Earlier this spring, the federal government sold a 940,000-square-foot office building in southwest Washington for $25 a square foot. That’s less than the price of decent carpet. The buyer plans to turn it into apartments. Across the city, the same thing is happening: D.C.’s Housing in Downtown program expects to convert 6.7 million square feet of underused office space into 8,400 new homes. Nationwide, some 90,000 apartments are under construction in former office buildings, a figure that has nearly quadrupled since 2022.
This is being celebrated as a success story. And on one level, it is. But I spent a decade running a Federal Reserve Bank in a city full of anchor institutions, the universities and hospital systems that are typically a region’s largest employers and most stable economic engines, and I can tell you: We are not being honest about the math underneath these conversions, or about the larger forces bearing down on American cities.
Start with remote work. The debate about whether it’s permanent is over. About 28% of all U.S. working days are now remote. That number has barely moved in two years. Three-quarters of companies have settled into some version of hybrid. The firms pulling people back to the office tend to be older and shrinking. The ones expanding flexibility tend to be younger and growing. The headline rate looks stable because those two forces cancel out. But nobody should confuse that for a return to 2019.
What this means for cities is simple and uncomfortable: The commuter-economy tax base is permanently impaired. Converting offices to apartments is good urbanism, but residential properties generate a fraction of the tax revenue that commercial office space did. New York’s own Comptroller has published the numbers. On a single conversion project, the city is foregoing over half a billion dollars in tax revenue over 37 years to produce 338 affordable units. That amounts to $1.6 million in lost revenue per unit. I’m not saying it’s the wrong trade. I’m saying let’s not pretend it’s free.
Now layer on artificial intelligence, and the picture gets harder. Goldman Sachs calls AI the big labor market story of 2026, and the early evidence is specific enough to worry about. Workers under 25 in AI-exposed occupations have seen their share of employment decline since ChatGPT launched. Experienced workers in those same fields? Holding steady or gaining. The Dallas Fed has a useful framework for understanding why: AI is very good at replicating the kind of knowledge you can get from a textbook, and much less good at replicating the judgment that comes from years of doing a job. So it’s automating the entry-level tasks and augmenting the senior ones. Wages in high-experience, AI-exposed occupations are actually rising.
We are not being honest about the math underneath these conversions, or about the larger forces bearing down on American cities.
Think about what that means for cities. The economic logic of urban life has always been: Move here, pay high rents, and get on the career ladder. The density of experienced people around you taught you what no classroom ever could. If AI hollows out the bottom rungs of that ladder, the whole bargain starts to unravel. You’re not just losing jobs. You’re losing the reason people put up with the cost of being there.
Learning From the Past
We have been here before. The last time a technological revolution reshaped the relationship between work and the physical city was the automobile, and we made a series of decisions we are still paying for. Urban renewal, redlining, highway-driven displacement, the hollowing out of downtowns. None of that was inevitable. It happened because the people making decisions weren’t thinking hard enough about who would get hurt. The question now is whether we repeat those mistakes with remote work and AI, or use the assets we already have.
I don’t have to theorize about this. I watched it for a decade from the Philadelphia Fed, and I’m watching it now. John Fry arrived at Temple University and did something that too few university presidents do: He said out loud that Temple’s future and North Philadelphia’s future are the same thing. He’s planning a research triangle along Broad Street, partnering with local K-8 schools, and running a $1.5 billion campaign that treats the university’s surrounding community not as a problem to manage but as an economy to build. That’s what an anchor institution looks like when it takes the job seriously.
That experience shaped how I think about what cities actually have going for them. The biggest, most stable employers in most American cities are not tech firms or banks. They’re universities and hospital systems. They don’t relocate. They don’t go remote. Their employees spend money locally, their students stay in the region, and they’re embedded in exactly the places that need economic restructuring most. At the Philadelphia Fed, we built an entire research initiative around this idea, and the data bore it out: Cities with strong anchor institution networks weathered the pandemic measurably better than those without.
The biggest, most stable employers in most American cities are not tech firms or banks. They’re universities and hospital systems.
Investing in Humans
So, what do we do? Mark Hagerott, Ramayya Krishnan, and I have proposed what we call a Digital-AI Land Grant Act. The concept is simple. In 1862, Lincoln signed the Morrill Act because an industrializing economy needed a new kind of institution to train workers at scale. We’re at an analogous moment. Every state needs at least one institution with the mission and the money to provide AI-era workforce training, and not as a side project. Fund it the way we should have been funding it all along: Require the tech companies building data centers in these communities to invest in the human capital around them, not just the server racks.
This isn’t hypothetical. When I was president of the University of Delaware, we built exactly this kind of partnership with JPMorgan Chase: an on-campus innovation center where students and bank technologists worked side by side on real problems, feeding a pipeline of technology talent into Delaware’s largest private employer. In the Philly Fed’s district, I watched Saint Francis University in Loretto, Pennsylvania, do the same thing with Curry Supply, a local manufacturer: They built a $7 million innovation lab in downtown Altoona where engineering students solve real production problems, from robotic painting arms to snowmaking systems, and the region gets a workforce it couldn’t grow any other way.
There are $213 billion in office loans coming due by 2027. The federal budget is getting tighter, not looser. Cities that wait for Washington to solve this will be disappointed. But the anchor institutions are already there, already invested, and operating at a scale that matters. Remote work and AI are going to keep reshaping American cities whether we’re ready or not. We can use the tools we have, or we can spend another generation fighting the last war.
