Kunlunxin, Baidu’s AI chip unit, is preparing a dual IPO on Shanghai’s STAR Market and in Hong Kong. According to reporting by Qianer Liu at The Information, the unit has asked prospective IPO investors to do something beyond writing a check: buy its semiconductors.
Baidu holds about 58% of Kunlunxin, which carries a reported valuation north of $3 billion. The unit is in the middle of a pivot — from a captive supplier that mostly fed Baidu’s own data centers to a third-party merchant chasing outside customers. ByteDance and Tencent have both surfaced as potential buyers. The IPO is the moment that pivot gets priced.
The key insight: Asking investors to also be customers collapses two questions into one. “Will you fund us?” and “Is there a market for us?” are supposed to be answered by different people. When they’re the same people, a strong order book can be a financing artifact rather than proof of demand.
U.S. export controls didn’t just restrict chips — they created a captive market on the other side of the wall. Inside that market, the scarcest two resources are capital and validated demand. Bundling investors and customers is what you do when both are scarce at once: you use the capital raise to conjure the demand signal, and the demand signal to justify the raise. It’s efficient. It’s also circular.
“A captive market hides the line between a customer and a shareholder. Kunlunxin’s IPO is where that line gets tested in public.”
Three Implications
1 · Demand you can’t fully trust
If anchor buyers are also anchor investors, the order book reflects portfolio interest as much as product-market fit. Underwriting that requires unbundling the two — hard to do from the outside.
2 · Sanctions are a domestic-champion subsidy
Every restriction on Nvidia and the U.S. frontier is, functionally, a demand transfer to Kunlunxin, Huawei, and peers. The wall that contains American chips is the same wall that incubates Chinese ones.
3 · The compute layer is decoupling for real
Two AI stacks are forming, not one — and the split is moving down from models to silicon. A standalone, publicly traded Chinese AI-chip champion makes the separation structural, not temporary.
Business Engineer Framework
The Permission Layer · Compute Tier
Export controls reshape the bottom of the AI stack, not just the top. When access to silicon is gated by geopolitics, demand and capital fuse — and whole national champions get built in the gap. Map where the compute layer is splitting before you bet on it.
Asking your IPO investors to become your customers is unusual everywhere except a sanctioned market, where it’s almost rational. It tells you how badly China wants a homegrown AI-chip champion, and how thin the line between belief and demand has become in getting one to scale. Kunlunxin may well clear its listing and grow into the order book it’s pre-selling. But the move itself is the story: the U.S. built a wall to keep its chips in, and on the far side, a chipmaker is now financing its own demand to climb out.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
