DeepSeek is reportedly in talks to raise about $1.5 billion at a $71 billion pre-money valuation — a jump of roughly 42% just weeks after it closed its first-ever external round at around $50 billion. If the round closes at that number, it would be one of the fastest valuation step-ups the AI sector has seen this cycle — and further evidence that investors will pay up for teams shipping frontier-class models at a fraction of the usual cost.
Editor's note: This is a developing story based on reporting from Bloomberg, the Financial Times, TechCrunch and other industry outlets. The round is in talks, not closed — we'll update this post as details are confirmed.
What we know
- The number: DeepSeek is talking to investors about a $71B pre-money valuation on a raise of about $1.5B, first reported by Bloomberg and picked up by TechCrunch and the Financial Times.
- The step-up: Its previous round — roughly $7B at a ~$50B valuation — closed only weeks earlier, in mid-June, per SiliconANGLE and Asia Financial. That makes the new talks a ~42% re-rating in about a month.
- The bigger arc: DeepSeek only opened to outside capital this spring — reports put its valuation near $10B in April, ~$45–50B by late May, and $71B in the current talks. That's roughly a 7x re-rating in three months.
- What's next: The company is reportedly planning a China IPO in 2027, per TechCrunch and Tech Funding News.
- Company status: DeepSeek has not issued an official statement on the talks at the time of writing.
Why it matters
DeepSeek built its reputation on doing more with less. Its open-weight releases stunned the industry by matching much larger, more expensive models on reasoning and coding benchmarks — while giving the weights away and undercutting incumbents on API pricing. That combination reset expectations for how cheaply a frontier lab can operate.
A 42%-in-weeks re-rating signals three things:
- Open-weight momentum is compounding. Capital is flowing to the labs proving that state-of-the-art doesn't have to mean closed and costly.
- The efficiency story is the valuation story. DeepSeek's edge has always been cost-per-capability. Investors appear to be pricing that moat, not just headline benchmark scores.
- The gap between "cheap" and "frontier" keeps closing. Every step-up like this pressures closed incumbents on price and pace.
Where DeepSeek sits in the open-model race
For teams choosing a model today, DeepSeek is already a serious option: strong reasoning, genuinely open weights you can self-host, and an API that costs a fraction of the majors. A larger war chest would mean more compute, faster release cadence, and broader tooling — good news for anyone building on open models.
If you're weighing it against the alternatives, see our ranked guide to the best AI chatbots and the open-source AI directory for self-hostable options.
What's still unconfirmed
We're holding the following as unverified until primary confirmation lands:
- Whether the round closes, and at what final valuation and size — the $71B/$1.5B figures reflect talks, not a signed deal
- The identity of the lead and participating investors
- The structure (primary, secondary, or a tender) and the close date
- Any official comment from DeepSeek
- The 2027 IPO timing, which remains a reported plan rather than a filing
We'll update this post as details are confirmed.
Want funding rounds, model launches and industry moves like this first? Join the ChooseMyStack Weekly Brief — one email a week, curated and filtered.