DeepSeek, the Chinese AI lab that sent shockwaves through global tech markets when it built a frontier model for a fraction of the typical cost, has reportedly reached an annualized revenue run rate between $400 million and $500 million. That’s roughly double the $200 million to $220 million estimates that circulated in 2025.
From research lab to revenue machine
DeepSeek is backed by High-Flyer, a quantitative hedge fund, and has built its reputation on open-source AI models. The R1 model, which reportedly cost around $6 million to develop in January 2025, triggered what some crypto analysts called an “iPhone moment” for AI agents.
Previous estimates pegged DeepSeek’s daily revenue at roughly $562,000 against daily costs of approximately $87,000, suggesting gross margins north of 500%.
In June 2026, DeepSeek raised over $7.4 billion at a valuation exceeding $50 billion.
Why crypto markets should pay attention
DeepSeek’s open-source models, including the R1, V3, and V4 series, have fundamentally pressured global AI pricing structures. When inference costs drop, AI agent adoption gets cheaper. And when AI agents get cheaper, the blockchain projects building agent-based infrastructure suddenly have a viable cost layer to work with.
What this means for investors
The $400 million to $500 million figure, while significant, carries an important caveat. This is an annualized run rate, not audited annual revenue.
Bloomberg and the Financial Times have reported that DeepSeek may pursue an IPO as early as 2027. At its current $50 billion-plus valuation, the company would be among the most valuable AI-focused IPOs in history.