DeepSeek Liang Wenfeng Four Hour Investor Meeting Transcript

Editor's note: This DeepSeek Liang Wenfeng investor meeting transcript is translated from the user-provided PDF. The source describes itself as a compilation of 52 remarks collected from multiple channels and states that some wording may differ from the original while preserving the intended meaning. EasyGlobe has not independently verified the meeting or each quotation. For clarity, this DeepSeek Liang Wenfeng investor meeting transcript preserves all 52 items and their section order.

Last month, "elsewhere" reported on DeepSeek's financing story. The part that drew the most discussion was the legendary four-hour investor meeting.

Over the following month, various remarks attributed to Liang Wenfeng circulated widely. We also collected some of that material through multiple channels.

During the meeting, Liang said "no" many times: not geniuses, no unreasonable profits, no pursuit of user volume, no closed source, no 3D or video generation, no world models, and no attempt to build the next super app. He said restraint is a strategy used to increase the probability of achieving AGI.

Within the limited material available to us, the most frequent terms included models, cost, AGI, time, and open source.

Most of the time, Liang spoke with qualification and used plain, unadorned language. Only when discussing a few issues he deeply cared about did he sound sharper: "As long as I can keep the team stable, I will definitely achieve AGI. It is that simple."

Below are the 52 remarks we collected. Some wording may differ slightly from the original while preserving the intended meaning.

DeepSeek Has Only One Main Line

1. This is not the time to maximize returns from products. Products are a step on the road to AGI, but we do not need to spend too much thought or energy building consumer or enterprise products. When you stand at a high point in one technology and apply it to a lower-level technology, you have an overwhelming dimensional advantage. Products are a by-product on the road to AGI.
2. Many things are not on our main line, such as 3D and video generation. World models are another example; they do not have much to do with the upper limit of intelligence.
3. Multimodality is important for products and for consumer users. But it is only a component, not the main line or intelligence itself.
4. There are certainly ways to solve hallucinations in large models, but it is a long-term problem. Internally, we classify hallucination as a product problem. We will solve it, but it is not the priority.
5. At this stage, the most important thing is still the coding agent. Given the situation in China, the most sensible approach should be to go all in on a general-purpose agent. Other agents for finance, healthcare, and similar fields have lower priority.
6. If the AI era produces many trillion-dollar companies, it would be good enough for DeepSeek to be one of them.

First Continuous Learning, Then AI Self-Iteration, and Finally Embodied Intelligence

1. AI does not currently lack taste or intuition. What it lacks is the ability to learn continuously.
2. Humans can keep learning, but for the same task, AI has to be given all the context. That is almost impossible, so AI cannot replace employees. The next generation of models must have continuous-learning ability before it can truly be called the next generation.
3. We hope the next-generation model can help with our own development. Put simply, the first objective of the model we build is not that everyone else finds it useful, but that we ourselves find it useful. This is the fastest way to achieve AGI.
4. No one in the world has yet found a good method, because "learning" is made up of many different things.
5. DeepSeek's long-term vision is AGI. If the route to achieving it is like climbing stairs, last year's step was CoT (chain of thought), and this year's step is the agent. The problem to solve after agents is continuous learning.
6. Once continuous learning is achieved, we may reach a gradual singularity: models could do everything humans can do, including developing more advanced AI models themselves. In other words, AI could accelerate AI research. Only after completing this step do we reach embodied intelligence.
7. The endpoint of intelligence may always be embodiment. For an ordinary person, the need is not really for a computer, but for labor.

A Full Turn Toward Commercialization Is Still Far Away

1. We only seek a reasonable profit. Our pricing is not designed to maximize profit.
2. We initially worried that demand for one of our models would be too high, so we set a relatively high price. Later we cut it to one quarter of that level, and many people in the company chat cheered. That is why we put so much care into making the model good: so everyone can use it fully.
3. Low cost is an outcome. Our model architecture has always moved toward lower cost. We also want the cost to be affordable, especially when compute is scarce. There is another reason: the lower the cost, the larger the model you can support. When compute is limited, greater computational efficiency lets you train a larger model. Large companies can solve the problem by adding resources; we prioritize cost efficiency.
4. From the outside, it may look as though we chose a very difficult model. In reality, we operate it quite easily. Price cuts are certainly not good news for our competitors, and they are definitely not cheering. I do not find the business of selling APIs that attractive. I only need a few people to maintain the API. We do not even have customer service or a sales team; users come on their own.
5. We have always been commercializing, but commercialization is not the goal. The point when DeepSeek fully turns toward commercialization should still be very far away.
6. I do not even need to think about securing a place in it when the time comes. If the commercial opportunity is that large, there will certainly be a way. DeepSeek is a product of its era and a response to real conditions, not the result of imitation.

Open Source Is the Sweet Spot for a Company of Our Size

1. Restraint is a strategy: give up some things in exchange for more of other things. Open source means giving up some benefit. Internally, employees gain a sense of achievement and the company gains cohesion. It also benefits society; peers and ordinary people are happy. I have no doubt that AGI will create enormous commercial value. Given that, my priority is not taking a larger share, but increasing the probability that we succeed.
2. Open source helps if you want to make AI commercially successful. That sounds somewhat counterintuitive. Historically, a software company's market might be worth only several billion dollars a year, so open-sourcing the product could eliminate the opportunity. But AI is large enough that it may ultimately account for ten percent of human society's GDP. If we tried to monopolize that benefit, history would cast us aside. That is an objective rule and a historical view.
3. The open-source model we provide is the same as the model we deploy ourselves. We will not open-source an inferior model while deploying a better one internally.
4. I am not worried that others will deploy our model and compete with us. Not every company has the willingness and ability to reach this goal. If a startup is too small, it lacks the strength to do it; large companies find it difficult to organize. This is a sweet spot for a company of our size.
5. Open source has no effect on our business model, provided the premise is "earning only a certain level of profit." If you want to earn one hundred times the profit, then open source will indeed affect you.
6. We do not want to become an opponent of any large or small internet company. Under that premise, we are very willing to assist anyone and help them improve, including Alibaba, Zhipu AI, and Moonshot AI.

