Chinese hedge funds that have generated outsized returns from this year’s artificial intelligence rally, including Everlead Capital and Hunjin Capital, are beginning to reduce exposure to some of the sector’s biggest winners, according to a report by Bloomberg.
The move signals a more cautious approach even as they remain broadly constructive on the long-term investment theme. Several managers have told investors they are closely monitoring valuation risks and identifying potential catalysts that could prompt a larger-scale exit from AI-related positions, according to recent investor letters.
Shanghai-based Everlead, whose Growth Strategy No3 Fund returned 164% through the end of May, said it had reduced holdings in optical communications and advanced semiconductor packaging companies after benefiting from the sector’s strong performance.
Meanwhile, Hunjin Capital’s Yueyang G1 fund, which gained around one-third over the same period, also pared back selected AI positions. The manager, which oversees more than RMB5bn ($739m), said it had established specific conditions that would trigger a broader reduction in AI exposure.
The moves reflect a growing shift among some of China’s best-performing hedge funds, which continue to back the structural AI investment story but are becoming increasingly selective as valuations rise.
Dan Bin, founder of Shenzhen Oriental Harbor Investment Management, cautioned investors against being swept up in market enthusiasm, writing that while investors can profit from periods of exuberance, they should avoid becoming part of the frenzy. His firm, which manages more than RMB10bn, believes AI capital expenditure could begin to slow during 2027 after several years of rapid expansion.
The caution comes as AI-related stocks have rallied worldwide. Semiconductor and memory-chip companies have been among the strongest performers, with global investors continuing to pour money into businesses expected to benefit from AI infrastructure spending.
For many Chinese equity hedge funds, AI exposure has become a key differentiator this year. Managers that positioned portfolios around the theme have generally outperformed peers that remained underweight or avoided the sector.
Despite their positive long-term outlook, managers are increasingly identifying indicators that could signal the rally has become overstretched.
Hunjin Capital highlighted several potential warning signs, including slowing improvements in AI models, weakening demand, a loosening technology supply chain and the possibility of public resistance to widespread AI adoption. The firm also warned that storage manufacturers could face margin pressure as customers develop alternative technologies.
Founder Li Yue said the fund had begun reducing positions where investor consensus had become too optimistic and where the balance between risk and potential returns had become less attractive, while selectively increasing exposure to more traditional sectors.
Shanghai Chaser Asset Management is also monitoring several indicators that could point to a cooling AI investment cycle, including slower revenue growth from AI models, declining cloud computing capital expenditure and falling prices for leased computing capacity.
Not all managers share the cautious outlook. Shanghai-based Banxia Investment Management argued in a recent investor letter that early signs of stress are already emerging, citing concerns over AI company revenue growth and rising operating costs as evidence that the sector may be approaching a turning point.
Nevertheless, most bullish managers are not yet abandoning the theme.
Everlead Capital said leading overseas storage companies continue to offer significant growth opportunities, while Hunjin argued the rally remains supported by genuine earnings growth rather than speculative momentum. Chaser Asset also maintained that current market conditions differ fundamentally from the technology bubble of the early 2000s because the AI investment cycle is underpinned by tangible industry demand.
Even so, Hunjin believes the sector is becoming more mature. The firm said its internal assessment of the AI hardware investment cycle has advanced to 60 on a 100-point scale, compared with 30 earlier this year, suggesting the opportunity is becoming progressively less attractive.
