The transcript of Liang Wenfeng's investor meeting held on 2026-07-22 reveals DeepSeek's strategic direction amid intensifying AI competition. Liang, founder of the Chinese AI startup, emphasized the company's commitment to open-source large language models while pursuing commercial applications.

During the session, Liang detailed DeepSeek's progress on DeepSeek-V3 and DeepSeek-R1 reasoning models, highlighting performance benchmarks against GPT-4o and Claude 3.5 Sonnet. He addressed investor questions about monetization pathways, noting API revenue growth and enterprise adoption in coding and research workflows.

The discussion covered compute infrastructure challenges, with Liang confirming partnerships for H100 and H200 GPU clusters. He outlined plans for multimodal capabilities and agentic systems, targeting 2025 release milestones. Regulatory compliance in China and global expansion strategies were also discussed.

Investors probed competitive positioning against Alibaba's Qwen, Zhipu AI, and Moonshot AI. Liang stressed differentiation through model efficiency, cost structure, and developer ecosystem investments via DeepSeek-Coder and community initiatives.