The Return of the Office for AI Teams
Why AI Startups Are Quietly Killing Remote Work
For five years the story was simple: remote work killed the office. In 2026 that story is reversing, and the companies leading the reversal are the same ones everyone assumed would be the most distributed of all. AI startups are choosing to be in the same room, on purpose, every day.
If you raised a round or started a company between 2020 and 2024, "remote-first" wasn't a perk. It was the default. Hire anywhere, work from anywhere, skip the lease, put the savings into headcount. For a while it looked permanent. Office towers in San Francisco and Seattle emptied out, sublease space piled up, and the prevailing wisdom held that the physical office was a relic that the smartest, most technical teams would be the first to abandon.
That assumption turned out to be exactly backwards. The most technical, fastest-moving teams in the world right now, early-stage AI companies, are the ones racing back into physical space. And they're not doing it because a CEO issued a mandate. They're doing it because building in a fast-moving field is a contact sport, and they've decided proximity is a competitive advantage they can't afford to give up.
The Reversal Nobody Predicted
The shift away from remote-first is real, and it's measurable. Roughly a third of organizations now plan to reduce or eliminate remote work heading into 2026, and the fully-distributed startup is quietly going out of style. But the headline isn't that big companies are forcing people back. It's that the teams with the most freedom to stay remote are voluntarily giving it up.
Early-stage AI companies could work from anywhere. Their talent is in demand, their tooling is cloud-native, and most of their founders cut their teeth in the remote era. They have every excuse to stay distributed. Instead, they're signing leases. AI companies leased millions of square feet across the major tech markets in 2025, and in the first quarter of 2026 they accounted for nearly a quarter of all office demand in the entire country. That is not a hybrid-work compromise. That is a category of company deciding the office is where the work happens.
"The teams with the most freedom to stay remote are the ones racing back into physical space, not because they were told to, but because they figured out that being in the same room is an edge."
Why AI Teams Specifically Need to Be Together
Not every company needs everyone in the building. A mature business with settled processes can run on a hybrid week and barely notice. But an early-stage AI company is the opposite of settled. The field moves weekly. A model that was state of the art in March is a baseline by June. The product roadmap is rewritten on a whiteboard, erased, and rewritten again. In that environment, the cost of latency, the lag between someone having an idea and the whole team acting on it, is enormous.
In person, that lag collapses to zero. An engineer overhears a researcher describe a problem and solves it before lunch. A founder sketches an architecture on glass and three people pressure-test it on the spot. Decisions that would take a day of asynchronous back-and-forth happen in the ten minutes between meetings. The intensity that defines a good AI team, the rapid iteration, the shared context, the willingness to throw out yesterday's plan, is hard to manufacture over video calls and harder still to sustain.
There's also the simple matter of judgment. Building at the frontier means making dozens of small calls a day that no tool can make for you, and the quality of those calls goes up sharply when the right people are physically present to weigh in. That's why AI-native teams treat the office not as a place to be seen, but as the medium they build in.
What That Means for San Francisco and Seattle
This behavior is the real engine behind the office recovery in the two markets where UnCommon operates. San Francisco is now the only major U.S. gateway market to eclipse its pre-pandemic office demand, and AI is the reason. The city recorded the highest number of AI-related leases of any city on earth in 2025, and 2026 leasing is projected to climb 15% to roughly 12.8 million square feet. The marquee deals make the trend concrete: Anthropic took 420,000 square feet at 300 Howard, and OpenAI now occupies close to a million square feet across Mission Bay.
Seattle is on the same trajectory. Puget Sound leasing reached 2.1 million square feet in Q1 2026, up from 1.8 million a year earlier, with the Bellevue CBD turning into a genuine AI cluster. OpenAI alone signed for nearly 250,000 square feet there. The recovery isn't evenly spread, though. Both markets still carry plenty of empty commodity space; what's scarce is the right kind of space for a team that's growing fast and doesn't yet know exactly how big it will be next quarter.
