Diversification Is an Illusion: How Does "Smart Money" Invest in a Market Dominated by AI?

The prospect of investing in a private equity fund looked promising—its manager had a strong track record, and the New York City Retirement System (NYCRS), with $327 billion in assets, had the means to commit funds. But Chief Investment Officer Monte Tarbox flatly rejected the deal—the fund’s portfolio was overloaded with assets related to artificial intelligence, according to Bloomberg.

“I told the manager: we have to say no precisely because we don’t know where the line is between ‘enough’ and ‘too much,’” Tarbox explained his decision, citing the current cycle of AI development. — “When that day finally comes, we don’t want to be the ones left out in the cold, hearing, ‘Well, why did you agree to everything?’”

Virtually all major institutional investors on Wall Street and beyond now face a similar dilemma. Their challenge is to allocate capital across various assets in a way that avoids excessive concentration. However, artificial intelligence has penetrated so deeply and widely across all asset classes that the traditional diversification on which funds have relied has become an illusion, writes Bloomberg. Oninvest summarizes the agency’s in-depth report on how major investors managing billions of dollars are seeking a way out of this situation.

According to Goldman Sachs, companies related to AI infrastructure account for about 40% of the S&P 500’s total market capitalization. The shares of just three semiconductor manufacturers make up more than a quarter of the emerging markets index. According to Apollo Global Management, AI accounts for nearly half of all investment-grade bond issuance this year and 87% of venture capital funding. An additional risk is posed by so-called circular financing schemes, in which large technology companies effectively finance one another.

AI has a significant impact even on broadly diversified multi-asset portfolios, according to a study by JPMorgan Chase analyst Thomas Salopek. He built four models with different asset allocations, reflecting the typical asset allocation of pension and endowment funds, and found that in recent years they have shown a strong positive correlation with AI-related risk factors.

In a recent Invesco survey of 90 sovereign wealth funds, more than half cited market concentration as the main risk associated with investing in AI. Behind this concern lies an understanding of the inevitable—and unpredictable—end of the current AI investment cycle, says Lisa Shalett, Chief Investment Officer at Morgan Stanley Wealth Management.

“At some point, there will be enough infrastructure. We’ve seen this happen hundreds of times throughout history—whether it’s railroads or the Internet. The only question is when it will happen,” she said.

The problem is that, unlike traditional classifications by asset class, industry, or geography, there is no single definition of what constitutes exposure to AI. Exposure to AI can range from chip manufacturers that are entirely focused on this market to companies that are merely integrating the technology. Therefore, attempts to assess such exposure are, in part, a subjective judgment about what exactly to consider a risk, notes Bloomberg.

At the Finnish company Elo Mutual Pension Insurance, which manages $41 billion in assets, portfolio managers use AI to identify companies that are most sensitive to developments in artificial intelligence. To do this, they use AI-based tools that help track the extent to which the portfolio’s publicly traded assets depend on the spread of this technology. Kari Vatanen, head of the asset allocation division, believes that strictly limiting the share of AI for the sake of diversification could hurt returns if the theme continues to dominate the market for many years to come.

"Having the right sector in your portfolio is more important than having the right asset class," Vatanen is convinced.

Investments in AI have indeed been extremely profitable for most of the past two years. The Bloomberg Intelligence index, which tracks the stocks of global AI-related companies, has outperformed the broader market by an average of 11 percentage points per year.

This may be why the largest pension funds, on the whole, are not yet in a hurry to deviate significantly from the market when it comes to AI-related risks, according to Bloomberg. Michael Markov of Markov Processes International, a financial analytics firm, conducted a factor analysis of approximately 50 of the largest pension funds and estimated that, on average, their additional exposure to AI — in both public and private markets — was close to zero, although it is significantly higher for some individual funds.

One such exception is the California Public Employees' Retirement System (CalPERS), the largest public pension fund in the United States.

According to Markov, his excessive exposure to AI has grown in recent years due to investments in large private companies. A spokesperson for the fund told Bloomberg that private equity investments are “thoroughly screened for exposure to AI and related risks.”

According to the article, some large investors are shifting to the so-called Total Portfolio Approach. This investment concept moves away from dividing investments by asset class and instead evaluates them relative to one another based on the interests of the portfolio as a whole. In recent years, the TPA has been gaining popularity among institutional investors. CalPERS, among others, has recently adopted this approach.

Opinions differ on the best way to apply TPA, but generally, deeper implementation of this approach requires the creation of a unified infrastructure that aggregates data across different asset classes and enables quantitative analysis. For example, it can be used to assess and test a fund’s aggregate exposure to factors such as inflation or AI.

In June, the Australian fund Aware Super, which manages $171 billion in assets, completed a multi-year project to upgrade its internal investment platform using tools from third-party providers, including BlackRock. “This is now the foundation for how we build our internal AI strategy. True diversification isn’t just about geography or sector; it’s about understanding the correlations between assets and themes,” said Michael Clavin, head of liquidity and markets at Aware Super.

Marsh Investments and Retirement, an American consulting group that manages $846 billion in assets, developed a system based on agent-based AI in just six months to track exposure to key themes, including artificial intelligence.

“The key is to understand exactly what you have, so you can figure out where to go from here,” says McDougall. “Those who already have well-established systems in place can, even before making an investment, largely assess where it will lead. But those whose systems aren’t working well will most likely find out only when it’s too late.”

At NYCRS, Tarbox’s team is looking for ways to mitigate risks. Portfolio managers are trying to identify the most vulnerable areas and develop guidelines—for example, setting a maximum allowable percentage of software development companies in a single fund. The main challenge, according to Tarbox, is keeping up with technology: AI is evolving so rapidly that its risk profile must be constantly reassessed.

“We need to do more than just increase the resolution of our microscope—we also need to increase the frequency of our assessments,” says Tarbox. “Virtually every fund we invest in will inevitably increase our exposure to AI.”

This article was AI-translated and verified by a human editor