The AI boom is easy to measure in dollars raised and models shipped. It is harder to see where it physically lands. A new report from the real-estate platform VTS offers one answer, and it is striking.
AI companies now account for nearly half of all the office space that America’s tech firms are chasing.
VTS tracks live, forward-looking leasing pipelines rather than deals already signed, which makes its data an early signal. Across the 17 US markets it follows, AI tenants make up 34% of active tech requirements, but 46% of the square footage. National AI office demand is up 85% year over year, and up 179% in the biggest AI hubs.
One caveat worth stating up front: VTS sells software to landlords, so this is not a neutral academic study. But the platform handles more than 60% of Class A office space in the US, so the pipeline it sees is real.
Where the money lands
The 💜 of EU tech
The latest rumblings from the EU tech scene, a story from our wise ol' founder Boris, and some questionable AI art. It's free, every week, in your inbox. Sign up now!
The capital behind this is staggering. VTS points to the first quarter of 2026 as the largest AI funding quarter in history, with global private investment of $226 billion. OpenAI alone raised $122 billion, at a valuation of around $850 billion. Anthropic and xAI added tens of billions more. That money is now turning into desks.
AI tenants are also taking bigger spaces. The average AI requirement is about 37,000 square feet, some 37% larger than the typical tech deal. The enterprise tier of leases above 50,000 square feet is back, after three years of shrinking tech renewals.
AI firms are committing to serious footprints early, before most have turned a profit.
Three cities, one street
The demand is not spreading out. It is piling up. San Francisco, Silicon Valley, and New York together hold 63% of all active AI square footage. San Francisco alone accounts for 5 million square feet, nearly a third of the national total.
Zoom in further and the concentration gets almost comic. A single San Francisco submarket, the SoMa and Mission Bay corridor, holds a quarter of all active AI office demand in the entire country. This is where the foundation-model labs cluster.
The report notes Sierra, the agentic-AI startup co-founded by OpenAI chairman Bret Taylor, is set to take about 300,000 square feet at one building, while OpenAI and Anthropic keep expanding nearby.
New York tells a different story. Its tenants lean towards application-layer and infrastructure firms serving banks and law firms. The standout deal came from Nscale, an AI cloud company, which signed for space at One Vanderbilt at $320 per square foot.
VTS says that is the highest office rent Manhattan has ever seen. Silicon Valley, meanwhile, draws the chip and hardware firms that want to sit near the semiconductor supply chain.
The new map of AI
Beyond the big three, a fresh geography is forming. Seattle is the fastest-growing market VTS tracks, with AI demand up 390% in a year, drawn by the city’s engineering talent.
Northern Virginia is up 169%, on the back of defence and government work. Austin, Chicago, and Atlanta are all climbing, on the strength of enterprise adoption rather than AI-native labs.
Washington, DC is its own case. In several of its submarkets, AI accounts for more than 80% of all active tech demand. That is not chatbots. It is defence and intelligence work, from firms like Anduril, Shield AI, and Palantir, and it answers to the Pentagon budget rather than to venture capital.
Two big names sit out the story entirely: Los Angeles and Boston, where AI demand is actually falling.
Is this 2022 all over again?
The obvious worry is a repeat of the last tech bust. In 2022, companies over-hired ahead of revenue, then dumped staff and office space when funding dried up. VTS argues this time is different. Today’s leading AI tenants are scaling revenue faster, from a higher base, and many have committed to build-outs that are years from finished.
A funding slowdown would cool growth, the firm says, but is unlikely to trigger a 2022-style collapse.
That case is reasonable, but it deserves scepticism, and this is where the bubble question bites. Demand this concentrated is also this exposed. If the capital that props up a handful of labs slows, the streets carrying a quarter of national demand feel it first. The broader office recovery remains patchy, and the wave leans heavily on a few names raising sums with little profit beneath them.
Why it matters
The real value of the report is what it says about hidden signals. Citywide vacancy figures still look soft, which tempts landlords to price space to the average. VTS’s point is that the average is a mirage. In a few corridors, concentrated AI demand is quietly tightening supply, while most of the market sees nothing.
There is a forward signal too. In past cycles, law and consulting firms followed the tech industry into new space with roughly a year’s lag. In San Francisco, professional-services demand is already up 33%, an early hint that the spillover is starting. Legal demand, by contrast, is down sharply, which may reflect AI eating into that work in real time.
Where AI plants its desks, in other words, is a map of where it thinks it is going, and it draws that map a year before the rest of the economy catches up. Right now, that map points at a very small number of streets.
Get the TNW newsletter
Get the most important tech news in your inbox each week.