DeepSeek's First Fundraise Targets $20B+ as Costs Pile Up

DeepSeek is raising outside money for the first time, targeting a $20B+ valuation — double its original ~$10B target just days ago, driven by unusually strong investor demand. Tencent and Alibaba are in active talks, per The Information.

The timing tells a story. DeepSeek's founder once said money was never the problem. Now it is. V4 has slipped on engineering delays, key researchers are being poached by Tencent, Alibaba, and ByteDance, compute costs are ballooning, and the team burned months porting models to Huawei and Cambricon chips — while still secretly training flagship models on smuggled Nvidia Blackwell GPUs.

Ironically, both Tencent and Alibaba have already backed every major Chinese AI rival — Zhipu ($50B+), MiniMax ($30B+), Moonshot (~$18B closing). DeepSeek is currently valued below MiniMax despite global name recognition.

Why it matters: The lab that redefined AI cost efficiency is now fundraising at VC-era multiples — a sign that even lean-and-mean Chinese AI can't escape the compute cost wall.