DeepSeek has reportedly closed a round of about $15 billion, two days after Washington told federal agencies to strip its software from their systems. Both facts are true at once, and that's the story.
What Bloomberg reported about the DeepSeek round
Bloomberg broke the story on October 22. Citing people familiar with the matter, the wire reported that DeepSeek raised roughly 100 billion yuan, or about $15 billion, in its latest round, one of the largest private fundraisings any Chinese company has pulled off.
The reported backers include CATL, the battery maker, and Tencent, the company behind WeChat. Other outlets have added GigaDevice and Meituan to that list. Those additions are secondhand, so treat the roster as reported rather than confirmed.
DeepSeek didn't respond to a request for comment from TechCrunch. No official announcement, no filing, no statement on the company's site. That matters when the number is this big, and when the timing is this awkward.
Why CATL and Tencent would want in
So what does a battery maker want with a language model lab? The logic sits one layer down, and it isn't about batteries.
CATL runs factories where scheduling, quality checks and equipment maintenance generate enormous amounts of data. Tencent owns the distribution channel: WeChat reaches over a billion people, and its cloud arm sells compute to everyone else. For both, a stake in DeepSeek is a way to influence where model capability lands in their own products.
Neither company has explained its reasoning publicly. Any strategic rationale you read, including this one, is inference from their businesses rather than a stated plan.
Money also buys priority. Investors in a round this size typically get access to model releases, price terms or capacity before the general market does. That's standard practice, not a scoop.
Two days between the US order and DeepSeek's round
On October 20, the US ordered federal civilian agencies to remove DeepSeek's technology from their systems, citing security concerns. On October 22, Bloomberg reported the $15 billion round.
The sequence looks like defiance, and it probably isn't. Fundraising rounds take months to negotiate, and a term sheet signed in September doesn't change just because a policy memo lands in Washington in October. What the timing does show is that the money was committed with the US restriction already visible on the horizon.
The practical effect splits in two directions. Domestically, the cash funds infrastructure and open models. Internationally, the US order complicates the enterprise and cloud path, because agencies and their contractors now have a reason to avoid the stack.
For a developer outside China, both halves matter. DeepSeek's pricing has been a floor for the whole market. If that floor holds, everyone benefits. If procurement restrictions push the company toward domestic buyers, the international product may drift.
From $7 billion in June to $15 billion in October
The pace is the part worth pausing on.
In June, DeepSeek raised more than $7 billion at a $50 billion valuation. In July, reporting put it in talks for another $1.5 billion at around $71 billion. By October, the round had grown to $15 billion, with some reports putting the pre-money valuation near $74 billion.
Those numbers come from different outlets at different moments, and none has been confirmed by the company.
A trend, not a ledger.
Even so, the direction is clear. DeepSeek has moved from a lab with a famous cheap model to a lab with a balance sheet.
For context on how unusual this is, Moonshot AI raised $2 billion at a $20 billion valuation in May, while Manus AI brought in more than $500 million earlier in October. Zhipu and MiniMax listed in Hong Kong in January and sit at roughly $29 billion and $7.2 billion in market value. DeepSeek's round alone is more than double what Moonshot raised in May.
What a Shanghai listing would change for DeepSeek
Reports have pointed to an IPO in China as early as 2027. If it happens, the funding stops being a private arrangement between a founder and a few industrial giants.
Liang Wenfeng, who founded DeepSeek in 2023 after building the quant fund High-Flyer, reportedly holds about 81.6% of the Chinese operating entity. A public listing would dilute that stake, and it would put quarterly expectations on a lab that has so far shipped on its own schedule, which is a very different way to run a research team.
For users, the interesting question isn't the ticker. It's whether a public DeepSeek keeps releasing weights. Open-weight releases are cheap marketing and expensive moat-building at the same time, and shareholders tend to notice which one they're paying for.
What DeepSeek's $15 billion buys in open models
Compute, mostly. Training runs at the frontier are priced in tens of thousands of GPUs and months of power, and export controls make both harder to source inside China, which turns a round this size into a multi-year budget for a constraint that isn't going away.
The second purchase is talent. Senior researchers at Chinese labs have options, and cash plus a credible IPO path is how you keep them.
The third is price. DeepSeek's API rates have forced rivals to cut prices for two years. A well-funded DeepSeek can keep undercutting, which is good news for anyone building on top of a cheap model and bad news for labs that planned to win on margin.
What to watch next in DeepSeek's funding story
Three signals will tell you how this lands.
Whether the investor list firms up. A confirmed Tencent or CATL stake, on the record, would answer the strategic question that reporting has only guessed at.
Whether DeepSeek ships another open-weight model this quarter. Funding announcements often precede a release, and a new flagship would show the money is already working.
And whether US restrictions widen from federal agencies to contractors and cloud providers. That's the line between a company with a domestic business and a company with a global one.
If you build on DeepSeek's API, the near-term answer is boring: keep an eye on pricing pages and keep a second provider wired up. The round changes the company's runway. It doesn't change the risk of building on a single vendor.
