Take the Money, Leave the KPIs: Liang Wenfeng Is Rewriting the Investor Contract
After reading the transcript of Liang Wenfeng’s investor meeting, my first reaction was blunt:
Is he trying to rationalize DeepSeek’s commercialization strategy?
On one hand, he has taken investors’ money. On the other, he repeatedly told them that when the company was founded, it never started by asking how much it would eventually make or whether it would go public. DeepSeek was not being built around maximizing commercial returns. The company had no KPIs, only a vision.
In plain English, the message was roughly this:
You can give me money, but do not pressure me with revenue, profit, or the speed of commercialization. I have taken your money, but I am not going to rush to become the next ByteDance or Tencent. I am going after AGI. If you want to invest, you should accept that premise first.
One investor even volunteered that the round represented recognition, support, and respect for DeepSeek’s mission and vision.
Well, then.
Investors provide the money; Liang Wenfeng pursues the dream. As for when the company will make money, how much it will make, and how it will make it, investors should not rush.
Isn’t this just pushing commercialization expectations down to the floor?
But I kept reading. By the end of the final hour, I realized that this judgment was only half right.
Liang Wenfeng is indeed lowering investors’ expectations, but not about whether the company should make money. What he really wants to take away is the right to define the pace at which it makes money.
That is where things get interesting.
He Really Does Not Want Commercialization to Direct the Technical Roadmap
In the transcript, Liang Wenfeng was unusually direct about profit maximization.
DeepSeek’s API could originally have been sold at a much higher price. By his account, even if the price doubled, usage might not change very much, so company revenue could rise almost in step. But they did not do it. When the team saw prices fall, people in the company group chat actually celebrated because the models had become cheaper and more people could afford to use them.
For a normal commercial company, that scene would be rather strange.
When most companies cut prices, the finance department first calculates the impact on ARR, the sales team calculates the impact on commissions, and management holds three more meetings to decide how to balance “user value” and “shareholder value.”
DeepSeek’s response was: ARR may fall, but everyone is happy.
Liang Wenfeng calls this kind of choice restraint. His logic is that earning a little more in the short term does not necessarily increase the probability of building AGI. Lower prices, open source, and sharing more value may instead strengthen team cohesion and win broader ecosystem support.
He also said that during the past three years, spending time on advertising, e-commerce, product lines, and commercialization paths would have been a waste. Technology was changing too quickly. A business model carefully designed today might lose its underlying premise six months later.
This is not a polished summary we invented for him. He said it clearly himself:
We have always been commercializing; commercialization simply has not been the goal.
That sentence matters.
DeepSeek has consumer users and business revenue, but in Liang Wenfeng’s narrative those businesses look more like byproducts that naturally fall out of the road to AGI. If something can be sold, sell it. If it can generate revenue, let it generate revenue. But do not let a successful byproduct make the whole company turn around and start optimizing for the byproduct.
He did not even want to chase users who were already within reach. When user numbers suddenly surged, DeepSeek did not rush to build a super app or force the traffic into the advertising, e-commerce, and local-services monetization machines that internet companies have tested countless times.
Liang Wenfeng’s analogy was direct: there is a watermelon farther ahead; what is in front of you may only be sesame seeds.
From his point of view, the logic is coherent. If AGI succeeds, its commercial value may be too large for us to estimate today. If that is the case, spending scarce research attention on a little more API profit, a larger consumer audience, or several product lines with short life cycles may be counterproductive.
So does he not want commercialization?
No, no, no.
He simply refuses to let current commercialization turn around and dictate the long-term technical roadmap.
AGI Is the Dream; the API Business Is the Floor
Near the end of the meeting, an investor asked the question nobody can avoid:
DeepSeek may enter the capital markets and eventually face public shareholders. How should a company balance pure AGI research with the capital markets?
It was a good question. Capital can believe in a vision, tolerate short-term losses, and accept risk. But it is hard to hear “trust me; once AGI works, everything will be fine” forever.
Dreams do not go directly onto a financial statement.
Liang Wenfeng’s answer was not “then we should never go public,” nor was it “investors who cannot accept the vision should leave.” He said both sides were necessary.
He believes DeepSeek already has business revenue and consumer users, giving it a basic commercial foundation. If technical progress stopped at some point, the worst-case scenario would be to focus fully on selling APIs and delivering good service. That business alone might be enough to support a public company.
