DeepSeek raises $7.4B at $50B+ valuation, plans 2027 IPO

DeepSeek, the Chinese AI startup that stunned Silicon Valley with its cost-efficient open-source models, has raised $7.4 billion in the largest private AI financing in Chinese history, valuing the three-year-old lab at over $50 billion. Founder Liang Wenfeng, previously elusive, held a four-hour pitch meeting from Hangzhou where he described his team as “ordinary people.” The company plans to go public in 2027, likely raising billions more. With these funds, DeepSeek will aggressively expand its global reach, undercutting US rivals on pricing. Its V4 Flash model costs just $0.02 per standardized intelligence task on Artificial Analysis, compared to $2.75 for Anthropic’s Claude Fable 5 – a 100x difference.

Major US companies are already switching. Lindy AI moved from Anthropic to DeepSeek, cutting costs to 10% of previous levels – saving millions annually. TinyFish, an enterprise AI startup, switched from frontier models to Chinese open-weight ones and saw inference costs drop 90%. Airbnb runs a mix of Chinese open-source models, Pinterest touts their utility, and Microsoft CEO Satya Nadella has considered using DeepSeek in Copilot Cowork. Chinese open-source models now claim over 30% share on OpenRouter, up from 2% in 2024. This low-cost strategy aligns with Beijing’s push for accessible AI, threatening the proprietary business models of OpenAI and Anthropic.