Hang Seng Posts Worst Week Since April 2025 as AI Bets Unwind

The Hang Seng Index closed June 26 down 1.8% at 22,671.86, capping what South China Morning Post reported as the worst weekly performance since April 11, 2025 — a 5.2% loss for the week. The Hang Seng Tech Index shed a further 3.4% on the day, and across the border the ChiNext 50 tumbled 4.6% while the CSI 300 slid 3%. The origin of the selling was not in Hong Kong: South Korea's KOSPI plunged 9% and triggered a second trading halt as leveraged AI stock positions unwound, with Taiwan's Taiex falling 3.6% in parallel, and that pressure then spread into Hong Kong and mainland markets.

Gary Dugan, CEO of The Global CIO Office, put the underlying dynamic plainly: "Investors want AI exposure but are less willing to pay a single multiple for long-duration growth, margin expansion and market leadership simultaneously." He characterized the broader shift as a move from "concept to execution" — a recalibration rather than a rejection of AI as a theme. The Federal Reserve's hawkish pivot and weakened consumer confidence in China affecting internet platforms were cited as compounding factors, adding fundamental weight to what might otherwise look like purely technical margin-call contagion.

The honest caveat is that a week of selling triggered by leveraged position unwinding is not necessarily a verdict on AI's commercial trajectory. The reporting does not detail which specific companies bore the deepest cuts, or how much of the Korean and Taiwanese leverage remains to be unwound — those figures would tell you whether the worst is past or still ahead. What the article does not give you is any AI revenue data from the affected companies that might have catalyzed the move; the trigger appears to be cross-market contagion from a technical correction, not a fundamental earnings miss.

The clearest implication of Dugan's framing is that it sets a visible test: companies with demonstrable near-term AI monetization become relatively more defensible when markets stop paying for long-duration growth stories in a single bundle. The recalibration hurts the narrative plays first and the revenue plays last, which means the current volatility may end up clarifying the AI landscape more than a year of analyst reports could.