DeepSeek Hits $1B Revenue Run Rate, Eyes Shanghai STAR Market IPO

Last January, DeepSeek shook Silicon Valley with the claim that it had trained a frontier-grade model for just $6 million, proof that a lab could compete with the biggest U.S. AI players without massive capital or compute clusters. A year and a half later, DeepSeek faces a different test: turning that technical shock into an actual business.

According to The Information, DeepSeek’s annualized revenue run rate has crossed $1 billion, more than doubling from just a few months ago. Founder and CEO Liang Wenfeng reportedly disclosed the figure himself during a recent investor meeting. The jump is attributed to a mix of continued demand for DeepSeek’s models and a recent price hike of roughly 2.3 to 4.5 times. It’s worth noting that a $1 billion run rate isn’t $1 billion already booked — it’s what the current pace would produce over 12 months if it holds. Liang reportedly told investors that more than 70% of DeepSeek’s compute is still going toward training new models, with less than 30% allocated to inference — a sign the company is betting on its next generation rather than milking current demand. Earlier this month, DeepSeek released DeepSeek-V4.1-Flash, which it says offers faster inference, higher throughput, and support for scaling to larger models.

DeepSeek is a Hangzhou-based AI lab founded in 2023, bankrolled early on by founder Liang Wenfeng’s quant hedge fund, High-Flyer, which he co-founded in 2015. High-Flyer funded DeepSeek entirely in-house for years before the lab sought outside capital.

That first external round closed in June. Per Reuters, DeepSeek raised roughly $7.4 billion (about 51 billion yuan) at a post-money valuation exceeding $50 billion. Tencent, CATL, JD.com, NetEase, and China’s National AI Industry Investment Fund participated — through a limited partnership structure managed by Liang, giving investors economic exposure with a five-year lockup but no voting rights, preserving Liang’s roughly 84% economic stake and near-total voting control via dual-class shares.

Now DeepSeek is back for a second round. Reuters reports the company is raising about $7.45 billion (50 billion yuan) at a reported valuation of 500 billion yuan (about $75 billion), aiming to close by the end of October. Separately, Reuters reported DeepSeek has hired CITIC Securities to prepare for a listing on Shanghai’s STAR Market, though the timing, size, and target valuation remain undecided.

The race to prove real revenue is heating up across China’s AI labs. Moonshot AI, the maker of chatbot Kimi, raised $2 billion at a $20 billion valuation — nearly five times its valuation six months earlier — with Kimi’s annualized revenue topping $200 million. Among labs that have already gone public in Hong Kong, the fortunes have diverged. Zhipu AI‘s stock surged 524% in the 43 days after its January debut, well above its initial $7.1 billion valuation. MiniMax doubled on its own debut the same month, but its stock later sank to roughly a quarter of that valuation after the company announced plans for its own STAR Market listing — a cautionary sign for what a mainland listing announcement can do to a Hong Kong-listed stock, the same path DeepSeek is now eyeing.