Four logical judgments behind the financing rumor of DeepSeek with a valuation of tens of billions of dollars

The biggest news in China's AI circle this weekend is that DeepSeek is rumored to be releasing around 3% of its equity for financing at a valuation of $10 billion. For a company that has long adhered to "self - financing" and whose founder, Liang Wenfeng, directly and indirectly holds 84.29% of the shares and has almost 100% voting rights, this news alone is enough to trigger intense industry discussions.

However, it's worth noting that just two days after the news started to spread, information from various channels is highly consistent: A person from a large state - owned equity institution said the news is "very likely to be true," but "it's currently impossible to invest." Many venture capitalists also admitted that for popular projects like DeepSeek, financing shares usually need to be "grabbed." In other words, even if the financing news is true, it's highly likely that only a few external institutions will be able to get a share.

Regarding this rumored event, we present four levels of logical judgments and break them down layer by layer as follows.

First - level logic: Essentially, it's the structural design of equity incentives for a non - listed company

DeepSeek's uniqueness lies in the fact that since its incubation by Magic Square Quant in 2023, it has never received any external equity financing. This fact brings about a structural problem that is easily overlooked: The stock options held by employees lack a market - based pricing anchor.

An investor who has invested in large models accurately analyzed to Yicai Global: Even if DeepSeek opens up for financing, it's not a game for most people. And according to Liang Wenfeng's ideas, the terms will definitely be extremely strict. Regarding this shift in financing, the investor believes that it's most likely for the pricing and redemption of employee stock options, and "it's too late."

The logical deduction is as follows.

In the employee equity incentive system of a non - listed company, the value of stock options needs to be confirmed through external market - based pricing. Without external financing, it means there is no valuation anchor verified by real money. The equity commitments in employees' hands cannot be converted into a clear wealth expectation, lacking sufficient liquidity and premium reference in the eyes of top - tier talents.

The talent competition in the AI field has reached a white - hot stage: Luo Fuli, a key contributor to the DeepSeek - V2 architecture, joined Xiaomi; Guo Daya, the core author of the GRPO algorithm, joined ByteDance; and Ruan Chong, a core multi - modal researcher, joined DeepRoute. The salary packages offered by these competitors can be double or even more than what DeepSeek currently offers.

Introducing a small - scale round of financing essentially uses a market - based transaction price to officially price the stock option pool for all employees. $300 million corresponding to about 3% of the equity is a large enough transaction volume to generate a legally valid and market - referenceable price anchor, but it's not enough to shake Liang Wenfeng's absolute control. From this perspective, the primary function of this round of financing is to "account to the internal," giving clear return expectations to past contributors and clear incentive benchmarks to future talents.

This also explains why the "shares are hard to grab." If the core purpose of the financing is pricing rather than introducing strategic resources, then Liang Wenfeng will tend to choose investment parties with the highest degree of cooperation in terms of terms, the weakest strategic demands, and the lowest willingness to intervene in business decisions. In the eyes of an idealistic founder, the entry of external capital is a necessary compromise, and his natural inclination is to minimize its impact.

But there is a deeper logical consideration here: Is introducing external financing really the only way to solve the problem of stock option pricing?

In fact, the pricing of stock options for non - listed companies does not necessarily depend on equity financing. Under a mature legal and financial framework, the company can completely hire a third - party evaluation institution for an independent valuation, or Magic Square Quant can contribute funds to establish an internal repurchase fund to repurchase employee stock options at a fair value. These methods can also provide a liquidity outlet for stock options without diluting the founder's control at all.

So, why did Liang Wenfeng specifically choose the path of financing? The possible answer lies in the essential difference between "market - based endorsement" and "internal valuation."

No matter how high the price is set for internal repurchase, in essence, it's still the company using its own money to buy its own stocks, lacking the trading behavior of external market entities as support for fair value. In the perception of top - tier talents, the "credibility" of such an arrangement is far lower than that of introducing strategic investors, which means an independent third party has confirmed the market value of the company's equity with real money. In other words, financing is not the "only solution" for stock option pricing, but it is the most credible "optimal solution."

Second - level logic: A valuation of $10 billion is an unreasonably low price, and "outsiders" are unlikely to get a share

If the essence of this round of financing is the structural design of equity incentives, then pricing is the variable most worthy of in - depth study. The figure of $10 billion is unreasonably low in the current AI valuation coordinate system.

