A retired teacher in Pennsylvania may not know it. Her pension leans on Nvidia (NVDA). David Uberti and Heather Gillers of the Journal counted the holdings of 13 big public pension funds. A basket of 15 AI-linked stocks drove about 57% of their U.S. stock gains since late 2022. Even a board member of her fund says she struggles to tally its AI exposure. “It’s like being Nancy Drew,” she said.
The boom has been good to pensions. Federal Reserve data put state and local pension assets at $7.3 trillion in June, up from $5.8 trillion in 2023. The 13 funds held about $195 billion in those 15 stocks in June. That’s about 27% of their U.S. stock holdings, up from 12% when ChatGPT launched. The basket produced $138 billion in paper gains. Nvidia led, then Alphabet (GOOGL), Broadcom (AVGO), Amazon (AMZN) and Micron (MU). Funding levels tell the bigger story. In the 2007-09 crisis, the money on hand covered 93% of promised benefits, then only 63%, by the Equable Institute’s count. Funding is back to 86% in its latest projection.
Here’s the wedge. Officials in those same states are fighting data centers in the last stretch of the midterm campaign. New York put a moratorium on many new data centers in July. Pennsylvania’s governor slowed them in August, and the state treasurer who helps oversee the teachers’ fund wants a pause. A New York City councilman grilled Meta (META) and Alphabet on the risk to retirement funds. “We’re on both sides of the bet,” one New York candidate said. Officials from both parties are in the fight, and we don’t pick a side. President Trump, who backs fast building, said last week that data centers must supply their own power or pay host towns a dividend. We covered that rule on Oct. 23.
What a fund can’t easily sell
Pensions mostly track indexes. Trillion-dollar tech names sit at the top of every index, so a fund can’t dodge them without giving up returns. Politics has pushed pension funds to sell before, over apartheid South Africa and over guns. Retirees sometimes lost money on those exits. The exposure goes beyond stocks too. California’s $650 billion fund put $500 million into a data-center vehicle. A Texas teachers’ fund has $2.6 billion of data-center exposure in private markets. Equable’s Anthony Randazzo says a board shouldn’t change its investing lightly, because that means choosing a side.
Our read
This is Retirement, M10, and the lesson is bigger than pensions. If big pensions with 27% of their U.S. stocks in AI names struggle to measure the bet, so do most of us with an S&P 500 fund. We hold the names in the chart, and Meta. Nvidia is held at weight in 31 books, Broadcom in 29, Alphabet in 28, Micron in 23, Meta in 20 and Amazon in 19; nothing added. We size them on purpose, and we count them as one theme, because they tend to rise and fall together.
The one thing in your plan to look at is your AI exposure, counted across every account. This week, pull up your 401(k)’s biggest fund and read its top ten holdings. Add up the AI names. If the number surprises you, that’s the point of the exercise. You check the roof before the storm, and the same goes for a portfolio that’s done well. We’re glad to count it with you.