PE’s AI exposure has institutional investors on alert

Institutional investors are demanding greater transparency from private equity and other alternative asset managers over their exposure to AI, as billions of dollars flow into data centres, power infrastructure, software and other parts of the AI ecosystem, according to a report by Bloomberg.

Pension funds, insurers, sovereign wealth funds, endowments and foundations are increasingly asking managers to identify AI-linked investments across their portfolios, according to executives at investment firms and large institutional investors.

The concern is that AI exposure may be considerably larger than it appears because investments connected to the technology can sit across several strategies. Data centres, for example, may be classified as real estate or infrastructure, while semiconductor investments can appear within broader technology allocations.

Some investors are now considering whether AI-related assets should be excluded from infrastructure strategies altogether. One Canadian pension fund reportedly declined a co-investment opportunity in a data centre because of concerns about increasing its overall AI exposure.

Alternative asset managers have become important sources of capital for the infrastructure underpinning the AI boom, financing everything from data centres and electricity generation to energy networks and semiconductor-related assets. But the rapid expansion has prompted concerns that an eventual downturn in AI could have a wider impact on private market portfolios.

The interconnected nature of the sector is another source of concern. Investors are increasingly examining so-called circular transactions, in which AI companies, technology providers and infrastructure businesses invest in or finance one another, potentially creating additional vulnerabilities if valuations fall.

AI-related lending is also raising questions for infrastructure investors. Some alternative managers are financing semiconductors and offering chips-as-a-service through infrastructure strategies, despite the relatively short useful lives of the underlying assets.

Critics argue that loans repaid over the roughly five-year life of semiconductor equipment may not fit comfortably within infrastructure strategies that traditionally target assets designed to remain productive for decades.

Some chip manufacturers and AI hyperscalers have attempted to address the issue by offering residual-value guarantees intended to assure lenders that leased equipment will retain a minimum value. Investors remain concerned, however, that such assumptions could prove optimistic if demand for computing capacity weakens.

For limited partners, identifying the aggregate exposure is becoming increasingly difficult as AI-related investments spread across multiple funds managed by the same sponsor.

The issue extends beyond individual assets to the concentration of AI investments across a manager’s broader platform. Brookfield Asset Management, for example, operates a dedicated AI infrastructure strategy, while other parts of its business, including energy, can also benefit from increased electricity demand associated with AI.

Blackstone likewise has investments connected to the AI buildout across multiple strategies, including data centres and power generation. Such cross-fund exposure can make it difficult for limited partners to determine the potential impact of an AI correction on their overall commitments to a manager.

The scrutiny comes despite some institutional investors benefiting significantly from the AI boom. Ontario Teachers’ Pension Plan, for instance, attributed part of its 9.5% first-half return to a pre-IPO investment in SpaceX, which has interests spanning rockets, satellites and AI.

The CAD303.2bn pension fund is nevertheless working to establish the scale of its direct exposure to assets connected to what its chief investment officer for asset allocation, Stephen McLennan, describes as the “AI complex”.