AI Is Filling Office Towers. It's Also Likely To Empty Them

The tech sector is driving a curious rebound in office leasing that's far from widespread, instead mostly playing out in a handful of the highest-quality properties in just a few U.S. markets.

While firms of all stripes — especially companies focused on artificial intelligence — sign new deals to pile into the best buildings, the tech giants are cutting headcounts, potentially presaging a white-collar job landscape primed for layoffs. The seeming paradox reflects the fast-evolving world of AI at work that is set to define the post-pandemic positioning of the U.S. office market. 

“In the office market, what we're seeing is a scramble to implement AI at the business level, and that's disrupting the way we see the office recovery,” said Erin Patterson, who co-leads research and strategy at Manulife. “I actually hesitate to call it a recovery just yet, but I think we are starting to see some normalization. The problem with the office market is the supply side.”

Tech firms leased 11.5M SF of office space in the first quarter, a decade-high 23% of all U.S. leasing activity, according to CBRE, as the companies race to develop the next market-making artificial intelligence tools.  

But on the other side of the AI revolution are the 116,854 tech sector layoffs tracked through the first week of June by online aggregator Layoffs.fyi, surpassing the 104,266 job losses the site logged across all of 2025.

Tens or hundreds of millions of square feet of effectively obsolete office space dot the country, unlikely to ever find a tenant in the post-pandemic marketplace.

Total U.S. office vacancy has hovered around 20% for years, even as the highest-quality properties continue to pull in new tenants and start to reach capacity. But AI adoption is expected to keep office-using job growth relatively flat, and occupiers have broadly been downsizing in recent years and trading for better-appointed offices.

“We're in a low-hire, low-fire environment. Businesses aren't really laying off, but if they're spending money, they're probably spending money on AI instead of hiring a person,” Patterson said. 

Giants in the sector like OpenAI and well-funded startups are acting as a counterbalance to downsizing across the office sector. Tech firms signed 36.7M SF worth of leases in 2025, nearly 17% of all leasing volume, according to CBRE.

But the activity is highly concentrated in San Francisco, Silicon Valley and Manhattan, where tenants continue to cluster, while other markets like Miami pull in outliers. AI firms leased 1M SF of office space in Manhattan in the first three months of the year, eclipsing their total for all of the prior year. In San Francisco, the 3.5M SF of Q1 deals from AI firms was 60% of the 2025 total. 

But on the other side of the rapid expansion is the push, mostly from public companies, to realize the promised efficiencies of AI. Oracle, Amazon, Intel, Microsoft and Meta have cut a combined 118,483 jobs since the start of 2025, or 57% of tech cuts tracked by Layoffs.fyi. 

Bank executives have been wooed by tech demos and are preparing to reshape their workforce around AI, with far fewer junior staff, Bloomberg reported this month.

Amazon is planning to cut its office footprint this year by 49,000 desks, the equivalent of more than 14M SF, even as it leases space in key markets, according to leaked audio from an internal meeting earlier this year. 

The U.S. added 172,000 jobs in May, according to government figures released Friday. But 93% of the new roles were in healthcare, hospitality or local government. Meanwhile, the financial sector shed 22,000 jobs in May and has lost 107,000 from its peak a year earlier.

More layoffs are on the horizon, with 99% of the 825 executives and 1,650 human resource leaders responding to consulting firm Mercer’s Global Talent Trends report saying they expected to cut jobs due to technology sometime in the next two years. 

The layoffs from corporate giants allowed them to redirect capital toward AI investments as they compete with one another to quickly build out the data center infrastructure needed for their promised AI solutions, said Colin Yasukochi, executive director of CBRE’s Tech Insights Center.

“They're obviously spending a ton on AI infrastructure, and so they're probably having to make some of these cuts in advance of the real efficiency gains,” Yasukochi said.

Some firms are plowing ahead with detailed road maps in place, but other companies are taking a different approach.

“It's more like, ‘Do you know how to swim? No? Well, I'm going to throw you in the pool. You're going to figure it out,’” he said.

