DeepSeek, the Chinese lab that rattled Silicon Valley in early 2025 by matching frontier models at a fraction of the cost, has hired CITIC Securities to prepare an IPO on Shanghai’s STAR Market. It is currently raising at a reported 500 billion yuan, around $75 billion, up from a $50 billion-plus round in June. A filing could come this year, with a public debut plausibly in 2027.
The number is eye-catching, but the location is the story. DeepSeek is listing at home, on a mainland Chinese exchange, not in New York or Hong Kong. That is a choice loaded with meaning: capital controls, US-China tech tension, and a deliberate signal that China’s flagship AI champion will be funded by Chinese capital and answerable to Chinese regulators. For Western investors, the practical upshot is that the most interesting pure-play AI listing of the cycle may be one most of them cannot easily buy.
Why it matters beyond the ticker
DeepSeek’s whole reputation is built on doing more with less, training competitive models without the top-tier Nvidia chips export controls denied it. An IPO turns that scrappiness into a war chest, and the stated use of funds is telling: compute infrastructure, model development, and talent retention. In other words, the efficiency story was partly necessity, and given capital, DeepSeek intends to spend like everyone else. That undercuts the tidy narrative that China had found a cheaper path to frontier AI; it may simply have found a cheaper path to the starting line.
For anyone tracking the AI investment landscape, DeepSeek’s raise sits alongside a broader Chinese AI IPO rush, and it is the clearest sign yet that the compute-arms-race framing (see our piece on China’s state compute plan) is now flowing into public markets. The valuations are enormous, the geopolitics are inseparable from the financials, and “can you even access this?” is a real question. (Not investment advice; cross-border listings carry particular risks.)