wealth does not lie in the valuation story of large models, but in the financial business of High-Flyer Quant. High-Flyer Quant was once one of the top quantitative private equity firms in China, with its management scale once exceeding 100 billion RMB. In 2025, based on the management scale of about 700 billion RMB, High-Flyer achieved an annualized return of more than 56%. Calculated according to industry practice, the performance remuneration alone may generate considerable cash flow. Liang Wenfeng stopped introducing external funds for High-Flyer in his early years and held most of the equity of the company, which enabled the profits of High-Flyer to be fully converted into R&D resources for DeepSeek.

The layout in terms of computing power requires a more precise division of the timeline. The total investment of the "Firefly No.1" training platform built by High-Flyer from 2019 to 2020 is nearly 200 million RMB, equipped with about 1100 GPUs. The "Firefly No.2" invested with about 1 billion RMB in 2021 is equipped with about 10,000 NVIDIA A100s. At that time, there were no more than 5 enterprises in China that held GPU clusters at the 10,000-card level. It is inaccurate for the original text to directly correspond "Firefly No.1" with the 10,000-card A100; it is "Firefly No.2" that has formed the 10,000-card reserve. This advance layout constituted a significant computing power buffer in the large model entrepreneurship boom in 2023, which is valid.

Two Games of Capital: External Constraints and Internal Freedom

OpenAI and Anthropic represent the typical external capital-driven path of Silicon Valley AI companies: continuous financing, pushing up valuations, and deep binding with cloud giants. Anthropic's prospectus discloses that the total amount of cloud computing agreements it has promised in the next ten years is about 4.5 trillion US dollars, of which about 1.11 trillion US dollars is for Google Cloud, about 1.1 trillion US dollars for Amazon AWS, and about 310 billion US dollars for Microsoft, and most of them are irrevocable long-term obligations. While this structure endows the company with huge computing power, it also forms rigid financial commitments and capital return pressure.

Before the financing in 2026, DeepSeek was indeed in a state of "not needing to deliver results to external VCs" for a long time. Even after the completion of the first round of financing, Liang Wenfeng still retained highly concentrated voting rights, and external investors have relatively limited say in governance. This structure gives him great strategic freedom: he is in no hurry to commercialize, open sources the model weights, and fully open sources the internal inference framework DeepInfer in October 2026. DeepSeek's open source strategy and low-cost API pricing have objectively reshaped the industry's cost expectations, but it is a narrative simplification to attribute it entirely to "using personal wealth to support the AGI dream"; more accurately, the highly concentrated control reduces the obedience to short-term capital returns, but does not eliminate the constraints of commercial sustainability.

Differences in Wealth View: Freedom Brought by Control

In the current AI entrepreneurship ecosystem, founders often need to repeatedly weigh between financing rhythm, valuation expectations and commercialization timelines. The reason why Liang Wenfeng's situation seems special is not that he "never deals with capital at all", but that he still retains the control to determine the direction of the company after introducing capital. This control, superimposed on the financial foundation accumulated during the High-Flyer Quant period, enables him to have more maneuvering room than most peers in model architecture selection, open source rhythm and talent recruitment.

Conclusion

The statement that "Liang Wenfeng is richer than Altman and Dario combined" is generally true in terms of book figures, but a limitation needs to be added: the sources of wealth and statistical calibers of the three are different, and the precise sum comparison lacks a unified benchmark. What is really worth noting is not the ranking itself, but the structural difference it reflects: in the AI race, the amount of book wealth depends to a large extent on the founder's position in the company's equity structure, and the strategic freedom endowed by this position that does not need to continuously compromise with external capital.

This article is from the WeChat official account "Tech Business", author: Tech Business, published with authorization from 36Kr.