News
DeepSeek Lands at Least ¥80B Raise: Tencent, CATL Are the Big Backers
Bloomberg reported on October 6 that DeepSeek's new funding round is nearing close at at least ¥80 billion (~$12 billion). Tencent Holdings and CATL are the round's largest investors. The round started targeting just ¥50B and kept growing as subscriptions overshot expectations.

Bloomberg reported on October 6 that, according to people familiar with the matter, DeepSeek’s new funding round is nearing close at at least ¥80 billion (about $11.93 billion). Tencent Holdings and CATL are the round’s largest investors.
The round originally targeted just ¥50 billion. Investor demand overshot expectations and the number kept growing. Reports say the final figure could approach ¥100 billion. DeepSeek, Tencent, and CATL all declined to comment; Reuters says it cannot independently verify the report for now.
What ¥80 Billion Means
No Chinese AI startup has ever raised this much in a single round. Globally, it would rank among 2026’s largest fundraises. Just these past few days, OpenAI has also been negotiating at least $30 billion in new funding at a $1.4 trillion valuation target. Both sides are making big autumn bets.
Where the money comes from is worth noting too. The largest checks come from Tencent and CATL: one an internet platform, one a battery giant — neither a traditional dollar VC. DeepSeek’s capital map is shifting from financial capital to industrial capital. That suggests investors want more than financial returns — industrial capital like this usually wants a channel for technology collaboration too.
Where the Money Burns
The report didn’t specify use of proceeds. Judging from DeepSeek’s recent moves, the directions are guessable.
First, compute. Bigger models mean training and inference both cost real money. Last month DeepSeek released V4.1-Flash, optimized for speed and throughput. Driving inference costs down is its signature play for grabbing share in the US developer market. Cheap prices rest on massive compute investment.
Second, domestic chip adaptation. DeepSeek has been working with Huawei on programming tools around the Ascend chips, centered on the open-source TileLang. The goal is clear: run models on domestic chips and reduce dependence on NVIDIA. Building a chip toolchain is a cash-burning long-term project.
Third, going public. Reuters reported last month that DeepSeek hired CITIC Securities to prepare for a STAR Market IPO. Timing, valuation, and raise size are all unsettled. Aggregated reports suggest a listing could come in early 2027. Raising at this point is likely also paving the way for the IPO.
A Few Caveats
First, be clear: this comes from people familiar with the matter — none of the three companies confirmed. In the reporting, “at least ¥80 billion” is “nearing close” and “¥100 billion” is “possible.” How much is signed versus indicated is unclear right now. Read the news with care and wait for final numbers.
Also, a mega-round doesn’t mean the company faces no operating pressure. DeepSeek wins markets with a low-price strategy; the API price war runs on compute efficiency. Whether money converts into a moat depends on the next generation of models, not the raise amount itself.
Two things to watch next: the round’s final number, and whether the STAR Market listing makes substantive progress.