When a researcher resigned from Anthropic out of fear that artificial intelligence might destroy all of humanity, the startling revelation panicked everyday Americans and spurred policymakers around the world to action.
It also troubled the supervisors for the largest public pension in the United States, setting off a debate last month about whether the funds they oversee might also be at risk — and what they should do about it.
With more than $600 billion in assets, the California Public Employees’ Retirement System had long invested heavily in Amazon, Facebook, Google, Nvidia and the vanguards of the dawning A.I. age. It was hardly alone in that embrace of Big Tech, and like many, its well-placed bets had yielded soaring returns benefiting its more than two million workers.
But the question of whether its investments had come at a cost seemed to weigh on the administrators of the California system, known as CalPERS, at a routine meeting in September. As they pored over the latest numbers, some on the board openly grappled with their responsibilities in a world in which A.I. could disrupt American life.
“The market is even more reliant now on tech than it has ever been. We have certainly prospered greatly over the last two decades from our investments there,” said Lisa Middleton, one of the board’s members. “That always raises the question: At what point do we need to start looking to what the downside could be?”
The moment in California underscored one of the great uncertainties in America’s grand economic transformation. For millions of workers, the fate of their hard-fought retirement savings may well hinge on the boom-or-bust potential of artificial intelligence.
At a time when technology giants dominate financial markets, the exuberance on Wall Street around A.I. has been to their benefit, sending stocks soaring and boosting retirement accounts accordingly. But that jubilation could just as easily turn to doubt, delivering a major blow to the money that many Americans rely on in older age.
The risks are evident in state and local retirement programs, which together serve more than 37 million people. In states like California, New York and Texas, the officials who oversee these funds have started to question their exposure to A.I. — and some have even turned away investments — out of fear that the sky-high financial returns may not last forever.
The work to protect retirees’ savings has not exactly been easy, given the financial needs of pension funds and the extent to which technology companies dominate financial markets. With the arrival of data centers, the ambitions of Silicon Valley have only expanded further into everything from energy to real estate — and, at times, Americans’ retirement savings have helped to supply the capital for these expensive endeavors.
With roughly $6 trillion in assets, the public pensions are “feeding the capacity of the A.I. companies to grow,” said Anthony Randazzo, the executive director of Equable Institute, a nonprofit that studies pensions. In an analysis of available data, his firm estimates that state and local retirement plans’ exposure to A.I.-related investments could be as high as $700 billion.
But, Mr. Randazzo said it raised many questions: “If there is a systematic reduction in the value of A.I. companies, how does that map onto all the rest of the assets of these plans? Does it take everything else down with it?”
For the moment, A.I. is powering Wall Street while setting the stage for highly anticipated public offerings from the likes of Anthropic, which unfurled details about its finances on Monday. The excitement has factored heavily into the thinking at the White House that it should refrain from issuing regulations that might slow down the industry.
“I will never stifle the growth of technology that will be bigger than the Industrial Revolution,” President Trump said in late September.
Increasingly, though, there is worry that the excitement may not last. The concern stems from the otherworldly pace at which tech companies are constructing data centers and other related infrastructure, a set of investments that could reach $10.3 trillion between 2025 and 2032, according to a new paper from Stijn Van Nieuwerburgh, a professor and economist at Columbia University.
In one scenario, the A.I. investments could contribute heavily to the nation’s economic growth, making workers more productive. But there is also the possibility that sufficient demand — and revenue — do not materialize quickly enough to warrant all of the newly created computing resources, causing financial havoc.
The tension was top of mind when state and local retirement officials from around the country gathered in August for an annual industry forum, held in Chicago. Amid the sessions on pension finance and management, Paulo Carvão, a research fellow at the Harvard Kennedy School, warned that A.I. was “entering the stress test period.”
In general, public pensions are funded through worker and employer contributions as well as investments. These retirement systems, on average, are able to fund about 86 percent of the benefits that they are obligated to pay, according to an October analysis from Equable, which described some funds’ finances as improving yet “fragile.”
