Summary: Reports on May 6, 2026 said DeepSeek was negotiating its first outside funding round, with the discussed valuation rising from $20 billion to $45 billion in weeks. The Big Fund was reported as the likely lead, while Tencent and Alibaba were negotiating stakes, and nothing was signed. The outside round would change its founder-funded capital structure and reportedly aims to enable employee equity and retain researchers.
Word arrived on May 6, 2026: TechCrunch, drawing on reporting from the Financial Times and Bloomberg, says DeepSeek is negotiating the first outside funding round it has ever taken — with the valuation on the table climbing from $20 billion to $45 billion across a matter of weeks. The China Integrated Circuit Industry Investment Fund, the state vehicle better known as the “Big Fund,” is reportedly in line to lead, and Tencent and Alibaba are said to be negotiating places in the round.
What matters here is not the size of the check but who writes it. Until now, the most famous open-weights lab on the planet ran entirely on capital generated by founder Liang Wenfeng’s hedge fund; letting state money and cloud heavyweights onto the cap table is a decision that will redraw how China’s AI camp competes.
Weeks Separate $20 Billion From $45 Billion
Nothing is signed, and the numbers could still move. The FT’s reporting has the valuation under discussion going from $20 billion to $45 billion in only a few weeks — and if that figure sticks, DeepSeek lands in the top rank of the world’s most valuable independent AI labs. A valuation that more than doubles inside a month is less a financing than a bidding contest, and it reveals as much about the scarcity of eligible buyers as about the asset itself: only one lab with DeepSeek’s standing is still on the market, and every negotiator at the table knows it. Bloomberg’s account adds that the expected lead is the Big Fund, the national platform at the center of China’s push for semiconductor self-sufficiency, while Tencent and Alibaba work out follow-on stakes. TechCrunch reached out to DeepSeek and had gotten no answer by press time.
What Prompted the Raise
The company has gone its entire life without a single outside investor — the starkest line separating it from every other frontier lab. The FT’s version: founder Liang Wenfeng, holder of close to 90% of the shares, is raising not because compute runs short but because staff are leaving — rivals have begun picking off DeepSeek researchers, and holding on to them takes a mechanism for putting equity into employees’ hands.
Frame it that way and the round reads as a defensive play for talent first and a war chest for growth second. The lab that upended the industry with training budgets nobody thought possible has run headlong into the market truth that any star researcher has a price — a lesson for every team whose advantage lives inside a few heads. The shape of the deal will matter as much as its size, too: handing employees equity brings governance changes, thins the founder’s stake, and introduces reporting duties a lab funded solely by its founder never had to answer to.
National Money Attached to the Round
Bring in the Big Fund and the deal stops being a purely commercial affair. The vehicle sits at the center of China’s semiconductor strategy, and its leading this round would put state capital directly behind the flagship of open-weights models. The reported interest from Tencent and Alibaba stacks on another layer: the cloud players want DeepSeek’s models lashed tightly to their own compute, yielding an arrangement in which the state fund contributes legitimacy and the clouds contribute reach. Teams running DeepSeek weights in production should read that composition both ways — deeper capital and long-term staying power on one side of the ledger, geopolitical exposure and possible strings attached to usage on the other.
Open Weights, Huawei Hardware
Ever since its low-cost training runs stunned the world in early 2025, DeepSeek has pressed near the frontier on reasoning and coding without giving up open weights — anyone can pull V4-Pro from Hugging Face. Its models have likewise been tuned for Huawei chips. Pair the software half (DeepSeek) with the hardware half (Huawei) and you get the trunk road China is building toward homegrown AI capability under export controls: when Nvidia’s best chips are difficult to get hold of, a leading open model plus domestic accelerators stops being a contingency and becomes the default plan.
Set that beside Meta’s reported lean toward closing up its frontier releases — the open-source split that opened 2026 — and open weights now carry different meanings across the Pacific. On the American side it remains a strategic option; for China it has turned into the national team’s badge. No matter where the final terms land, developers should assume DeepSeek’s open-weights policy stays put for the near term — that very openness is what the arriving capital is paying for.