Key Takeaways
- • Shanghai Everlead Capital, up 164% year-to-date, has begun trimming positions in optical and chip-packaging stocks—core infrastructure plays in the AI compute layer—signaling profit-taking rather than panic selling among China's top AI hedge funds.
- • Compute stocks (chips and hardware) gained 62% over the measured period but fell approximately 13% in the most recent month, while applications and software lagged all year (down 9%) before recovering roughly 5% last month, indicating a late-cycle rotation across AI market layers.
- • Two major Chinese hedge funds—Everlead and Hunjin Capital—are rotating capital away from crowded AI hardware positions into cheaper traditional stocks, with Hunjin estimating the AI hardware cycle is now 60% complete, double its February assessment.
- • Power and infrastructure stocks, which rose 11% earlier in the period, have stalled alongside fading compute stocks, with power equities now re-trading as AI-linked assets rather than on traditional rate and regulation narratives.
- • The data suggests a market-wide layer rotation rather than a full bubble collapse, with leadership shifting from hardware and compute infrastructure toward software and applications as institutional capital reallocates.
Summary
China's highest-performing AI hedge funds have begun booking profits on their most crowded positions, prompting renewed debate about whether the market faces a correction or a healthy consolidation. Shanghai Everlead Capital, which has delivered 164% returns year-to-date, has trimmed holdings in optical and chip-packaging stocks that form the compute infrastructure backbone of AI data centers. The fund's moves follow substantial rallies in these names: Zhongji Innolight has reached a trillion-yuan market capitalization, while Yangtze Optical Fibre has rallied twelvefold—gains substantial enough to trigger profit-taking across the sector.
A second major player, Hunjin Capital, has executed similar tactics, reducing exposure to memory-chip names facing anticipated pricing-power deterioration and rotating proceeds into undervalued traditional equities. Notably, Hunjin estimates the AI hardware cycle is now 60% complete—a significant acceleration from its February assessment of 30% completion. Importantly, neither fund is signaling an imminent crash; rather, they appear to be managing portfolio risk after outsized appreciation and repositioning for what they view as a maturing cycle phase.
Broader market data confirms these fund-level moves reflect a systematic rotation across AI investment layers. Compute stocks appreciated 62% over the measured period but declined roughly 13% in the most recent month, while power and infrastructure infrastructure gains of 11% have subsequently stalled. Conversely, applications and software—which underperformed significantly (down 9% year-to-date)—captured approximately 5% of inflows last month as capital rotated into the sector. This pattern is consistent with late-cycle dynamics rather than capitulation, as institutional money shifts from saturation plays in hardware toward less-crowded software and applications exposure.
Power equities occupy a critical inflection point in this rotation. These stocks have historically traded on interest-rate and regulatory narratives but are increasingly being repriced as proxy AI infrastructure plays. The sector's transition in investor framing and valuation drivers may signal that markets are normalizing AI-related allocations rather than abandoning the theme entirely. The outcome hinges on 2027 capital expenditure trajectories—a variable that will determine whether current profit-taking represents healthy rebalancing or the opening chapter of extended de-rating across AI infrastructure assets.
A second major player, Hunjin Capital, has executed similar tactics, reducing exposure to memory-chip names facing anticipated pricing-power deterioration and rotating proceeds into undervalued traditional equities. Notably, Hunjin estimates the AI hardware cycle is now 60% complete—a significant acceleration from its February assessment of 30% completion. Importantly, neither fund is signaling an imminent crash; rather, they appear to be managing portfolio risk after outsized appreciation and repositioning for what they view as a maturing cycle phase.
Broader market data confirms these fund-level moves reflect a systematic rotation across AI investment layers. Compute stocks appreciated 62% over the measured period but declined roughly 13% in the most recent month, while power and infrastructure infrastructure gains of 11% have subsequently stalled. Conversely, applications and software—which underperformed significantly (down 9% year-to-date)—captured approximately 5% of inflows last month as capital rotated into the sector. This pattern is consistent with late-cycle dynamics rather than capitulation, as institutional money shifts from saturation plays in hardware toward less-crowded software and applications exposure.
Power equities occupy a critical inflection point in this rotation. These stocks have historically traded on interest-rate and regulatory narratives but are increasingly being repriced as proxy AI infrastructure plays. The sector's transition in investor framing and valuation drivers may signal that markets are normalizing AI-related allocations rather than abandoning the theme entirely. The outcome hinges on 2027 capital expenditure trajectories—a variable that will determine whether current profit-taking represents healthy rebalancing or the opening chapter of extended de-rating across AI infrastructure assets.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.