That gap, between abundant generic square footage and the small amount of space that actually fits a fast-growing AI team, is precisely where UnCommon sits. UnCommon operates historic, character-rich buildings in both cities, run as communities rather than desk inventory, by an owner that has a stake in whether the businesses inside them grow.
The UnCommon Fit: Room to Grow, Without the Friction
Here's the bind a five-person AI startup is in. They know they need to be together. They also have no idea whether they'll be five people or thirty-five by the end of the year. A traditional lease forces them to bet: sign for too little and you're renegotiating in six months; sign for too much and you're paying for empty desks while you're still pre-revenue. Either way, the lease becomes a drag on exactly the speed that made the team want to be in person in the first place.
The whole point of UnCommon is to take that bet off the table. A team can start with five desks and scale to thirty without the friction of a new lease, a build-out, or a personal guarantee. The space flexes as the company does. Think of it the way the best incubators think about the teams they nurture: give ambitious people a real place to grow, surround them with operators and builders solving similar problems, and let the environment do some of the work. The difference is that UnCommon's buildings come with a landlord that genuinely cares whether you make it, because UnCommon is operated by Rubicon Point Partners, the firm that owns the buildings, not a coworking middleman flipping memberships.
The Rotunda
A 1913 Beaux-Arts landmark, fully restored and operating as UnCommon's flagship. The ecosystem here leans hard toward tech founders and AI builders, with private offices, coworking, and event venues alongside a podcast studio, wellness studio with cold plunge and sauna, conference rooms, the UnCommon Lounge, and Awaken Cold Brew Coffee. It's the kind of building where a small team can plant a flag and grow into the space, and where the staff know members by name and by business model.
Explore The Rotunda →
Masin Block
In a Seattle market where AI demand is real but the right kind of space is scarce, Masin Block offers something a sublease never will: an environment built for teams that actually ship. Light-filled coworking floors, glass-walled meeting rooms, and a setting that signals seriousness without stuffiness. The building survived the Great Seattle Fire of 1889, and it's run as a place where things get built, not where people show up to look busy.
Explore Masin Block →
200 Pine
In the market leading the entire national office recovery, 200 Pine puts a growing team steps from Montgomery Street BART in the Financial District. The building offers traditional offices for companies that need a serious San Francisco address, and because RPP understands the value of investing in community, one floor is being transformed into UnCommon Coworking, intentionally curated so the right founders, operators, and AI teams end up in the same building.
Explore 200 Pine →"Most landlords think about vacancy rates. We think about whether the businesses in our buildings are actually growing. AI teams want to be in the room together, so our job is to give them a room worth being in, and the room to grow when they do. If they're growing, everything else follows."
Razmig Boladian, Rubicon Point Partners
Why It Matters Right Now
If you're a founder or ops lead searching for space today, the timing matters. The reversal is still early enough that the best buildings in San Francisco and Seattle aren't yet picked over, but AI demand is moving fast, and the gap between generic vacant square footage and space that actually fits a fast-growing team is widening. The teams that win the recovery won't be the ones who found the cheapest desks. They'll be the ones who landed in a building designed for the way they actually work.
The remote-first era taught a real lesson about flexibility. The in-person reversal is teaching a second one: flexibility and presence aren't opposites. The right space gives you both, everyone in the room when it counts, and the freedom to grow without a five-year bet getting in the way.
AI startups didn't kill remote work with a mandate, they out-competed it. Being in the same room turned out to be an edge, and the office came roaring back in the exact markets where UnCommon operates. The smart move for a growing team isn't the cheapest lease or the most generic tower. It's flexible space, in a serious building, run by an owner who actually wants you to grow.
Grow from 5 desks to 30, without the lease that slows you down.
Tour UnCommon's flexible coworking and office space in Oakland, Seattle, and San Francisco, built for AI teams that need to be in the room together, and operated by the people who own the buildings.
