Baidu’s (BIDU) Kunlunxin Chip Unit Draws ByteDance Interest Ahead of Potential IPO
1 min read
The story
Baidu's Kunlunxin unit — which designs AI training and inference chips to reduce China's dependence on Nvidia — is attracting interest from ByteDance as a potential anchor investor ahead of a standalone IPO. Baidu overall posted FY2025 revenue of $18.5B (only +1.2% YoY) with a thin 4.2% net margin and $15.30 diluted EPS, underscoring why the market has struggled to price the chip and AI cloud assets embedded inside a slow-growing core business.
A Kunlunxin IPO would force a re-rating of the stub BIDU equity and create a cleaner public comps set for Chinese AI silicon alongside Cambricon. The key watch items are: whether ByteDance formalizes a strategic stake (which would be a commercial win for Kunlunxin), any disclosed IPO timeline or valuation, and whether US-listed BIDU ADRs can sustain a rally given ongoing delisting risk and China macro headwinds.
The case — both sides
If ByteDance formalizes an anchor commitment and an IPO valuation leaks — comps like Cambricon suggest AI chip units in China can command 10-20x revenue — a credible sum-of-the-parts would imply BIDU's current market cap ascribes near-zero value to Kunlunxin.
With FY revenue growth of just 1.2% and a 4.2% net margin, the core Baidu business is deteriorating, and Chinese tech IPO markets (both onshore and offshore) remain choppy, meaning the Kunlunxin spinout could drag on for years or be shelved entirely, leaving BIDU with no near-term catalyst.
The house read
Two-sidedBIDU holders face the question of whether a Kunlunxin spinout with ByteDance as a backer unlocks enough conglomerate discount to re-rate the ADR, or whether thin margins, slow growth, and geopolitical risk cap any rally.
Wrong ifUS-listed ADR delisting risk, further regulatory crackdown on Chinese tech IPOs offshore, or ByteDance walking away from talks would each deflate the narrative quickly; core revenue growth at 1.2% provides little fundamental support if the IPO story fades.
Published read · research, not advice