Liang Wenfeng says DeepSeek is chasing AGI, not commercialization, in a 4-hour investor meeting
创业邦 · wechat · 2026-07-23
What Liang Wenfeng said in a 4-hour DeepSeek investor meeting
A WeChat deep-dive reconstructs the key points from Liang Wenfeng’s roughly four-hour investor meeting, focusing on DeepSeek’s strategy, organization, compute constraints, and roadmap toward AGI.
- Vision over process: Liang framed DeepSeek as a mission-driven company with no KPI-heavy management, no rigid organization, and a shared belief that the work is meant to be useful to humanity rather than purely money-driven.
- Pricing and commercialization: He said API pricing is set around a “reasonable profit” target—roughly enough to recover hardware costs in about ten months—and described commercialization as a byproduct of pursuing AGI rather than the primary goal.
- AGI roadmap: Liang described AI progress as a staircase: CoT was last year’s step, agents are this year’s step, and the next major hurdle is continuous learning. Only after that, he said, would self-iteration and embodied intelligence become feasible.
- Team and organization: DeepSeek’s biggest core interest, in his view, is keeping the team stable. He said talent retention, shared belief, and organizational cohesion matter more than money or formal structure.
- Compute and China-U.S. gap: He argued the biggest gap with the U.S. is compute resources, not talent. DeepSeek reportedly has around 20,000 H-series equivalent compute, while frontier U.S. efforts are training on much larger clusters.
- Domestic chips: Liang was optimistic about国产算力, saying the window for Chinese chip adoption is opening because CUDA’s moat is weakening and supply constraints are forcing the ecosystem to adapt.
- Efficiency and scaling: He said DeepSeek believes in scaling and wants lower model costs, but is constrained by resources. In his view, training size is determined by available compute rather than a fixed “enough” threshold.
The article’s main takeaway is that DeepSeek sees itself less as a product company and more as an AGI-oriented research organization with unusually strict discipline about focus, cost, and team stability.
Related event: Liang Wenfeng's 4-Hour Investor Meeting: DeepSeek's Sole Focus is AGI(10 posts)→
More from Companies & People
- Tesla says Robotaxi unsupervised miles are growing over 10% week over week — davidpattersonx · 2026-07-23
- A surreal AI startup drama thread about public beef and falling apart — DionysianAgent · 2026-07-23
- OpenAI seeks AI safety researcher for real-world misalignment investigations — tomekkorbak · 2026-07-23
- Liang Wenfeng says 16,000 Huawei 950 GPUs are only enough for current-gen training — teortaxesTex · 2026-07-23
- Crawlravenseo pitches a way to turn old search data into AI-era opportunities — ayushtweetshere · 2026-07-23
- A YouTube download Skill wraps yt-dlp and adds fixes for the usual failure modes — vista8 · 2026-07-23