The AI boom is a totally different tech boom than anything the city has seen before.
That was what Ted Egan, San Francisco’s chief economist for the last 19 years, said at a Wednesday night panel about the city’s economy, moderated by Io Yeh Gilman, a reporter at Mission Local.
Over the past three years, 60 percent of U.S. venture capital investment in AI went to companies in San Francisco, said Egan. Last year, that was close to $190 billion in venture capital, distributed among 2500 AI startups.
This investment, however, carries a totally different economic impact than the dot-com boom of the late 1990s, or the surge in local tech investment that followed the 2008 mortgage crisis, he said.
The AI boom is adding much more money to San Francisco’s economy, said Egan, but its effects are different.
During the dot-com boom and the pre-pandemic tech boom, companies spent freely on office space. In the past three years, many of those leases have expired, resulting in a net loss of about 7.5 million square feet of occupied office space across the city.
While the tech boom of the 2010s created new jobs, the AI boom appears to be jettisoning them. The city lost more than 30,000 tech jobs in the past three years — accounting for nearly all of the job losses reported citywide.
In 2018, tech companies based in San Francisco paid almost half of their payroll expenses here, according to their business tax filings, Egan said. But, between 2021 and 2024, that share has declined from 44 percent of payroll to 11 percent.
“One of the things I’ve kind of struggled with is, why is the apartment market so hot when the labor market is so cold?” he pondered. “It does appear that, whether people are hiring in San Francisco or not, tech workers are moving back to San Francisco after several years of saying, ‘I’m going to Miami or Austin or wherever.’”
Still, Egan said, the startups and the money involved are reasons for long-term optimism about the city’s economy. “I do think a certain corner has been turned,” he said. “The AI thing could be a bubble that blows up tomorrow, but I think that as long as that money is there, the worst of it is behind us.”
Still, panelists made it clear that there is an abundance of bad ahead to be dealt with. Seated next to Egan was his boss, San Francisco Controller Greg Wagner, who was sworn in for a 10-year term as the Controller in February 2024. Wagner’s office was responsible for calculating the $643 million two-year deficit that San Francisco now faces.
“How painful will fixing this structural deficit be?” Gilman asked Wagner.
“The elusive answer is it’s going to be up to the mayor and the board,” said Wagner. Labor is the single biggest cost the city incurs, Wagner said. “About half of the city’s budget is for people.”
And there are more jobs buried elsewhere in the budget.
“The second biggest piece is on grants and contracts,” Wagner added. These contracts often pay the salaries for people who work for organizations that contract with the city.
Earlier this month, Mayor Daniel Lurie announced plans to cut city spending by $400 million this year to reduce the deficit. He plans for $100 million in cuts to come from eliminating positions and layoffs.
His office recently gave 127 city employees layoff notices as part of that plan, though labor unions have pledged to fight every one of the proposed layoffs.
Usually, Wagner said, the solution to a budget shortfall is to try to squeeze a little more efficiency out of the government, rather than cutting the services provided by city employees and contractors.
However, “the size of this deficit is beyond squeezing efficiency,” he explained. “It’s big enough that it really does mean some conversations about service levels, the number of positions that are in the city government.”
“How much control does the controller have?” asked Gilman.
Not as much as you might think, said Wagner. A controller’s job is to prevent misuse or misappropriation of city funds.
“There’s a lot that feels completely outside of my control,” he said. Wagner’s office can calculate the deficit, but it has no power over the city budget, which is controlled to a large degree by the Mayor’s Office, and to a lesser degree by the Board of Supervisors.
The city’s economy is “always in a state of transition,” said Egan.
The dot-com boom brought an influx of money and new residents into the city, but did little to disrupt 30 years of stagnant job growth that had been underway since the 80s. Post-2010, San Francisco became one of the two fastest growing cities in the United States — “entirely unexpected,” Egan said.
During the COVID-19 pandemic, the city lost more population than any other city, and experienced the biggest rent drop. Now, the city’s population is on the rise again.
That’s a good thing, said Wagner.
“Our financial recovery is based on people coming to San Francisco and working in San Francisco,” he said. If the two upcoming ballot measures to fund local and regional public transit systems don’t pass, he said, that could impact the city’s recovery.
“It’s a little bit hard to square significant reductions to the public-transit system with an economic recovery that works for the city’s budget.”
Some have expressed hope that AI could be used to increase efficiency in city government, and reduce the impacts of the budget shortfall on city services.
There’s no “off-the-shelf solution” to do that without considerable trial and error, Egan said. AI’s tendency to hallucinate makes it problematic for use in contexts like building code or critical services.
“There is a lot of conversation about how to use AI in the city,” said Wagner. “City Administrator Carmen Chu is really pushing us to do some structured thinking about how to use AI.” City governments tend to be late adopters of technologies, he said.
