Vertical Applications, Capital Market Balance, and "No Role Model"
Discussion on the priority of Coding Agent, the balance between capital markets and pure research, and how DeepSeek's organizational form has no role model and is self-exploratory.
First, we have no role model. Every step is based on our actual situation, seeking truth from facts.
I haven't fully figured out what our domestic business model will look like, or what the smoothest path will be. We're not at that stage yet. The situation domestically may differ from abroad. It's still hard to tell what it will be like domestically.
Given the current situation domestically, I think the most sensible approach is to go all in on general-purpose Agent, and prioritize other Agents lower, including those for finance, healthcare, etc. We need to do Coding first, because Coding Agent can accomplish a lot, and there are many vertical Agents. At this stage, we believe the most important is still Coding Agent.
That's very clear, thank you. Another question we'd like to ask. We really admire how you've been running DeepSeek with a very pure research approach.
But this industry has indeed moved toward the capital market, toward capitalization. And you're a very responsible person, both to your team and to investors. So between pure research, purely pursuing AGI, and the capital market, you'll inevitably need to go to the capital market, with public shares. How do you balance that in the future? How are you thinking about it?
I think we can achieve both. If this year we can have a few hundred million USD in B-side revenue, plus our C-side users, that already provides some commercial foundation. If next year our B-side revenue grows, the company could be close to net profit, maybe even already profitable.
It might no longer be a phase of pure cash burn. So I feel we'll have considerable room to maneuver. Or in the worst case, just selling API could sustain a listed company. If technology doesn't advance further and freezes here, we could just focus on selling API and delivering good service, which would be enough.
So I'm confident; it's really not that difficult. Because we're in a high-leverage area and a fast-moving field, it might just not be that hard. We hope for bigger dreams, but we also have baseline performance to show.
Thank you, that was well said. I'm done with my questions, thank you.
Thank you for sharing. Touching on some earlier points, I'd like to ask a follow-up. I think the biggest difference between DeepSeek and other companies is that our organization is different, or our organizational form is different.
But organizational form is tied to our goals, and we also need to consider its boundaries and efficiency. From a macro perspective, is there a good role model for our organizational form? Historically, maybe Bell Labs? Or some other ideal form? Or do we feel there's no ideal, and we have to explore gradually?
Maybe in this new era, we can only explore on our own. Because our organizational form is definitely different from the American companies, right? The three US companies are different from each other, but at least they all start from a commercial perspective. I mainly want to ask about organization.
First, we have no role model. Every step is based on our actual situation, seeking truth from facts, making decisions based on reality, and figuring out what we should do. So it's a product of the times, or a reflection of reality, not a result of imitation.
In this context, the optimal solution probably is this, or at least, the path we chose is this. Every step we've thought through, we've made choices. The result of those choices is what we have. It's not because we saw someone else choose this, but because we analyzed the pros and cons.
Going forward, it will be the same-we won't imitate anyone. I think we're different from Bell Labs, because Bell Labs didn't need commercialization-it was... but we do need commercialization. We ultimately have to survive; we are a company, and the government won't give us a cent.
So we can have a very grand mission, but at the end of the day, we are a company, and we need to consider how to stay alive. So B-side is definitely important to us, because we might rely on it to survive later. It's just not important now, because it's a cost center.
So I think it's different, different from Bell Labs. We are essentially a company. Historically, many companies have had pursuits beyond profit, but that doesn't mean they aren't companies.
Many great companies have a pursuit beyond profit. That pursuit, far from hindering commercialization, often enhances it. We are essentially a company; we just think about which money to earn, when to earn it, how much, and how. We make trade-offs.
Mr. Liang, two quick questions. First, you just mentioned MILOS might not be a definite goal, but you'll move in that direction. And you also mentioned activation parameters might, for example... The next generation might be around 150 to 250 billion parameters. In that case, do you think 150 to 250 B is comparable to O4.7, or could it be comparable to something else? That's the first question.