AI binge tests risk appetite

PRIVATE equity’s aggressive push into artificial intelligence (AI) is likely to face closer scrutiny from institutional investors as they assess whether their portfolios have become too dependent on a single technology theme.

With money flowing into everything from data centres (DC) and power generation to chips, software and credit, investors are expected to demand greater transparency from private-market managers over just how much of their portfolios are ultimately tied to the AI boom.

According to a recent Bloomberg report, some of the world’s biggest investors are already asking buyout firms for more detailed information on their AI-related holdings, as investment firms including Apollo Global Management Inc and Blackstone Inc pour billions of dollars into the technology.

The concern is not simply about whether AI will succeed.

Investors are increasingly worried that a downturn could spread across seemingly different parts of their portfolios because so many investments are linked to the same underlying growth story.

“We are concerned that so much exposure is tied to a single thread, AI, be that through DCs, through software, through our venture portfolio,” John Bradley, senior investment officer at the Florida State Board of Administration, which managed US$306.8bil as of May 31, tells Bloomberg. “We’re trying to be mindful about our overall exposure to AI.”

That is making portfolio-wide exposure a bigger issue for pension funds, insurers, sovereign wealth funds, endowments and other institutional investors that allocate billions of dollars to private capital.

For some, the next step could be putting limits on AI-related investments within infrastructure funds.

Such funds have become important financiers of the physical foundations of the AI economy, including DCs, power sources and other infrastructure needed to support growing computing demand.

Bloomberg reports that one Canadian pension plan had even rejected a potentially lucrative co-investment in a DC because of concerns that it would concentrate too much of its portfolio on AI.

The hesitation comes as private-market managers increasingly finance an interconnected AI ecosystem.

Capital is moving into memory chips, DCs, electricity generation and energy grids, creating a web of investments whose fortunes can ultimately depend on sustained AI demand.

Uncertain returns

The problem for investors is that the potential returns from many of these investments remain uncertain.

It can also be surprisingly difficult to identify AI exposure in the first place.

A DC, for example, may be classified as a real estate investment, while a semiconductor maker could sit within a broader technology allocation.

This means an investor looking at individual funds may not immediately see how much of the overall portfolio depends on AI.

The issue is becoming particularly complicated in chip financing.

Some asset managers are investing in chip-financing or chips-on-demand providers through infrastructure platforms, according to Bloomberg, yet some advisers to pension funds argue that loans tied to semiconductor assets do not necessarily fit the traditional definition of infrastructure, which typically focuses on durable assets expected to last for decades.

Chip leases, by contrast, may be structured around the five-year useful life of semiconductors.

Some chipmakers and AI hyperscalers have attempted to address the concern by offering residual-value backstops, promising that leased equipment will retain a minimum value.

But investors and advisers remain wary that such assumptions could underestimate the risks.

“Assuming demand for chips is well in excess of supply is a concerning assumption built into some of the pricing of these loans,” John Nicolini, a partner at Cerity Partners who advises institutions on real assets, tells Bloomberg.

“It’s important to manage exposure and take a critical view of loans that rely on residual value as a material part of the upside on return.”

As private capital becomes increasingly involved across the AI buildout, investors are therefore likely to focus less on individual holdings and more on the combined exposure sitting across their portfolios.

“Understanding how much of a portfolio is ultimately tied to AI has become one of the top questions investors are asking,” Sud Murugesu of Partners Capital, which manages about US$75bil for endowments, foundations and wealthy families, tells Bloomberg.

Another concern is the interconnected nature of the AI financing boom.

Deals in which companies invest in businesses that subsequently buy their products or services have become increasingly common.

Amazon.com Inc, for instance, has invested in OpenAI, which uses the technology giant’s computing power for ChatGPT. Microsoft Corp and Nvidia Corp are major backers of rival AI company Anthropic PBC, which is also tapping computing capacity linked to those firms.

Such circularity could magnify losses during a downturn. It may also create conflicts of interest and make it harder for investors to assess the underlying economics of individual transactions.

Ty Gellasch, a former US Securities and Exchange Commission official who heads investor advocacy group Healthy Markets Association, tells Bloomberg that institutional investors and congressional aides are increasingly asking questions about circular financing.

“As risks of AI financing become more clear, compliance officers at pension funds want to dig in,” he is quoted as saying by Bloomberg.

Valuations concerns

Valuations are adding another layer of concern.

The rapid rise in the value of AI-related companies has helped drive returns for some investors, but it has also made it harder to judge how much risk is being accumulated across multiple funds managed by the same investment firm.

Brookfield Asset Management, for example, has a dedicated AI infrastructure fund, while other strategies, including energy, also have investments benefitting from AI-driven demand.

Blackstone likewise has AI-related exposure, including DCs and power producers, spread across several strategies.

For limited partners, that makes a potential AI downturn harder to model.

“Getting the necessary transparency into DC and AI investments is definitely a cause of concern for limited partners who don’t fully understand their exposure,” Neal Prunier of the Institutional Limited Partners Association, a trade group representing investors, tells Bloomberg.

Still, institutional investors have also benefitted from the AI boom.

Ontario Teachers’ Pension Plan, which manages about C$303.2bil (US$219.3bil), credited its first-half return to gains including a pre-initial public offering investment in SpaceX, Elon Musk’s rocket, satellite and AI company.

SpaceX also raised the largest-ever public debt in June.

Yet, even successful investments are prompting a closer look at concentration.

Stephen McLennan, Ontario Teachers’ chief investment officer for asset allocation, tells Bloomberg that the pension fund is trying to gauge its overall direct exposure to assets linked to what it calls the “AI complex”.

For investors, the challenge ahead may increasingly resemble the early stages of the Covid-19 pandemic, when markets moved so quickly that institutions had to rapidly work out which industries would be winners and which would be hit.

AI is creating a similar problem, except that the boundaries between winners, losers and beneficiaries can shift just as quickly as the technology itself.

“It’s like Covid in the sense that things are moving very quickly and there is less clarity on the health of portfolios,” Prunier says.