DeepSeek V4.1下月发布500亿融资抢先到位 (English)

Generated: 2026-06-22 11:03:17

---

Okay, no problem. As an editor, I'll carefully fact-check, correct data, and refine the language style to make the article more solid and the rhythm more natural.

Here is the revised final version:

---

Final Version

The Man Who Said "Never Raise Money" Put Up 20 Billion Himself

On the evening of May 8th, I was scrolling through my phone when two pieces of news exploded at the same time.

One was that Alibaba and DeepSeek had called off the talks. The other was even bigger—Liang Wenfeng was putting up 20 billion of his own money.

My first thought was: Wait, didn't he say he would never raise money?

In July 2023, he did say exactly that. But by April 2026, he had changed his mind. From "no funding" to raising 50 billion, the three years in between were also filled with wave after wave of talent raids.

To be honest, I felt a bit like it was a slap in the face at first. But when you lay the whole thing out and see every move clearly, you might just want to stand up and applaud.

This isn't surrender. This is being a grown-up.

---

Why Alibaba Was Kicked Out

Let's start with the juiciest part.

Alibaba, with that massive size, got rejected.

Why?

One sentence: Liang Wenfeng wants control, Alibaba wants ecosystem integration. These two things are inherently at odds.

What's Alibaba's usual investment playbook? Invest in you, then tie you into its ecosystem. Cloud, chips, enterprise clients—all linked together. Over the past decade, Alibaba has invested in so many companies, and it's pretty much been this way.

But Liang Wenfeng doesn't buy it.

DeepSeek has been independent from day one. Open source, low cost, no taking sides. If it tied itself to Alibaba Cloud, would Tencent Cloud users still use them? What about Huawei Ascend compatibility? That would be cutting off its own arms.

One insider said something that really stuck with me: What Liang Wenfeng values most isn't the amount, but the offer with the fewest strings attached.

A few years ago, when AI companies were rushing to pick sides among the BATs, which founder wasn't begging for capital? Meeting investors, asking "Who should I choose?" with anxiety written all over their faces. Now? Liang Wenfeng just says, "If it's not a good fit, I don't want it," and turns away Alibaba's money.

Times have truly changed.

---

National Team Enters the Game – Do You See the Move?

The biggest variable in this funding round is the National Big Fund.

You read that right. Previously, the Big Fund only invested in integrated circuits; it had never touched AI large model companies. This time, they made an exception.

The signal is clear: Shifting from "filling hardware gaps" to "building a coordinated software-hardware ecosystem."

What does state investment mean for DeepSeek?

First, the terms are loose; no requirement for ecosystem binding. Second, its strategic position is directly elevated, becoming "AI infrastructure." Third, the door to the government and enterprise market suddenly swings wide open.

With state funds in hand, DeepSeek becomes eligible to bid on projects like smart city initiatives by local governments and digital transformation of state-owned enterprises. These are resources that purely market-driven VCs simply can't provide.

This move has paved all the backup paths perfectly.

---

Why Is Liang Wenfeng Putting Up 20 Billion Himself?

This is the part I most want to dissect.

The total funding for this round exceeds 50 billion, with Liang Wenfeng personally contributing 20 billion, accounting for 40%. In the history of global AI financing, almost no founder has ever done this.

On the surface, it doesn't make sense. Funding is about spending other people's money, so why are you chipping in?

I went through dozens of reports, sorted out the logic, and boiled it down to three reasons.

First, control.

Look, before this funding round, Liang Wenfeng's stake was only about 1%, with most held by High-Flyer Quant. On April 27, 2026, DeepSeek's registered capital increased from 10 million to 15 million, and the entire increase was subscribed by Liang Wenfeng personally. His direct stake immediately jumped to 34%. Combined with High-Flyer's indirect stake, he controls about 84.29% of equity and nearly 100% of voting rights.

Think about it: taking 50 billion in external investment while maintaining absolute control. In a financing case of this magnitude, it's almost impossible.

Liang Wenfeng used 20 billion not for the money, but to ensure decision-making power doesn't slip away.

Second, giving employees peace of mind.

This is the most practical reason.

DeepSeek hadn't raised money in three years, so employee stock options had no market pricing. Stock options, at the end of the day, are like an empty check from the company saying "might be worth something in the future," but no one knows the current value.

Here's the problem. When competitors poach talent, they offer "cash + immediately vesting stock." What terms did ByteDance, Tencent, and Alibaba offer? Two to three times the salary, four-year total packages worth hundreds of millions. How could DeepSeek's core researchers resist?

Over the past six months, at least five core R&D backbone members have left. Guo Daya, a core researcher for R1, went to ByteDance. Wang Bingxuan, lead author of the first generation large model, went to Tencent. Ruan Chong, head of multimodal, also left.

Talent drain was the most critical trigger for Liang Wenfeng changing his funding strategy.

Now? With the 40 billion valuation set, a 5% employee option pool corresponds to 20 billion. Everyone can immediately calculate the value of the options in their hands. It transforms from "an empty promise" into "visible wealth." And everyone is tied to the same boat—if the company's valuation rises, everyone wins; if core talent gets poached and the valuation plummets, everyone's money shrinks.

Liang Wenfeng putting up 20 billion of his own money at the same price per share is him sending a message to the employees: I'm in this together with you; I'm not just cashing out.

Third, preventing reverse talent poaching by investors.

DeepSeek laid down a hard rule at investor meetings. You can invest, you can earn dividends. But there are two red lines that no one can cross—don't poach our employees, and don't encourage key members to leave and start competing ventures.

Think about it: Tencent, JD.com, NetEase are all working on general-purpose large models. CATL is doing in-vehicle AI. They are all potential competitors for DeepSeek. Without this red line, investment would just be "using investment to steal technology and people."

Liang Wenfeng shut this path down. Ruthless. Truly ruthless.

---

Cutting Prices Right After Funding – Such a Brilliant Move

The week of June 16th, when DeepSeek's first funding round hit their account, they announced a permanent price drop for V4-Pro.

How much of a drop?

Cache hit input: 0.025 yuan per million tokens.

What does that mean? The industry usually talks in terms of "cents per level"—this was unprecedented. GPT-5.5 output price is about 216 yuan per million tokens. DeepSeek is 36 times cheaper. Tongyi Qianwen Qwen3.5-Plus output is 1.6 yuan. Baidu's ERNIE flagship output is 20 yuan. Zhipu GLM-5 input is 12 yuan.

DeepSeek's cache hit price: 0.025 yuan.

And this price cut happened precisely when other domestic large model companies were raising prices collectively. ByteDance's Doubao launched paid subscriptions. Zhipu GLM raised API prices three times. Alibaba Cloud and Tencent Cloud API prices went up by 5%.

Only DeepSeek went in the opposite direction.

This isn't just "competing on price." Liang Wenfeng is using price to lock in the developer entry point. Cut prices right after funding, making it clear to the market: "I have money and can keep subsidizing."

The ones feeling the squeeze aren't just competitors; it's ByteDance, Alibaba, Tencent.

I did the math: 50 billion in funding plus the national team's backing plus permanent price cuts