DeepSeek, the Hangzhou-based open-weight lab whose R1 and V3 releases rattled US frontier pricing in early 2025, is closing its first external funding round at approximately 50 billion yuan ($7.4 billion), according to people with knowledge of the matter cited by CNBC on June 3.
The post-money valuation is expected to land between 350 billion and 400 billion yuan ($52 billion to $59 billion). Fewer than 10 investors are participating. Tencent is considering a 10 billion yuan commitment and battery giant CATL is looking at 5 billion yuan, making them the two largest external backers. China’s national AI fund, NetEase, JD.com, IDG Capital, and Monolith Capital are also in final discussions, per the same people.
The number that reframes the round is not the headline total. It is the 20 billion yuan that founder Liang Wenfeng is contributing from his own balance sheet, covering roughly 40 percent of the raise. At this size, most founders take dilution and pocket capital. Liang is doing the opposite: paying to concentrate his position. That is a specific structural signal. Founders who buy into their own rounds at scale are pricing in asymmetric upside they do not want to share. The governance implication is equally direct: a cap table with fewer than 10 names, anchored by the founder at 40 percent, is one where external pressure to change strategy, open the model further, or accelerate an IPO will be limited.
The investor lineup also carries strategic weight beyond pure capital. CATL’s involvement is not a passive bet on AI software. The company is a dominant player in EV battery supply chains and has been building out AI data center power infrastructure. A DeepSeek relationship gives CATL a direct customer for the energy storage solutions it is developing for large-scale compute workloads. Tencent, whose own Hunyuan model trails ByteDance’s Doubao and DeepSeek domestically, gains proximity to the lab it cannot currently match. For both, this is less a portfolio investment and more a strategic hedging position within China’s AI supply chain.
No IPO statement has accompanied the round, and DeepSeek has made no public comments on future listing plans. That silence reads differently now that the round is structured the way it is. A sub-10 investor syndicate with a dominant founder stake is not a cap table that needs a public market to provide liquidity on any near-term timeline.
The contrast with Anthropic’s capital path is instructive. Anthropic, a closed-weights US safety lab, has pursued a series of massive external rounds, disclosed a commercial agreement with Google worth up to $300 billion over time, and is widely expected to pursue a public offering as the next source of growth capital. Two labs, both operating at the frontier of capability, have chosen opposite structures: one distributing governance outward toward institutional backers and public markets, the other consolidating it inward around the founder with open weights as the product surface.
That divergence matters for how each lab behaves under competitive pressure. Anthropic has external shareholders whose return expectations constrain the option space. DeepSeek, after this round, has a founder who has personally written a nine-figure check to stay at the table, a national strategic backer in the AI fund, and a product that ships weights publicly. When a company like DeepSeek cuts prices or drops a new model release, the decision tree has fewer nodes to clear.
For teams currently evaluating open-weight models for production deployments in the next two quarters, this round removes one of the few remaining reasons to treat DeepSeek as a flight-risk vendor. Liang has put his own capital behind continuity.
Source: CNBC (cnbc.com), published June 3, 2026.
