DeepSeek Surpasses $1B ARR and Prepares $7.5B Funding Round Ahead of IPO
DeepSeek has reportedly surpassed $1 billion in annualized run-rate revenue—doubling its revenue within a few months—and is finalizing a $7.5 billion funding round at a 500 billion RMB (~$70 billion+) valuation. The company is completing this round ahead of a planned initial public offering (IPO) on the Shanghai Stock Exchange. This financial milestone highlights strong commercial momentum and user retention for DeepSeek even after raising its API prices. It also underscores the massive capital requirements driving China's leading AI laboratories toward public capital markets to sustain frontier model development. DeepSeek founder Liang Wenfeng informed investors that the lab allocates over 70% of its compute power to training next-generation models, reserving less than 30% for inference. Internal tests also showed that DeepSeek's lightweight models can run stably on consumer-grade gaming GPUs while handling the vast majority of everyday user tasks.
## BACKGROUND
In AI development, compute resources are divided between training (building new base models from vast datasets) and inference (running existing trained models to generate outputs for users). Model lightweighting encompasses compression techniques like quantization and distillation, which enable large language models to run on standard consumer hardware rather than expensive cloud servers.