First, let's look at the horizontal comparison. In January 2026, Zhipu AI was listed on the Hong Kong Stock Exchange, with a market value of about $6.8 billion on the first day and a latest market value of about $50.7 billion. MiniMax had a market value of about $13.7 billion on its first day of listing and a latest market value of about $34.4 billion. As another large - model unicorn that is yet to be listed, the valuation of Dark Side of the Moon has risen from $4 billion in November 2025 to $18 billion.

Now, let's look at the vertical logic. Magic Square Quant, the parent company behind DeepSeek, had an average annual return rate of up to 56.6% in 2025, with a management scale of over 70 billion yuan. It ranked second in the performance list of hundred - billion - level quantitative private funds. Roughly estimated according to the industry's common practice of "1% management fee + 20% performance fee," in 2025 alone, Magic Square Quant brought Liang Wenfeng an income of about 5 billion yuan, equivalent to over $700 million.

What's even more thought - provoking is the "earnings - based valuation": Magic Square Quant is essentially a financial machine with stable profitability. If there is a clear value connection between DeepSeek and Magic Square Quant, whether it's a capital channel or technological synergy, then a valuation of $10 billion corresponds to a price - to - earnings ratio of just over ten times. For a complex that has both top - tier AI R & D capabilities and top - tier quantitative trading capabilities, this pricing is difficult to justify under any reasonable financial model.

In both the horizontal and vertical dimensions, a valuation of $10 billion is significantly lower than the market reference. This naturally leads to a deeper question: Why is Liang Wenfeng willing to introduce external capital at such a low price?

A reasonable explanation is that the low valuation itself is a screening mechanism. In a financing where "outsiders can't get a share," the pricing level is not the founder's primary consideration. On the contrary, an obviously low valuation can effectively filter out those investment institutions with strict requirements for financial returns and strong bargaining intentions, and screen out partners who truly accept the game rules set by Liang Wenfeng. In other words, this price is not the result of market game, but an entry threshold actively set by the founder.

However, this explanation can only answer "why it's low," not "why it's this specific number."

So, what is the real anchor for the figure of $10 billion? The answer is likely to be found in Magic Square Quant's accounts.

Since its incubation in 2023, all R & D investments, computing power purchases, and team salaries of DeepSeek have been borne by Magic Square Quant. This is an internal transfer cost that can be accurately calculated. According to industry - cross - verified data, Magic Square Quant's cumulative investment in DeepSeek over the past three years is approximately in the hundreds of millions of dollars.

Releasing 3% of the equity at a valuation of $10 billion exactly means that: The $300 million raised in this round is approximately equal to Magic Square Quant's total investment in DeepSeek over the past three years.

We can regard this as a very precise financial signal: If $300 million corresponds to Magic Square Quant's cumulative investment in DeepSeek over the past three years, then after the completion of this round of financing, DeepSeek will be officially financially independent from Magic Square Quant. This means that in the future, DeepSeek's continuous losses will no longer be covered by Magic Square's profits, and it will need to face the capital market on its own. This round of financing will be the starting point for its independent operation.

Third - level logic: Dimension - reduction anchoring, using equity swaps to lock in structural advantages

The above two levels of logic explain "what this round of financing is" and "why the pricing is like this," but they don't fully answer the question of "what the money will be used for." A financing scale of $300 million is a drop in the bucket in the current AI computing power competition.

Let's do a simple calculation. In March 2026, OpenAI completed a $122 billion financing, with a post - investment valuation of $852 billion. Anthropic completed a $30 billion Series G financing in February this year, with a post - investment valuation of $380 billion. (It should be noted that although Anthropic has grown rapidly recently, with its annualized revenue exceeding $30 billion and surpassing OpenAI, and it has received a valuation offer of about $800 billion from investors, its post - investment valuation in the previous official financing round was still $380 billion, not exceeding OpenAI's current valuation of $852 billion.)

Regardless of which set of data is used as a reference, the single - round financing scale of leading players is often in the tens of billions or even hundreds of billions of dollars, while DeepSeek's $300 million financing is not even enough to purchase a medium - sized ten - thousand - GPU cluster. Not to mention that the upcoming DeepSeek V4 will have a total parameter count of up to one trillion, facing exponentially increasing computing power and power consumption demands in the Agent era.

