How a narrow first round became DeepSeek's US$70B overshoot

Capital concentration around a single national champion is now the operating model for Chinese AI funding, and DeepSeek's round cadence — a limited Series A almost immediately chased by a US$70 billion target — shows the mechanism in motion. DeepSeek is the Hangzhou-based lab behind the widely-cited R1 reasoning model; a Series A is a startup's first priced funding round, and "limited" here means deliberately narrow. The first close admitted only Tencent, JD.com, NetEase, and CATL — the world's largest EV battery maker, a strategic surprise for an AI round — alongside VCs Monolith, Loyal Valley Capital, and Shixiang.

The cap was the trigger. Excluding most of the market turned the next raise into a demand pull, not a capital push. South China Morning Post's own sourcing names the gate: the first round was "limited to a few selected backers, leaving other investors eager to support a national AI champion and still seeking ways to get involved." That eagerness is the second round.

The repeat: any time a national AI champion is treated as a narrow balance-sheet pool, the next round arrives faster and at a steeper valuation. Caveat: the $70 billion figure is preliminary, sources spoke on condition of anonymity, and structure remains undisclosed.