If you’re an AI skeptic, Egan said, that doesn’t mean that you disapprove of AI. The models are just so new. In five years, we’ll be able to see what they can actually do.
Also, bureaucracy has a tendency to grow, regardless of the amount of work to be done.
“You would know better than me,” Egan said to Wagner. “Have we ever rolled out a big technology system that left us with less people than before?”
“Never seen it yet,” said Wagner.
Let’s hope the CEO law passes. We should be taxing them more instead of cutting services. What’s the point of being one of the most expensive places to live in the world if it’s crap to live here? Thanks Tech and AI. 🙁
We should have stuck with tourism.
To the editor:
Yujie Zhou’s coverage of the April 15 panel with Controller Greg Wagner and Chief Economist Ted Egan (“AI boom is city’s weirdest tech boom”) captures a conversation whose substance deserved harder pushback than it received, from either the moderator or the page.
The piece’s most genuinely interesting datum is buried in the middle: San Francisco–headquartered tech companies went from booking 44 percent of their payroll locally in 2018 to 11 percent by 2024. That collapse — capital routed through SF addresses while the actual labor is performed elsewhere — is the structural fact that explains Egan’s otherwise puzzled observation about a hot apartment market coexisting with a cold job market. It is the story. It goes underdeveloped in favor of quotes that sound weightier than they are.
Consider what your panelists actually said. Egan dismissed AI’s usefulness to city government by invoking “hallucination” and concluding that in five years we will be able to see what the models can actually do. This is the chief economist of the city at the global center of AI capital deployment — a fact his own remarks established moments earlier — proposing a five-year wait-and-see on the industry reshaping his tax base. The working professionals in his jurisdiction are not waiting. Building-code compliance, which he specifically cited as too risky, is in fact a bounded, text-heavy, rule-based domain where automated review is already in commercial use. The answer treats a 2023 talking point as a 2026 analysis.
Wagner, asked whether any city technology rollout has ever reduced headcount, answered that he has never seen it happen — and the room, and the reporter, received it as rueful civic humor. It is not humor. It is the controller of a city facing a $643 million deficit, on a panel where he has just conceded that the shortfall cannot be closed by efficiency gains, cheerfully confirming that every prior technology investment has failed to translate into the staffing reductions such investments were presumably sold as enabling. That is an indictment — of procurement, of political will, of labor settlements, or of some combination — and it deserved a follow-up question instead of a chuckle.
The throughline is a problem of listening. Wagner said something self-indicting and was received as folksy. Egan said something analytically empty and was received as measured. Your readers would be better served by reporting that treats the city’s senior fiscal officials as subjects of scrutiny rather than as sources of copy.
Requesting moratorium on term “weird” in headlines
“Some have expressed hopes that AI could be used to increase efficiency in city government”. Let’s start with corruption, nepotism, rent seeking, pension spiking, overtime abuse and other such “inefficiencies”. Obviously.
Further, in 2005 the City&County had 27k employees, 2025 it’s 34k, are we getting our money’s worth? I’m not seeing it. Considering how we’re being told how we remain short on nurses, teachers, cops, frontline workers in general, it stands to reason that the 7k growth is in middle management, no? In other words, ignoring all the handwringing and doomsaying that would come with it, if we went back to 2005 levels, we wouldn’t see a lick of a difference.
People don’t see 99% of the AI industry. You see a website that doesn’t have AI crap everywhere – that doesn’t mean it’s not using it. Server management, ticket management, codeing support, even managing backups and deployments. You’re looking at a monitor right now – that company is using AI to help manage supply chain, analize orders, etc…
People think it’s all ChatGPT, but that’s just the public view. It is more and more embedded in software and regular business work every day.
So, maybe it’s a bubble, like the dot-com “bubble” was, but when that started, there was practically zero internet compared to today. That “bubble” was in SF, but resulted in world-wide job creation, even years after the SF bubble ended. Look for the same now. The AI money coming to SF may not be as visible, but the jobs are coming. Right now the layoffs aren’t, “Oh, we have AI do all that now”, that’s just BS news click-bait for like 1% of these things. The reality (that I’ve seen among several clients) is that they’re laying people off while they bank money getting ready for AI development. The jobs are just moving from one department to another.
I hope Lurie will focus on getting these new companies to *hire* locally and get people who live here a good paycheck, and not focus on giving them massive tax breaks that never pay off, like we saw during the latter end of the dot-com.
“Over the past three years, 60 percent of U.S. venture capital investment in AI went to companies in San Francisco, said Egan. Last year, that was close to $190 billion in venture capital, distributed among 2500 AI startups.”
That is a stunning statistic. It is hard to imagine how bad the city’s fiscal situation and deficit would be without AI. The two big AI monsters, OpenAI and Anthropic, are both SF companies, and both will go public in 2026 probably with trillion dollar market caps.
San Francisco minus AI equals Detroit.