Large - model training follows the Scaling Law. Performance improvement requires exponential computing power input, and in the operating costs of AI large models, power costs account for 60% to 70%. In this structure, tokens can be regarded as a kind of "power derivative" to some extent. With the release of V4 and the opening of Agent capabilities, DeepSeek will face an exponential increase in call volume, which will lead to a simultaneous surge in power costs.

This leads to an inference: The $300 million cash financing is a drop in the bucket for computing power purchases. However, if a part of it is used to lock in partners in power infrastructure through equity swaps, for example, exchanging partial equity for long - term low - price power supply agreements with power companies or data center operators, the strategic value of this transaction will be completely different. The power cost in China is less than one - fifth of that in the United States. If this comparative advantage can be locked in and magnified by DeepSeek through equity ties, it will be a far more important infrastructure layout than the financing amount itself.

Looking further, power may just be an entry point. This logic can be extended to a general model of "dimension - reduction anchoring": After the large - model competition enters the agent era, the competition dimension is expanding from the model ability itself to the infrastructure level.

DeepSeek can use its own equity as a "high - dimension currency" to anchor any "low - dimension node" with structural cost advantages in the industrial chain. Power is just the most prominent one, and potential targets also include domestic chip production capacity, data center cabinet resources, cross - border network bandwidth, etc. The essence of equity financing is redefined here: It is no longer just exchanging equity for cash, but exchanging equity for structural barriers.

Frankly speaking, this part is a speculative conjecture lacking solid information support. Among all publicly available reports, there is no information indicating that DeepSeek will use the financing for power infrastructure swaps. Linking power costs with the equity structure has not become a precedent in the AI industry. Therefore, this judgment is more like a logical possibility deduction rather than a factual assertion.

Fourth - level logic: Signal hedging, the art of balance between certain narratives and uncertain realities

Let's return to the most fundamental question: Why did Liang Wenfeng choose to finance at this time?

A dimension ignored by most analyses is "signal hedging." The multiple delays of DeepSeek V4 have led to the accumulation of negative expectations in public opinion. It has been 15 months since the release of R1. During this period, competitors have iterated multiple times, and Doubao, with a monthly active user count of over 331 million, firmly holds the top position in the domestic AI application market. The release of V4 has been postponed from February this year to March and now to late April according to the latest news. Each delay erodes the market's certain narrative about DeepSeek being "always ahead."

In this context, launching the first - round financing is a powerful hedging signal. Its subtext is: We are evolving from a pure research institution into a commercial company with a capital governance structure. This is not because of technological bottlenecks, but because the organization needs to enter the next stage.

Using the financing narrative to hedge against the product delay narrative and using the certainty of "organizational evolution" to hedge against the uncertainty of "technological rhythm," this layer of signal value may be far more strategically significant than the $300 million in cash.

This also explains why the financing news was released at such a low valuation. If Liang Wenfeng's goal was just to raise money, he had every reason to wait until after the release of V4 and the restoration of market confidence before setting the price. But the value of the "signal" lies in its pre - emptiveness. Releasing a positive structural signal when market expectations are at their most fragile is far more powerful than adding icing on the cake when market confidence is high.

Conclusion

Based on the four levels of logic, the picture of DeepSeek's current round of financing is gradually becoming clear:

It is a highly restrained equity structure design: Using small - scale equity transactions to complete market - based pricing for employee stock options; screening investment parties with high cooperation degrees through an obviously low valuation; using equity as a "high - dimension currency" for dimension - reduction anchoring at the infrastructure level; and using the signal of "organizational evolution" to hedge against the negative narrative of product delays.

These four levels of logic all point to one conclusion: The rules of this round of financing are entirely set by the founder, and the role of "outsiders" has been carefully restricted from the beginning. For investors who rushed to book flights to Hangzhou over the weekend, the real test is not whether they can meet Liang Wenfeng, but whether they are willing to accept a set of game rules completely defined by the other party.

This article is written based on publicly available information and is only for information exchange purposes. It does not constitute any investment advice.

This article is from the WeChat official account "Jinduan" (ID: jinduan006), author: Yuantai, published by 36Kr with authorization.