The capital flowing into artificial intelligence has reached a scale that makes previous technology cycles look modest. CB Insights put US private AI funding at $206 billion in the first quarter of 2026 alone, approaching what the entire prior year produced. That money is going into model development, data center capacity, chips, and the engineering talent required to build any of it.

Hiring engineers means putting them somewhere, and the companies doing the hiring have reached a conclusion about where that should be. Nick Romito, co-founder and CEO of VTS, has watched the posture of tech tenants reverse over the past few years. “Post COVID the tech companies were pushing back against the return to the office,” he said. “Now they are driving the return, a lot of companies are even bringing people in on Saturdays because they know they can’t win this race to innovate if they are remote.”

VTS tracks live office requirements across 17 markets, and its data shows national AI office demand up 85 percent year over year, with growth of 179 percent across the largest AI hubs. AI companies now account for 46 percent of the square footage tech firms are actively pursuing. The requirements themselves are larger than typical tech deals, averaging 37,000 square feet against 27,000 for tech overall.

That demand is not spread evenly. San Francisco, Silicon Valley, and New York together hold roughly two-thirds of all active AI square footage, and a single San Francisco submarket covering SoMa and Mission Bay accounts for a quarter of national AI demand on its own. San Francisco’s 81 active AI requirements average 62,000 square feet, more than double the average tech requirement across the markets VTS follows.

The composition differs by market. San Francisco skews toward foundation model labs, while New York draws application-layer companies and AI infrastructure firms selling into its financial, legal, and media base. New York recently moved past Silicon Valley into second place for active AI demand, with 2.79 million square feet against Silicon Valley’s 2.78 million, and it now has more requirements over 50,000 square feet than Silicon Valley does.

Secondary markets are starting to register. Seattle posted the fastest growth of any market VTS tracks at 390 percent year over year, and Northern Virginia grew 169 percent on defense and government technology work. Los Angeles and Boston are the only markets moving the other direction.

What makes the current cycle different is how quickly it has compressed. After the financial crisis, San Francisco anchored tech demand in 2010, New York followed by 2013, and Seattle entered by 2016. This time those markets activated close to simultaneously.

The broader expectation about AI and office space ran in the opposite direction. If AI automates enough knowledge work, companies need fewer people, and fewer people need less space. That logic shaped a good deal of commentary about where office demand was headed, and it has not yet shown up in leasing data.

Ryan Masiello, co-founder and chief strategy officer at VTS, thinks the reason has to do with the limits of distributed work for the kind of problem-solving these companies are doing. “Unless you are a company that started as a remote company, it is really hard, if not impossible, to have the same productivity over Slack,” he said.

The broader market numbers point the same way. New York became the strongest office market in the country in the second quarter, supported by a 113 percent year-over-year increase in technology demand. San Francisco ranked second with trailing twelve-month tech demand up 161 percent. Excluding San Francisco, which was coming off a record first quarter, technology demand rose 19 percent quarter over quarter.

There is an early signal that the effect is spreading beyond AI tenants themselves. Professional services requirement velocity in San Francisco is up 33 percent year over year, which follows the pattern from prior cycles where tech leasing led and other industries followed into the same submarkets.

Whether that holds depends on things nobody can forecast with confidence. AI funding could contract the way tech funding did in 2022, when companies that had hired ahead of revenue cut staff quickly. The productivity effects that were supposed to reduce headcount may simply be arriving on a longer timeline than the leasing data covers. And the concentration in a handful of submarkets means a landlord underwriting to metro averages is reading a market that does not exist at their property.

For now, the companies building this technology are leasing more space than their predecessors did at the same stage, and they are doing it while telling their employees to come in on weekends.

The post AI Was Supposed to Reduce Office Demand. So Far It Has Done the Opposite. appeared first on Propmodo.

​