DeepSeek set to raise $7.4B at a $74B valuation
The biggest private model lab in China is about to get materially bigger. DeepSeek is set to raise $7.4 billion from existing and new investors at a $74 billion valuation, according to the Wall Street Journal — the capital earmarked for research and for an aggressive expansion into the computing infrastructure it currently rents. The deal would cement DeepSeek not just as an open-weights disruptor, but as a company with the balance sheet to build the scale behind its famous price advantage.
The raise is the culmination of a funding saga that has run for months. DeepSeek completed its first outside capital raise back in June — a $7.4 billion round that reported valuations landed between $50 billion and $59 billion, led by Tencent and CATL. That round was reported to have restarted and expanded over the summer (several outlets earlier flagged an $8 billion target at the $74 billion mark), and the WSJ's latest news appears to lock the round near that level. The pattern behind it is consistent: DeepSeek has repeatedly said it wants to own the compute infrastructure that underpins its cost edge rather than keep renting it, and it raised API prices in July even as rivals raced to undercut it.
Why this matters: DeepSeek is one of the few labs outside the US that runs at genuine frontier-grade scale, and it built a global reputation on shipping capable open weights at a fraction of Western prices. A $74 billion valuation with a war chest aimed at matched compute puts that model on a fundamentally more durable footing — it stops being a clever cost playbook and starts being an infrastructure company. Chinese reports have also framed the round as a pre-IPO step, with a possible market debut as early as 2027; that would turn the private lab that rattled Western AI pricing into a public one.
What to watch: whether the round's control structure (DeepSeek's founder has kept unusual decision-making rights through the June deal) survives intact at this valuation, and how much of the new capital actually goes to building versus the ongoing price war with Alibaba and Zhipu, who have both been open-sourcing cheaper models this week.
Do you think owning compute becomes the moat that decides which AI labs last — and which get bought before they get big? Tell us in the comments.