After the conference, Mr. Carvão recalled that attendees had expressed a range of views about the promise and peril of A.I.: a skepticism about the technology, an eagerness to “increase my exposure” to its gains and even, at times, a concern about whether it was “morally acceptable” to support an industry that could displace the very workers whom pensions serve.
Nevertheless, Mr. Carvão said he emphasized there was a “lot of wealth created around” A.I., for now, but upheaval “could materially affect” pension finances.
Already, that concern about blowback has prompted some pension officials and the people they serve to question if they might be vulnerable to an A.I. bubble.
This June, for example, the retirement system for public workers in Los Angeles found that as much as 19 percent of its investment portfolio was directly implicated by artificial intelligence. Officials overseeing a retirement system for school employees in Ohio said in October that they were undertaking their own review.
One of the nation’s largest teachers unions wrote the Trump administration last month that it was “gravely concerned” for the pension savings of its nearly two million members, and it urged the federal government to rein in risky A.I. practices. And, in New York City, pension leaders grew troubled enough about A.I. concentration recently that they started looking differently at their own investments.
“You see a systemic level of risk that is making us wary,” said Mark Levine, who oversees the city’s pension systems as comptroller.
This year, New York City turned away a chance to invest about $250 million in a private equity fund involving small and medium-size A.I. companies. Mr. Levine declined to offer specifics, but said he passed on the pitch, which had good fundamentals, because he was “worried it was too concentrated in A.I.”
The cautiousness among some pension officials offers a stark contrast to the scramble by tech giants, chip makers and data center developers for the capital needed to build out artificial intelligence. To source that money, they have increasingly looked beyond their own balance sheets and adopted a series of novel, evolving and potentially risky A.I. financing tactics.
Much of that has involved debt, including bonds by cash-flush technology giants and other private debt deals that have supported a wide array of A.I., real estate and infrastructure projects. The many buyers include public pensions, which had been increasing their exposure to private capital and private equity even before the current boom.
“It’s a huge opportunity,” Christopher Sheldon, the co-head of credit and markets for KKR, an investment firm, said at a conference hosted by Barclays in September. Mr. Sheldon, whose firm is among the institutions raising capital for Nvidia and others, added that he expected “pension is going to follow” with interest.
The extent of the risk depends on the level and nature of the debt and other investments that pensions hold. Still, the trend has troubled watchdogs at the Open Markets Institute and Americans for Financial Reform, two left-leaning advocacy groups.
In an October report, they warned that an A.I.-fueled downturn “would hit pension funds hard, particularly since A.I.-linked assets dominate stock and corporate bond markets as well.” Maya Jenkins, a senior policy analyst at Americans for Financial Reform Education Fund and one of the authors of the report, added that there was “an open question about who is going to absorb the losses should these large tech firms run into problems.”
For pension officials and the workers they serve, some of the anxiety stems from past crashes. That includes the global financial crisis, which delivered a major financial blow to public pensions, swelling the gap between their existing assets and future liabilities by hundreds of billions of dollars, according to David Draine, the principal officer with the state fiscal policy project at Pew Charitable Trusts.
It took states many years to recover, and some pensions in states like New Jersey and Illinois still suffer from severe shortfalls. Mr. Draine said another major economic calamity could force states to commit their own funds to plug any holes.
That painful history appeared to weigh on officials in Texas, who gathered in September to discuss the finances for a pension system that serves about two million local teachers.
Jase Auby, the program’s chief investment officer, began his presentation by highlighting the healthy growth of the system’s portfolio. But he eventually reached a slide comparing the A.I. boom with past periods of economic transformation, from electrification to the internet.
Some of those inflection points “ended in tears for the investment community,” acknowledged Mr. Auby, who described the current moment as a “cautionary tale” as tech giants pour huge sums into A.I.
His diagnosis: “So far, so good, actually.” But Mr. Auby added: “We’ll continue to monitor that and watch that, but history definitely is giving us some lessons here today.”
The post U.S. Pensions Are Riding the A.I. Wave. For Better or Worse. appeared first on New York Times.
