Chinese profits rose 25.7%. The CSI 300 fell 9% and the Star 50 fell 29%

Onshore-listed Chinese companies’ profits rose 25.7% in the three months to June (fastest in nearly five years), but gains are heavily concentrated in AI-linked firms. At the same time, the CSI 300 is down ~9% this quarter and the tech-heavy Star 50 has fallen ~29%, as investors shift from viewing AI spending as a “promise” to treating it as near-term cost.

Analysis

The key signal is not that Chinese earnings improved; it is that the market is still refusing to pay for them unless the profit shows up in cash flow and not just capex rhetoric. That leaves a narrow leadership set: domestic AI infrastructure, semis, networking, power management, cooling, and industrial automation should keep gaining relative share, while internet/platform names that are funding the spend but not yet monetizing it likely stay de-rated. The second-order effect is dispersion: when profits are concentrated in AI-linked firms, benchmark performance becomes more index-agnostic and more stock-specific, which is typically bearish for passive flows and factor products.

Near term, the drawdown in the tech-heavy complex looks more like a positioning unwind than a clean fundamental reset, so a tactical rebound is plausible over the next 2-4 weeks if earnings revisions stabilize. The bigger catalyst path is 1-3 months: if AI capex starts translating into order growth for suppliers and margin expansion for enabling hardware, the market will re-rate the supply chain first, not the large platform names. Falsifiers include any evidence that spending remains front-loaded with no revenue conversion by the next reporting cycle, or any policy shift that starves credit to the same capex names the market is trying to own.

The contrarian miss is that investors may be treating “AI spending” as uniformly bad, when the losers and winners are diverging sharply. Broad China tech exposure can still underperform if monetization stays weak, but a domestic AI supply-chain basket can outperform even in a risk-off tape because it is one of the few areas with visible earnings elasticity. The move is likely under-discriminated rather than outright overdone: headline beta may stay weak, but the supply-chain leg of the trade still has room if revisions keep improving.

Key Decisions for Investors

• Long ASHR vs short KWEB for 1-3 months: own onshore A-shares where AI capex is already showing up in reported profits; fade offshore internet/platform names that are still being valued on uncertain monetization. Target is relative outperformance of domestic beneficiaries; stop if KWEB starts outperforming on clear earnings-guidance upgrades.
• Tactical long CQQQ on pullbacks, but only as a supply-chain expression, not a broad China-tech beta trade. Use a 4-8 week horizon and require follow-through in semicap, networking, or power-management names; if the next earnings cycle shows no order acceleration, cut quickly.
• Short a broad China tech basket into strength if the rally is driven by multiple expansion rather than revisions. The risk/reward improves if CSI 300 rebounds before STAR 50 breadth improves; that would be a clean signal that the market is still punishing AI spenders more than rewarding AI enablers.
• Watch-list, not a trade yet: build a basket of Chinese domestic AI infrastructure names only if we see confirmable revenue conversion over the next quarter. Missing data: capex-to-revenue payback and free-cash-flow conversion; without that, the earnings surprise could still be dismissed as subsidy- or timing-driven.
• If policy shifts toward broader consumption stimulus instead of capex support, reduce exposure to AI-linked cyclicals and rotate toward consumer proxies. That would be the main structural falsifier for the supply-chain thesis over a 6-18 month horizon.