He added:
We hope to pursue a bigger dream, but we also have results we can put on the table as a floor.
On the next page, he made the point even more plainly. DeepSeek definitely needs commercialization. The company has to survive; the government will not give it a penny. B2B business is not the highest priority today, but it may eventually have to keep the company alive.
So Liang Wenfeng was not telling investors, “I do not care about performance.”
The structure he offered was closer to this:
At the top is the enormous return that AGI might create. At the bottom is an API business providing a survival floor.
If I had to describe it as a financial product, investors are buying a long-term AGI call option with the API business as its floor.
If AGI works, everyone shares the enormous upside. If AGI does not work for the time being, DeepSeek can still live on its model capabilities, users, and API revenue, and may still be able to show the capital markets respectable results.
That is not simple idealism.
It is a very sophisticated capital negotiation.
Here, Idealism Is a Financing Term
Many people watching the meeting may describe Liang Wenfeng as a technical idealist who does not care about money and only wants to change the world.
I think that reading is too shallow.
He certainly has an idealistic side. But once a company accepts outside investment, idealism cannot remain merely a personality trait. It has to become a set of transaction terms that capital can understand, evaluate, and ultimately accept.
The terms Liang Wenfeng offered were roughly these:
DeepSeek will commercialize and take responsibility for the company’s survival. If technology stops advancing, the API business provides a floor. But capital cannot see short-term user growth and then demand a super app immediately. It cannot see that APIs could be priced higher and then require the team to maximize profit. And it cannot use the product KPIs of a traditional internet company to schedule an AGI research organization.
Investors receive financial downside protection, but must give up some control over the company’s pace.
Liang Wenfeng receives money, compute, and team stability, while retaining room to decide what matters most right now.
At the meeting, he said this round was carefully selected. The company wanted partners with aligned interests, minimal hostility toward DeepSeek, and a strong desire to see it succeed.
In plain English, DeepSeek was not only choosing who was willing to provide money. It was choosing who was willing to accept this unusual contract.
The money can come in; the capital-market clock stays outside the door for now.
That is what I find impressive about Liang Wenfeng. He did not pretend the company did not need money, and he did not describe commercialization as something dirty. He simply used DeepSeek’s current technical position and market scarcity to secure a right that most founders could never negotiate:
I can prove that the company has the ability to make money, but you cannot therefore require me to make money the highest priority immediately.
That sounds entitled.
The problem is that when a founder truly holds scarce technology, user growth, commercial revenue, and a compelling future, entitlement can become bargaining power.
Capital was not persuaded by his moral appeal.
After calculating the trade-off, capital was willing to pay for that entitlement.
I Admire the Negotiation, and I Am Wary of the Governance
At this point, I should probably start praising Liang Wenfeng.
But honestly, after reading the transcript, I felt less reassured and more cautious.
The reason is also contained in the same document.
Liang Wenfeng said DeepSeek organizes around a vision rather than rules and KPIs. More subtly, that vision has not even been formally written down, and different employees may not understand it in exactly the same way.
That is a romantic description of an organization.
A group of top researchers does not need layers of reporting or quarterly targets. They explore important questions on their own because they believe in the same goal. Formal assignments should ideally not consume more than half an employee’s time; researchers should decide for themselves what to do with the rest.
In AGI research, this environment could indeed produce unconventional innovation. KPIs are good at optimizing things we already know how to do, but they are poor at asking a team to deliver a breakthrough that has not yet been invented.
But no KPIs does not mean no power.
Quite the opposite. When the vision is unwritten and the standards cannot be quantified, the organization’s most important power becomes concentrated in the question of who is qualified to interpret that vision.
What counts as helping AGI? What counts as premature commercialization? What is a reasonable profit? What is team consensus? When has a technical path reached the point where it must be narrowed, and when should the company continue spending on exploration?
None of these answers can be generated automatically by running a benchmark.
In the end, someone still has to decide.
Liang Wenfeng emphasizes that the company relies on consensus rather than his personal decision on everything. I am willing to believe that this is his genuine management philosophy. But when the founder holds extraordinary authority over technical direction, organizational culture, and capital negotiations, the questions remain: how is consensus actually formed? How far can dissent travel? How long can failure continue before adjustment becomes necessary?
