Margaret Chen has spent thirty-two years teaching fourth grade in San Bernardino. She retired last June with a comfortable pension from CalSTRS, California’s teacher retirement system. What she discovered in her first quarterly statement might have surprised her: among her largest investments was a $5.26 billion stake in Nvidia, the chip company whose processors power the artificial intelligence revolution that may, in time, transform or eliminate much of what we call teaching.
Chen’s situation reflects a broader American paradox. Across the country, millions of public employees have become unwitting venture capitalists in technologies that could displace them. Their pension funds, designed as bastions of prudent long-term investing, now ride the most speculative wave in modern capitalism. Teachers fund companies developing AI tutors. State workers finance algorithms that automate government services. Police officers invest in surveillance technologies that may redefine law enforcement itself.
This transformation happened quietly, through the mechanical processes of indexation and fiduciary duty rather than conscious choice. Because artificial intelligence companies now dominate stock market indices, pension portfolios carry heavier exposure to technology than during even the dot-com boom. To own the market is to own the future of artificial intelligence, whether that future serves pension beneficiaries or supplants them.
The numbers tell one story. CalPERS, the nation’s largest public pension plan, doubled its Nvidia holdings in late 2023, finishing the year with 7.5 million shares worth $5.4 billion. Texas Teachers Retirement System built positions across the AI ecosystem, from chip manufacturers to cloud providers to venture funds backing artificial intelligence startups. The Canadian Pension Plan Investment Board committed $225 million to hyperscale data centers. Norway’s sovereign wealth fund, managing over a trillion dollars, began automating its own investment processes and suggested that AI efficiency gains might reduce the need for human staff.
Yet behind these financial maneuvers lies a deeper question about the social contract we have constructed around retirement security. Pension funds emerged in the twentieth century as instruments of collective solidarity. Workers contributed portions of their wages to pools that would support them in old age, managed by trustees bound to act in their collective interest. The system worked because economic growth generally created more jobs, better jobs, higher wages. Technology displaced some workers but created opportunities for others. Retirement savings could grow alongside an expanding economy that needed more human labor, not less.
Artificial intelligence challenges this assumption. Unlike previous technological revolutions, AI’s promise lies not in augmenting human capabilities but in replacing them entirely. The same pattern recognition that makes large language models valuable for customer service also makes human call center operators obsolete. The computer vision that powers autonomous vehicles threatens not just taxi drivers but entire transportation industries. The reasoning capabilities that excite investors may ultimately reduce demand for many forms of white-collar work.
Some pension trustees have begun grappling with these contradictions. Railpen, which manages retirement funds for British railway workers, published research on AI’s social risks and pledged to hold portfolio companies accountable for ethical governance. CalSTRS established internal AI policies and sent trustees to what it called “AI bootcamps” to better understand their investments. These efforts reflect growing awareness that fiduciary duty may require more than maximizing returns. It may require ensuring that the economy their investments create can still support their beneficiaries.
The historical parallel most often cited is the dot-com bubble of 2000, when pension funds watched technology investments collapse and learned painful lessons about diversification and liquidity management. But this comparison misses crucial differences. The companies driving today’s AI boom are profitable giants with established business models, not speculative startups burning venture capital. Microsoft, Google, Amazon, and Nvidia generate enormous cash flows from existing operations even as they invest in artificial intelligence research.
More importantly, the dot-com crash hurt pension funds but did not threaten the fundamental premise of retirement security. Workers could still expect decades of employment before claiming their benefits. Today’s AI revolution raises more existential questions. What happens to pension systems when the economy needs fewer workers? How do we fund retirement in a world where human labor commands diminishing value?
These questions extend beyond individual funds to the architecture of modern capitalism itself. Stock markets reward companies for reducing labor costs and increasing efficiency. Pension funds, as major institutional investors, have helped drive this process by demanding higher returns and supporting management teams that deliver them. The irony is that public sector workers, through their pension contributions, have become some of capitalism’s most patient and influential advocates for automation.
The moral complexity deepens when we consider alternatives. Pension trustees who avoid AI investments risk failing their fiduciary duties, potentially leaving retirees with inadequate resources. Those who embrace these technologies may accelerate economic disruptions that undermine the social systems retirement security depends upon. There is no clear ethical path, only competing obligations that reflect deeper tensions in how we organize economic life.
Perhaps the most troubling aspect of this situation is how little conscious choice it involves. Margaret Chen never decided to invest in artificial intelligence. Her pension contributions flowed through institutional mechanisms that made those choices on her behalf, guided by fiduciary rules written for an earlier era. The teachers, firefighters, and state employees whose retirement security now depends on AI’s success had no voice in that dependency.
This points toward questions that extend beyond pension management to democratic governance itself. If artificial intelligence is reshaping society in fundamental ways, should the direction of that transformation be determined entirely by market forces and fiduciary calculations? Do workers whose livelihoods may be displaced deserve some influence over the pace and character of technological change their own savings help finance?
The answers will not come from pension trustees alone, however thoughtful their deliberations. They require broader social conversations about how we want technology to serve human flourishing rather than simply maximize returns. They demand new forms of economic organization that can distribute AI’s benefits more widely while supporting those it displaces.
For now, Margaret Chen’s retirement remains tied to Nvidia’s stock price and the uncertain trajectory of artificial intelligence development. Her situation embodies the broader challenge we face: how to navigate technological change that may be both economically necessary and socially destabilizing. The choices we make will determine whether AI becomes a tool for shared prosperity or merely the latest mechanism for concentrating wealth and power.
The summer of 2025 may be remembered as the moment when this tension became impossible to ignore, when the quiet calculations of pension trustees became inseparable from the larger question of what kind of future we are building together.
