DeepSeek, who once insisted on not raising funds and maintaining independence, finally lowered its noble head and started raising funds! On April 22nd, the news of DeepSeek's first round of external financing exploded across the internet - Tencent and Alibaba, the two giants, are actively negotiating to enter the market. Even more explosive is that the target valuation has doubled from the initial $10 billion to over $20 billion (approximately RMB 136.5 billion). Founder Liang Wenfeng was a tough nut to crack when he said, 'We won't accept external financing, we won't dilute our equity, and we won't be hijacked by commercialization schedules.'. What exactly made him 'bow down'? The answer is in five words: talent has run out. After the explosion of R1 model, ByteDance, Xiaomi, Tencent and other big companies launched a "double or even more" job hunt for the core researchers of DeepSeek. Specifically, let's take a look: ·Luo Fuli (Key Contributor to V2 Architecture) → Head of Xiaomi MiMo Large Model ·Wang Bingxuan (first generation pedestal core) → Tencent ·Guo Daya (R1 reasoning soul, main contributor to GRPO algorithm) → Byte Seed team, rumored to offer nearly 100 million yuan in annual salary ·Ruan Chong (Multimodal Core Researcher) → Yuan Rongqixing, Chief Scientist ·Several core members, including Wei Haoran (OCR direction), have also left one after another Five core backbone members, covering the four major technology lines of R1 inference engine, V3 architecture, multimodal, and OCR, collectively left. According to industry sources, headhunters are accelerating the poaching of DeepSeek members with high salaries and options that are 2 to 3 times higher. This directly led to repeated delays in the release of V4, dragging it from February to late April. Why can't we keep the person? The fundamental reason lies in the options - DeepSeek has never raised funds, and the options held by employees have not been recognized by external capital. In other words, they are just "paper wealth", and the realization is far away. And what big companies are throwing at are all clear prices of "cash+tradable options", how can emotions withstand real money and silver? So this financing, in other words, is a 'talent defense war' - establishing market anchors for employee stock options and stopping the loss from the root. As an investor once said, "This compromise is most likely to price employee options, and it was done too late It is interesting that by 2025, the parent company, Fantasia, will have a quantified return rate of 56.6% and a scale approaching billions. DeepSeek doesn't really lack the $300 million in annual hematopoietic capacity of $700 million. But when the pressure of talent loss approaches the doorstep, relying solely on "not lacking money" is no longer enough. Now DeepSeek has doubled its valuation to $20 billion based on the dark side of the month (estimated at $18 billion). On April 22nd, the official API has been upgraded to support 1 million token contexts, and V4 is also on the verge of success. In addition, for the first time, commercial positions such as "business procurement" appeared in the recruitment information, and Liang Wenfeng began to mention more "productization and commercialization" from the autumn of 2025. Technological idealism ultimately cannot withstand the cruelty of reality - when 'not taking investors' money' meets' those who cannot keep the core ', the once rock solid red line has to bow down to the market.