The Gap Is Not in Talent

1. In the future, we need to rewrite the AI narrative: use a fraction of the compute to shorten the gap, reducing it to six months or three months.
2. We believe in scaling; the larger the scale, the better the result. We train models this large not because I think this size is enough, but because these are all the resources we have.
3. There is almost no talent gap. It is the same pool of people. China does not lack talent. Talent shortages are temporary; historically, no category of people has ever remained in shortage forever.

In Model-Lab Competition, Cost Comes First

1. Anthropic's current lead over OpenAI is not long-term; it is temporary. OpenAI and Google will probably take turns moving ahead in the future.
2. There are too many model companies in China. Each one is doing the same thing, so resources are fragmented. The market will eventually converge, but the process takes time. If every company seeks only reasonable profit, the industry does not need so many groups building large models. Two large companies and two small companies might be enough.
3. I absolutely do not believe large-model companies can take most of the profit in the AI industry.
4. The ultimate differences in large-model competition will appear in three areas: cost, time, and user experience. Cost ranks first: at what cost can you provide a service of the same quality? Time comes second; being a few months early or late makes a difference. User experience can create some stickiness and barriers, but it is not fundamental.

No Intention of Becoming the Next Super App

1. We do not want to build the next super app. Become the next ByteDance? The next Tencent? We have absolutely no such idea.
2. We do not compete for that because there are watermelons farther ahead, while what is in front may all be sesame seeds. The sesame seed may be relatively large, but I still do not consider it large.
3. Last year everyone competed for chatbots and consumer traffic. This year they are competing for B2B revenue. But we do not think that is important. What the company truly cares about is the AGI roadmap and how to achieve the next technical breakthrough. It is strange: the thing you most want is hard to obtain, while what you care less about often comes more easily.
4. Becoming popular during last year's Spring Festival was not in our script.

Keeping the Team Stable Is the Core

1. There is only one thing on which we cannot compromise: we must maintain the team's stability. That is also a very large risk we face. Of course, this financing round has substantially reduced that risk.
2. Many of the things we do are intended to keep the team stable. We do not want to become an opponent of any large or small internet company. We hope to empower and help them. We do not want to create enemies. That also makes our own environment better.
3. Some people think our organization works from the top down, while others think it works from the bottom up. I think both are right. From the top down, there is "doing the proper work." We generally hope the "proper work" does not take more than half of an employee's time. The other half is bottom-up and unassigned. People can research whatever they want, explore on their own, and pursue whatever they consider important, with no prerequisites.
4. We generally do not work much overtime. The first reason is that research needs a relatively relaxed environment. The second is that we are highly focused. Many of our products are incomplete, but we have not gone back to fill every gap. That is also part of a culture of restraint.
5. The organization is dynamic, not fixed. As the company grows, there may be adjustments in the future. It will not completely become a traditional hierarchy, though some necessary structure may appear. The fact that we are vision-driven will not change.

Acting with Goodwill Toward the World

1. When we started this company, the original intention was not to make a great deal of money or list on the capital markets. The first few dozen people did not think that way at all. If someone had, they would not have joined. We are doing this with tremendous goodwill toward the world because we believe it is useful to humanity.
2. "Achieve XX KPI" is not our way. We are a vision-driven organization. That has advantages and disadvantages. In the future, we will find ways to build on the strengths and avoid the weaknesses, but this is our distinguishing characteristic.
3. The vision is not even written down. It exists in how we do things and in our attitude toward the world. Everyone in the company may understand the vision differently, but we agree on the broad direction.
4. About twenty years ago, the manager I admired most was Jack Welch, the former CEO of GE. Looking back now, most of what he said may no longer be right, but he was right about one thing: the most important thing for a company is its vision. A vision is not a slogan hanging on the wall. It is not what you say, but what you do.

Restraint Makes Us More Likely to Achieve AGI

1. AGI offers the greatest return. As for other things, we will do them if we have the energy; if we do not, we will not. Restraint is part of our vision.
2. AI is too large, and the benefits are also too large. If you can make it work, even a small share of the benefit will be enormous. The more restrained you are, the more likely you are to make it work.
3. I think that is intuitive, at least according to my intuition. Apart from our vision, we do not have many other advantages.
4. When we founded this company two years ago, we did not have much money, many chips, recognition, or drawing power. We were simply a group of very ordinary people. The story I like is "a group of ordinary people did something extraordinary," not "a group of geniuses did something extraordinary."
5. Open source is also part of restraint. In pricing, we certainly do not begin from the goal of maximizing company revenue or profit. In the short term, a higher price brings more revenue, but in the long term, it is hard to say. To me, restraint is a strategy.
6. Open source and low prices give employees a sense of achievement, which creates organizational cohesion. They benefit society, and peers and ordinary people are happy. From a long-term perspective, this restraint can increase the probability that we achieve AGI.
7. If your vision is to take more, you have already lost. You may face even greater difficulties. That is how the world works.

Related Reading

• A Practical Guide to AI Agent Skills
• How People Use AI and What It Means for the Future of SaaS
• OpenClaw and the CLI-Driven Agent-Native Era