DeepSeek Nears a $74 Billion Valuation as Its ~$7B Pre-IPO Round Closes — With CATL and China's National AI Fund Aboard
The Hangzhou lab is finalizing a 50-billion-yuan raise valuing it at 500 billion yuan pre-money, with a STAR Market filing possible by year-end. The investor list reads like an industrial-policy roster.
DeepSeek is closing in on a funding round that values the Hangzhou AI lab at about 500 billion yuan — roughly $74 billion — pre-money, according to the South China Morning Post. The raise targets around 50 billion yuan (~$7 billion) and is expected to wrap before the end of August, setting up a filing for a Shanghai STAR Market listing that could come as early as year-end, with a market debut targeted for 2027.
The valuation math is aggressive. DeepSeek’s previous round closed at a post-money mark of roughly 450 billion yuan; this one starts at 500 billion before new money. For a company that detonated the “compute moat” narrative in January 2025 by training near-frontier models at a fraction of US lab budgets, the trajectory from open-source disruptor to $74 billion pre-IPO candidate has taken under two years.
Look at who’s writing the checks
The investor roster is the most telling part. Returning backers include Monolith and Shixiang Capital, alongside China’s national AI fund, NetEase, and JD.com. The new names: battery giant CATL, plus CPE, Legend Capital, and Stony Creek Capital — a semiconductor-focused private equity firm. Bloomberg reported the round had briefly paused over leaked remarks from founder Liang Wenfeng to investors before reopening in early August.
A state AI fund, two internet majors, the world’s largest battery maker, and a chip-specialist PE firm is not a typical venture syndicate. It is an industrial-policy coalition. Beijing wants a domestic frontier lab listed on a domestic exchange, funded by domestic capital, running increasingly on domestic silicon — and DeepSeek is the chosen vehicle. The Shanghai STAR Market listing makes that explicit: no US exchange, no foreign-listing risk, retail and institutional Chinese capital funding the training runs.
The contrast with the US financing model is stark. OpenAI is leaning on SoftBank commitments and tens of billions in structured debt; Anthropic is pitching IPO investors a $30 trillion addressable market. DeepSeek raises a comparatively modest $7 billion — but with the implicit backing of a state that treats frontier AI as strategic infrastructure. Its cost discipline makes that sum go further: this is the lab that made efficient training its brand.
For developers, the practical takeaway is that DeepSeek’s open-weight releases now carry a pre-IPO commercial logic. A listed DeepSeek needs revenue lines, and its API pricing has consistently undercut US labs by an order of magnitude. Cheap frontier-adjacent inference isn’t going away; it’s about to get a public balance sheet behind it.
An IPO filing by December would make DeepSeek the first frontier AI lab to go public anywhere. The race to ring the bell first isn’t happening on Nasdaq.