“It’s another truly amazing gold rush!” Marc Benioff posted on X in September 2023. The founder and CEO of Salesforce was celebrating San Francisco’s AI-fueled revival, touting a report that pegged demand for new office space in the city at nearly 1 million square feet. By February 2024, The Economist was declaring that “San Francisco staged a surprising comeback.”
It looked like quite a turnaround for a city whose epitaph had been written again and again since the pandemic. Just months before Benioff’s exclamatory post, Salesforce had reduced its office footprint, leaving the city’s tallest tower a costly emblem of urban decay. According to the “urban doom loop” hypothesis, reduced demand for office space would lead to a collapse in commercial real-estate values and, in turn, a decline in city revenues and services—which would then push even more businesses and workers out of the city. San Francisco, which famously experienced a major exodus of workers during the pandemic, was long considered the doom-loop poster child. If it could rebound from its struggles, then perhaps the rest of America’s cities would also avoid that fate.
But the comeback is not what it seems, and a doom loop is still possible. Historically, a booming economy has reliably translated into a booming commercial-real-estate sector. Now, however, San Francisco and other so-called superstar cities have entered a kind of Schrödinger’s economy, booming and busting at the same time. City leaders must come to terms with the fact that pre-pandemic office demand is never coming back, and plan accordingly.
By mid-2022, San Francisco was in trouble. Tens of thousands of people had moved out of the city, notable venture-capital investors had relocated to Miami, and multiple local tech companies—most notably Meta—had announced plans to embrace remote and flexible work permanently. The municipal budget deficit continued to expand.
Nationally, most cities were doing better, but the average vacancy rate was still inching into record territory. In October 2022, Bloomberg’s economic forecast put the odds of a recession at 100 percent, and the situation looked like it would only get worse. Barring some kind of deus ex machina, San Francisco and other cities seemed destined to continue spiraling downward.
And then God stepped out of the machine. In November 2022, San Francisco–based OpenAI launched ChatGPT and kicked off a new technology boom. In 2023 alone, investors poured nearly $30 billion into artificial-intelligence start-ups and billions more into AI-related public companies, many of which are based in and around San Francisco. Economic conditions across the country were equally surprising. The “inevitable” recession failed to materialize. By early 2024, the S&P 500 reached a new all-time high, unemployment remained low, and technology stocks reached a level of valuation (perhaps overvaluation) that exceeded the dot-com bubble. In many cities, including San Francisco, net migration flipped from negative to positive.
[Rogé Karma: Whatever happened to the urban doom loop?]
But something still wasn’t right. In the first quarter of this year, the national office-vacancy rate reached 20 percent, the highest level on record—even higher, slightly, than during the 2022 doldrums. In San Francisco, more than a third of all office space was vacant. In fact, shortly after Benioff’s celebratory X post, Salesforce again shrank its footprint, this time by 700,000 square feet. The AI boom was real, but so was the threat of urban doom. A similar dynamic has been playing out in cities across the country dependent on a variety of other industries.
This is unusual. For decades, office demand has been correlated with macroeconomic indicators, meaning that when the economy is strong, so is demand for commercial real estate. A model developed by the Commercial Real Estate Development Association (commonly and confusingly known by the acronym NAIOP) has done a pretty good job of predicting and explaining office demand based on GDP growth, corporate profits, employment, and other economic indicators since the early 1990s. But starting in 2022, that historic relationship broke down. As the economy emerged out of the pandemic, the model predicted that net office demand would increase by 43 million square feet. In reality, net demand was nearly 90 percent lower than expected and, by the following year, had turned negative, meaning more space was vacated than leased.
What explains the divergence? The obvious culprit is the rise of remote work.
Four years after the initial COVID-19 lockdowns, more than a quarter of all paid workdays are performed from home, according to an ongoing survey by the Stanford economics professor Nicholas Bloom and others. The main reason companies are reducing their office footprint is because they can. As more leases come up for renewal, vacancy continues to rise. Even without a recession, this trend is likely to endure as tenants continue to express a desire to cut down or let go of existing offices ahead of a wave of lease expirations in 2025 and 2026.
Within the academic community, there is some debate as to whether factors besides remote work, such as interest rates or recession expectations, are also to blame for persistently high vacancy rates. One thing is clear: Even if the economy continues to grow and unemployment remains low, high office vacancies will have an adverse impact on municipal budgets and residents’ quality of life. Lower crime, a rebound in tourism, and a slight increase in population won’t be enough to offset the loss of revenue from commercial property and business taxes because of lower rents and lower spending from regular commuters. Cities can diversify their tax base, but that would require changes to the physical environment that take years to materialize, plus direct investment and tax incentives. It would also necessitate a sense of urgency and determination that has been lacking in many cities—particularly in light of the recent “comeback.” Stijn Van Nieuwerburgh and Arpit Gupta, two of the authors of the original doom-loop paper, have recently updated their estimates based on the latest data and project that, despite some good news, “many American cities … will face significant tax revenue shortfalls in the years ahead.”
[Dror Poleg: The next crisis will start with empty office buildings]
Van Nieuwerburgh and Gupta’s latest assessment includes a new concern that was not part of the original thesis. Artificial-intelligence advances may reduce the number of office jobs and improve the quality of remote collaboration. Data from the epicenter of the AI revolution offers a preview. In 2003, the year in which Google first passed the $1 billion revenue mark, the company employed some 1,600 people. Last year, OpenAI required less than half that number of workers to exceed the same milestone. Over the past 18 months, Big Tech companies laid off tens of thousands of employees while growing their revenue and hiring fewer—but higher-paid—AI specialists. The “1 million square feet” sought by San Francisco’s AI companies sounds like a lot, but it is overshadowed by the city’s 30 million square feet of vacant office space and the specter of many more lease expirations in the coming years.
AI threatens to make the connection between economic activity and office demand, and thus between economic activity and city-budget health, even less linear and predictable. The possibility alone is enough to inject more uncertainty into labor and office markets that are already on edge. An economy in which most companies can predict their needs in advance and commit to long-term leases is not returning any time soon.
Ever since the pandemic, many landlords, mayors, and bosses have been going through what one might call “the five stages of office grief.” First, in 2020, there was denial that working from home would have any lasting impact. Then, in 2021, there was anger at employees who wouldn’t return, followed by bargaining on the exact number of days people would spend at the office. By 2022, depression had set in, and cities seemed ready to accept the need for radical change. Now, however, the country’s economic rebound provides new ammunition for those who wish to slide back into denial.
Our cities will be better served by embracing the transition to a world that is less centered around offices. That will require diversifying their economic base, streamlining the construction and conversion of new housing and mixed-use neighborhoods, enhancing public services, and doubling down on what makes urban life attractive in its own right—not just as an employment destination. And the effort must start with the recognition that, in good times and bad, the relationship between economic activity and office demand has changed forever.