August 10, 2026 (InvestinChina.asia) — China’s artificial-intelligence equity complex does not harbour a systemic bubble, even after a blistering first-half rally and a sharp June-to-August pullback, according to Goldman Sachs chief China equity strategist Kinger Lau. The recent selloff has restored “reasonable and healthy” alignment between prices and forward earnings, Lau said in interviews and a strategy note published August 9, leaving three subsectors — power supply chains, hardware infrastructure and physical AI — best positioned to capture the technology’s real economic dividend.
No blanket bubble, but local froth cleared
Lau conceded that parts of the A-share market overheated in June, when select AI hardware names on the STAR Market and ChiNext touched five-year valuation highs. Since then, China’s AI hard-tech basket — which had averaged a 33% gain in the first half — has retraced more than 20% from its June peak, with the STAR 50, ChiNext and CSI 1000 indices all down over 20% from recent highs alongside the global AI derisking trade.
In Goldman’s read, the drawdown has flushed out four latent hazards: speculative long positions have been closed, sector valuations have normalised off historical extremes, onshore margin leverage has unwound, and crowding in AI hardware holdings has eased materially. The house’s ten-signal dashboard — spanning pricing, risk appetite, fund flows, positioning, policy and fundamentals — frames the move as a risk-clearing correction, not a thesis-ending bust.
The macro backing for that call is a market-cap-versus-upside gap. Chinese AI-related stocks represent roughly 11% of the global AI equity universe, yet overseas investors allocate only about 1% of their AI portfolios to China — a underweight Goldman treats as structural headroom rather than a warning sign.
Three subsectors Goldman likes
Power supply chain. Sustained build-out of domestic compute capacity keeps demand visibility high for Chinese electrical-equipment makers, whose global market share is still climbing. Goldman frames the segment as a multi-year structural winner rather than a cyclical trade.
Hardware infrastructure. Printed-circuit boards, optical modules and data-centre equipment carry what Lau called “high earnings certainty” over the next two to three years, with order books anchored to hyperscale capex that is less sensitive to near-term sentiment swings.
Physical AI. Industrial intelligence and humanoid robotics sit inside China’s manufacturing ecosystem and export value chain, giving them a competitiveness edge that pure-software aspirants lack. Goldman sees global export runway opening for Chinese robot and factory-automation platforms.
Beyond the hardware layer, Lau highlighted AI tokens — the billing unit for model inference — as a potential new export engine. Chinese large-model providers price single-token inference far below US rivals, accelerating commercialisation of agents and industry-specific clouds at home and, increasingly, abroad.
From single-track betting to four-line diversification
Goldman’s August 9 note, titled AI Changes the Game: Correction, Rotation and Diversification, pivots the allocation advice away from blanket overweighting of AI hardware. The team’s four post-correction lanes:
- Step into Hong Kong internet and AI soft-tech. After months of underperformance, H-share platforms and application names offer valuation repair and earnings leverage to AI monetisation, hedging hardware volatility.
- Policy-beta with dual demand. Pair domestic-demand AI names with export-oriented AI companies, leaning on Beijing’s industrial-policy tailwinds and the “15th Five-Year Plan” pipeline.
- Oversold non-tech with upgraded earnings. Traditional-sector stocks that sold off hard but carry upward profit revisions can serve as defensive ballast in a high-volatility tape.
- IPO alpha and cash-return stories. Primary-market listings and high-dividend, cash-generative equities offer correlation breaks from the secondary AI trade.
The throughline: stop crowding into one hard-tech theme, blend soft and hard tech, balance tech with non-tech, and link primary and secondary markets. Short-term A- and H-share tech volatility stays elevated, so the edge moves from beta to stock selection and cross-subsector balance.
Why the call matters now
The contrast with the US AI debate is deliberate. Where Washington’s AI complex is increasingly read as a debt-funded, concentration-heavy financialisation trade, Goldman positions China’s AI equity story as earlier-cycle, cheaper on forward multiples relative to delivered earnings, and anchored in physical supply-chain advantages Washington cannot easily replicate. That framing lets global allocators treat Chinese AI not as a satellite of the US bubble but as a diversifier with 23%-style low correlation to US tech tapes and a token-cost moat in inference.
Lau’s bottom line is that the summer correction priced out the excess, not the opportunity. The winners from here are segments where China already owns scale — electrons, substrates, robots — rather than segments where it is chasing Silicon Valley’s narrative.