Private-market investors can choose to accept this arrangement. They have reviewed the materials, calculated the risks, and chosen to give their money to a strong founder. If the investment fails, that is also a contract signed by professional investors themselves.
If the company eventually enters the public markets, the situation becomes much more complicated.
Public shareholders will not necessarily treat AGI as their highest mission. They may not be willing to trade current profit for an indefinitely scheduled future. At that stage, “we have a bigger dream” can remain part of the company culture, but it will be difficult for it to serve as the entire governance answer.
Capital is willing to hand Liang Wenfeng the clock today because DeepSeek keeps producing results.
But a governance system cannot only work while its founder continues to be right.
The real test always comes when the technical path suffers repeated setbacks, the team disagrees about direction, the financing environment cools, and API growth can no longer cover research spending. Who has the authority to hit the brakes then? Based on what? And what can investors do?
That is why I remain wary of this clever negotiation.
Winning the negotiation does not make governance problems disappear. It merely pushes them into the future.
Ordinary Founders Should Not Copy This
I also want to pour some cold water on the discussion.
I have never conducted a capital negotiation on Liang Wenfeng’s scale, and I will not pretend that I have sat at that table. But public materials do make one thing clear: he can tell investors not to pressure him with commercialization because DeepSeek has already delivered scarce technical results, genuine user demand, and API revenue that can continue to scale.
He did not bring investors a slide deck saying, “Please give me ten years to pursue my dream.”
He had cards in his hand.
This is exactly what many founders learn incorrectly. They hear Liang Wenfeng say no KPIs and no profit maximization, then start talking about long-termism themselves. When nobody uses their product, they say the market is not ready. When revenue fails to grow, they say early commercialization would damage the vision. When investors ask for data, they decide that capital does not understand innovation.
That is not long-term thinking.
It is probably avoidance of validation.
Liang Wenfeng can put commercialization behind AGI because even with commercialization placed second, commercial results continue to appear on their own. Users come voluntarily. Companies call the models voluntarily. The capabilities themselves create revenue.
For an ordinary founder without that kind of technical leverage and market pull, commercialization is never merely a distraction. It is the health check that tells you whether you have created real value.
So the thing founders should learn from this meeting is not how to reject KPIs. It is something much harder:
Can you give investors a credible floor that makes them willing to give you more time?
If you have no API business or other floor, if user growth disappears when subsidies disappear, if team stability depends entirely on the next funding round, and if nobody except the founder can explain the vision, then “mission-aligned capital” may simply be a story that both sides are temporarily unwilling to challenge.
Stories matter, of course.
But only a story that can survive has earned the right to talk about the long term.
Finally
So, returning to the question at the beginning: is Liang Wenfeng rationalizing DeepSeek’s commercialization strategy?
My answer is yes, in part.
He is managing investors’ expectations and explaining in advance why DeepSeek will not maximize revenue or copy the expansion playbook of traditional internet companies. Every founder who receives a large amount of funding has to build a narrative around how the money will be used. Liang Wenfeng is no exception.
But seeing only rationalization would underestimate the substance of the meeting.
He did not evade commercial results. Instead, he offered a fairly clear floor. What he actually accomplished was to separate capital returns from capital control: investors can share in DeepSeek’s enormous upside if it succeeds, but they do not automatically gain the right to define its technical roadmap and commercial pace.
That is the sophisticated part.
It is also the part that deserves caution.
I admire his ability to make capital hand over the clock, but I am wary that he has handed the company’s clock to an AGI vision with no KPIs and few written boundaries.
Capital can believe in a person.
A company ultimately needs to build something that can keep operating even when that person is not always right.
Take the money. Leave the KPIs.
But boundaries will come eventually.
Source note: This article analyzes the transcript of “Liang Wenfeng’s Investor Meeting.” The file states that it was automatically transcribed from audio and then organized with AI; speakers were not identified, and some proper nouns and figures may be inaccurate. The original recording should be treated as authoritative. On July 23, 2026, several media outlets publicly reported on senior-level views contained in the discussion. This article does not use sensitive financing, compute, or employee-option figures from the meeting. Statements about expectation management, capital negotiation, and governance risk are the author’s personal judgments based on publicly available written material and do not represent verified motives of the parties involved.
