Chinese hedge funds warn AI rally is a ‘super bubble’

Shanghai-based Banxia Investment Management and Beijing-based Wealspring Asset Management warned the recent surge in AI-related stocks has detached from fundamentals. Wealspring described the rally as a ‘super bubble’, citing moderating revenue forecasts for model developers and heavy reliance on ongoing capital across the AI supply chain.

Banxia told clients that signs of a correction may already be appearing. The firm pointed to slowing revenue growth expectations at leading AI model developers and weaker momentum at Anthropic as early indicators that hyperscalers may take longer to monetise AI services than investors expect.

Wealspring argued many companies tied to AI infrastructure are priced on aggressive growth assumptions and lack durable competitive advantages. The firm added that a large share of businesses in the supply chain depend on continuous capital inflows to finance expansion rather than on stable operating cash flow.

AI-linked equities have posted outsized gains globally, especially in semiconductors and memory. Micron Technology and SK Hynix have rallied on expectations of sustained spending on data centres and high-performance computing for AI workloads, lifting valuations above historical norms for hardware and components firms.

Hedge funds raising caution noted the sector is becoming more capital intensive. Building and running data centres, buying specialised chips and expanding memory capacity require large, ongoing investments, and some managers flagged uncertainty over whether expected revenue will arrive fast enough or at sufficient scale to support current prices.

Banxia’s founder urged investors seeking AI exposure to exercise ‘very, very’ high caution. Wealspring highlighted valuation gaps across the AI supply chain and added many firms rely on new funding rounds to pursue growth rather than stable cash flow.

Some institutional investors have cut exposure to AI names over valuation and capital-risk concerns, while others maintain heavy allocations. Analysts and portfolio managers are increasingly scrutinising forward revenue assumptions and profitability timelines for model developers, chipmakers and data centre operators.

Questions under review include whether hyperscalers can convert AI services into dependable, high-margin revenue at scale and whether hardware suppliers can sustain growth without periodic fresh capital. Fund managers issuing warnings recommended reassessing the assumptions built into current prices and weighing the sector’s capital demands and execution risks before increasing exposure.