Generative AI, like image and video creation, is novel, but agentic AI is a more natural fit for business applications, he said. Setting AI loose to execute digital plans in real-world settings is the next step in AI’s development and the first major glimpse of the tech’s transformative potential.

Early AI adoption is likely to lead to more destruction than job creation, but many believers in AI expect the tech will quickly create new businesses and a yet-undefined segment of jobs needed to support the burgeoning world that the technology will create.

What those jobs will be isn’t entirely clear — it will become apparent as the tech matures and new business cases evolve, AI’s boosters say — but Patterson expects the first new roles will be focused on compliance, regulation and data security inside AI tools.

The focus on efficiency and optimization today is really a reflection of the frothy macroeconomic climate where uncertainty and high interest rates drive decisions, Patterson said. 

“Right now, where we're faced with this volatile period where AI is coming into play, that's the best way to sell AI,” she said. “If I'm an AI saleswoman, I'm absolutely going to say this can create efficiencies, which a business manager knows means they can cost-save.”

The headcounts at the major AI firms are also poised to explode as they capture ever more market share, creating the need not just for more developers of AI tools but also the entire ecosystem of corporate hierarchies that come with large companies. 

Sam Chandan, a global real estate finance professor at New York University, expects the shift to job creation will be relatively quick.

“There are real downside risks, but I think the next couple of years will be a period where we actually do start to see some of the meaningful gains in productivity and new business opportunities opened up from these tools,” he said. “That's going to be key. If that takes a long time, then we're going to have a problem.”

The corporate world is shifting from the experimentation phase with AI toward defined policies around the tools employees can use and how they should be implementing AI into their work, Chandan said. 

Management is looking for clear evidence of productivity and efficiency gains from AI but running into cost concerns as they face massive expenses related to their employees using the tech. 

At Nvidia, the chipmaker that has had a meteoric rise along with AI, the team focused on applied deep learning is spending more money on computing power than it is on employees, Axios reported. Uber put caps on the usage of some AI tools by its staff this month to save on costs, giving each worker a $1,500 monthly token budget, Bloomberg reported. 

OpenAI CEO Sam Altman said spending concerns were “the most fair” criticism of AI in a CNBC interview this month, but he offered little in terms of a solution, saying only that he assumed “the industry will figure that out pretty quickly.” 

The cost of computing power makes human employees more economical than AI, but the massive push to build out infrastructure should pull costs down. Once the math makes sense, layoffs are likely to accelerate, said Stijn Van Nieuwerburgh, a Columbia Business School economist. 

“Companies can't really do all the work that they want to do with AI yet because there's simply not enough compute to do it, which means that they're not in a position to lay off a substantial fraction of the workforce,” Van Nieuwerburgh said. 

The big AI firms are racing to meet that demand, in part by signing sprawling office leases. OpenAI has leased more than 1M SF of office space in California and Washington since the start of the year across five deals, while Anthropic has taken roughly 1M SF across California since September. 

Just over one-fifth of office space nationally is vacant, but access to the best buildings in the tightest markets is a difficult challenge, with most at or near capacity. That has led some leading firms to sign leases for office space they don’t yet need, operating under the assumption that they will grow into it, a practice that in the past has signaled overexuberance. 

“One of the features of this that is reminiscent of the dot-com boom is that, given the rapid growth, we are seeing some of the AI firms leasing in anticipation of what their needs will be several years out,” Chandan said. 

Despite the demand for the best properties and the willingness to pay top dollar to move in, the reality is that there has been a fundamental shift in office usage that has left the country with too much space, Van Nieuwerburgh said.

“We have a structural surplus in office, and the only way to fix it is to get rid of it — and that takes the form of part demolition,” he said. 

Office-to-residential conversions offer some alternative to redevelopment, but it typically starts with a reckoning for the landlord.  

“The only thing that people forget is, for that conversion to take place, that old office must have lost most of its value,” Van Nieuwerburgh said. “Otherwise, the conversion doesn't make economic sense. Somebody's losing their shirt.”