Baidu says Chinese buyers want local AI chips due to ‘supply chain’ issues

Baidu told investors it expects “good days ahead” for its Kunlunxin AI inferencing chip business, citing rising inference demand and supply constraints that could persist. On the Q2 call, Baidu’s AI Cloud accelerated with cloud infrastructure rental revenue up 50% YoY to nearly $1.1B and GPU cloud revenue up 283% YoY, helping offset modest overall revenue growth of 4% YoY to $3.9B. The company is working to list Kunlunxin and will share “concrete info” soon, while the backdrop remains Nvidia/US-China chip access uncertainty that Baidu frames as driving domestic chip adoption.

Analysis

The main market implication is not the chip itself but the optionality around vertical integration: if Kunlunxin gets listed, Baidu can re-rate as both an AI cloud operator and a domestic semiconductor platform. In the next 1-3 months, that matters more for sentiment than earnings because a spinout can surface hidden value and let investors price the chip asset separately from the slower core search business.

For competitive dynamics, the likely loser is not just Nvidia but any China-facing AI stack that depends on imported accelerators or on a resumed cross-border licensing regime. If domestic buyers stay boxed in, Chinese cloud and hardware vendors will be pushed toward local silicon even at lower performance, which supports adoption of Baidu’s chips, Huawei-linked ecosystems, and any ODMs that can package a credible domestic bill of materials. The second-order effect is margin compression for incumbents that must either absorb higher inference costs or pass them through and risk share loss.

The contrarian point is that this is a policy story more than a pure product story. Consensus may be underestimating how quickly procurement behavior shifts once buyers optimize for supply certainty rather than peak performance, but it may also be overestimating how much near-term profit accrues to Baidu if the chip business remains captive and capital intensive. Falsifiers are straightforward: a real reopening of Nvidia sales into China, a delayed Kunlunxin listing, or any evidence that Baidu’s AI cloud growth decelerates once promotional pricing normalizes.

Key Decisions for Investors

• Long BIDU on a 1-3 month horizon into any Kunlunxin spinout announcement; use weakness to build because the setup is more about multiple expansion and hidden-asset monetization than immediate EPS. Risk/reward is attractive if the market starts assigning even a modest separate valuation to the chip business.
• Pair trade: long BIDU / short NVDA as a small, policy-driven relative-value position over the next 1-3 months. The thesis is that China demand shifts toward domestic inference chips faster than NVIDIA can regain share; keep size modest because NVDA’s direct China exposure is already constrained and the short can be noisy.
• Avoid pressing a naked short in NVDA; use it only as a hedge against China-semiconductor policy risk. The better expression is via relative value, since the downside case for NVDA from China alone is more incremental than catastrophic unless export approvals worsen again.
• Watch for a confirmed Kunlunxin listing timeline as the key catalyst. If Baidu fails to provide concrete terms within the next quarter or if AI-cloud growth slows meaningfully, fade the re-rating and reassess the long BIDU thesis.