The technology that has been the market's best-performing theme for two years has also woven itself so thoroughly into stocks, bonds, private equity and even Treasurys that some of the world's largest pension funds say true diversification is now an illusion.
Monte Tarbox, chief investment officer of the $327 billion New York City Retirement Systems, recently turned down a private equity fund pitch specifically because it was heavy on artificial-intelligence holdings, telling the manager directly, "we have to pass on this precisely because we don't know when enough is enough or when it's too much." His concern was not about AI's existential risks, the subject dominating public debate, but about a more basic portfolio management problem: a pension fund's job is to spread risk across asset classes, and AI exposure has become so pervasive that spreading risk no longer means what it used to.
The scale of that pervasiveness is now measurable. Artificial-intelligence infrastructure companies, chipmakers and hyperscalers together make up roughly 40% of the S&P 500's total market value, according to an estimate from Goldman Sachs. Three chipmakers alone account for more than a quarter of the benchmark index used for emerging markets. AI-related debt accounts for nearly half of all investment-grade bond issuance so far this year and 87% of all venture capital funding, according to data compiled by Apollo Global Management. Alphabet alone carries a $4 trillion equity market value, $130 billion of outstanding debt, and infrastructure connecting more than 30 data centers globally, meaning a single company's fortunes now touch a pension fund's equity, credit and, indirectly, real assets allocations all at once. In a recent survey of 90 sovereign wealth funds conducted by Invesco, more than half named market concentration as the top risk they associate with AI-related investments.
Lisa Shalett, chief investment officer at Morgan Stanley Wealth Management, said AI compounds a diversification problem that was already building before the current boom, since the technology is now deeply entrenched in the success of both equities and credit markets simultaneously. She compares the current dynamic to prior infrastructure buildouts in railroads and the internet, both of which eventually reached a point of overbuilt capacity, adding that the only real uncertainty with AI is timing, not whether that point eventually arrives.
Some of the largest pension systems have started trying to quantify their own exposure rather than simply worrying about it in the abstract. The Los Angeles County Employees Retirement Association, managing $94 billion, ran a review combining bottom-up analysis by its own money managers with a top-down assessment supported by MSCI, and landed on an estimated range of 8% to 19% of its holdings carrying AI exposure, a wide band that Chief Investment Officer Jonathan Grabel said still provides a useful baseline for future decisions even without more precision. The California Public Employees' Retirement System, the largest public pension fund in the country, has seen its own excess AI exposure climb through investments in large privately held companies and has adopted a framework known as the Total Portfolio Approach, which evaluates investments against each other for the whole portfolio rather than within separate asset-class silos, specifically to better track cross-cutting themes like AI exposure.
Not every allocator believes the answer is to pull back. Kari Vatanen, head of asset allocation and alternatives at Finland's Elo Mutual Pension Insurance, which manages 36 billion euros, said his firm is using AI tools to monitor its own public holdings' sensitivity to AI adoption, but argued that capping AI exposure purely for diversification's sake could hurt returns if the theme continues to dominate market performance for years to come, and that getting the theme right matters more than getting the asset-class mix right. A Bloomberg Intelligence index tracking AI-related stocks globally has outperformed the broader market by 11 percentage points annually, a gap large enough to make any allocator's decision to underweight the theme a genuinely costly one if the trend persists. That tension, between the discipline diversification normally demands and the real cost of missing out on the market's dominant theme, is now sitting at the center of how the largest pools of retirement capital in the world are being managed.