Artificial intelligence startup Hightouch’s new office lease came together after a whirlwind tour of about 20 Manhattan buildings.
The company, a marketing firm whose AI platform automates its customers’ administrative tasks, needed space fast after closing $150 million in series D funding, which will partly go toward doubling its 400-person headcount over the next year. Chief Financial Officer Cherry Miao looked at the 18,800-square-foot space at 275 Seventh Avenue in Chelsea and submitted a letter of intent less than 36 hours later.
“We were literally the first people to walk through it when it came on the market, and as soon as I walked through it, I was like, ‘This is special. Let’s roll,’” Miao said.
That frenetic pace of leasing is playing out all over New York City and San Francisco as a wave of AI firms snap up space at a pace that has stunned brokers and landlords. The start-ups are rewriting the rules of office leasing and creating a new class of tenant with different demands, timelines and risk profiles.
The result is a leasing process removed from the norms of traditional office shopping. Young founders show up to building tours with their parents, run office floorplans through ChatGPT and carry carbon dioxide monitors to test the air quality. Deals that once took six months now happen in fewer than 30 days, as the office becomes the manifestation of sky-high aspirations, hiring goals and venture capital funding.
San Francisco is the center of gravity for AI ascendance by number of firms and capital invested, with New York City in second place.
In the first quarter of 2026, San Francisco saw a positive net absorption of 1.6 million square feet, with artificial intelligence companies accounting for nearly 40 percent of the deal volume and leading to the best post-pandemic quarter after a year of bests. JLL predicts that AI’s footprint inside the city will double by 2030 to 14 million square feet. New York City’s economy has more diversity, so AI’s leasing figures are smaller but growing. AI tenants accounted for more than a third of the tech sector’s demand in the first quarter of 2026 with 670,000 square feet leased, according to Colliers, a sharp jump from its 12 percent share in 2025.
“I think the biggest shock is just the amount of companies that are looking in both markets right now,” Liz Hart, president of leasing for North America at Newmark, said. “They’re really tapping into those two markets simultaneously. What you commonly hear as a theme is the big sprint that they’re doing right now. If you’re going to capture the talent, you’re going to make a big push to do it now.”
Fast and loose
The industry’s urgency has reshaped leasing timelines, with some deals moving from first tour to signed lease in a few weeks. Hightouch’s speedy LOI process, while extreme, is no longer unusual.
“The priority for a lot of these firms is speed and certainty — not chasing every last dollar and penny-pinching,” Adam Henick, co-founder of Current Real Estate Advisors, said. “They’d rather lock down the right space immediately than risk losing it, negotiating over what someone else may consider minor economics.”
The hurry has office brokers clearing their schedules.
“I was introduced to a company on Thursday at 4 p.m., they wanted to hop on a call at 7 p.m., and we’re going to go tour spaces [the next] Tuesday,” Will Cassriel, of JLL in San Francisco, said.
But speed is only half of the equation. Flexibility is the other. Because startups can either scale quickly or fail, some AI companies are opting for shorter lease terms or negotiating termination clauses into deals. Others are signing longer-term leases for more space than they need in the near term, with the idea of growing into the footprint over time.
“There’s usually that getting-to-know-each-other phase, and now things are moving very quickly. It’s like, ‘Okay, we’re about to get married.’ It feels like it’s happening overnight.”
Landlords, meanwhile, are adjusting to a tenant base that is cash-rich but unpredictable. While revenue may trail that of traditional corporate tenants, balance sheets are often well capitalized. That has led to some aggressive lease terms, including security deposits far above the market norm of four to six months’ rent.
“Landlords are asking for massive security deposits, sometimes three years, and the companies are putting it up,” said broker Ruth Colp-Haber of Wharton Property Advisors.
Landlords are being forced to size up AI tenants on an accelerated timeline, reviewing financials and learning about their businesses far faster than in a typical leasing process.
“There’s usually that getting-to-know-each-other phase, and now things are moving very quickly,” Henick said. “It’s like, ‘Okay, we’re about to get married.’ It feels like it’s happening overnight.”
The whole negotiation can be complicated by the technology of the moment. In New York and San Francisco, companies have insisted on lease terms suggested by AI platforms like ChatGPT.
“Sometimes there is this inherent sense from some founders that, ‘Not only do I know my business better than you but I know your business better than you and I’m going to show you,’” said Mike McCarthy, a broker with Transwestern who works with landlords in San Francisco. “You see that cockiness from some founders.”
Diva desires
AI companies often come with unconventional demands.
One broker said a tenant asked to replace all the windows in an office for aesthetic reasons, a request that was quickly rejected by the landlord. Two separate start-ups touring with Jack Doherty of Cresa in San Francisco turned down space available on high floors because “looking down on people” wasn’t a culture fit.
Once, a company taking 3,000 square feet in a 1 million-square-foot building sought building signage after apparently running its lease terms through an LLM (large language model), forcing its broker to explain that the request was unlikely to sit well with the landlord.
“That’s when you have to tell them that the one-year deal is the win,” said Jim Walker, a landlord-side broker with Kidder Matthews in San Francisco who explains that ownership might lose patience with too many negotiating points on a short-term lease.
Some tenants are turning to AI tools to redesign office layouts, according to Gabe Marans of Savills, but the technology can overlook details like city fire code requirements. That’s forcing some brokers to step in and explain what’s actually feasible.
Air quality is a priority.
“We did a study for a larger tenant we were working with that was worried about the air quality, and I’ve had two other much smaller tenants say, ‘Oh, I should have brought my CO2 reader to the tour,’” said Alyssa Zahler, managing director of commercial leasing at Two Trees Management.
The antics can start before anyone boards the elevator.
“If they can do everything through text, that’s their preference,” Jay Shaffer, a tenant-side broker with San Francisco firm Colton Commercial Partners, said. He’s now shooting off messages to coordinate with tenants early in the morning, late at night and over the weekends.
The in-person interactions are full of surprises. Brokers say young founders sometimes arrive with extended teams — or even parents or spouses — to evaluate space.
“I have been on a few tours with AI companies where they bring their parents, or they bring, like, their father, the real estate lawyer,” Colp-Haber said.
The age gap can have repercussions on the relationship. After one start-up signed a lease for a 2,000-square-foot space in San Francisco’s South Financial District, Cassriel shot the founder a text invitation for drinks.
“Hey dude,” Cassriel remembers his client writing back, “I’m not old enough to drink yet. I’m only 19.”
The bright stars of AI don’t always know — or buy into — the etiquette of the real estate business, brokers noted.
Traditionally, tenant reps seek exclusivity agreements with their clients, but tech firms, especially those looking for small offices, have been known to shift alliances. An AI tenant might hit a morning tour with one broker and meet up with a different one in the afternoon.
“I think there are some trust issues and skepticism on the AI side,” Doherty said.
Veteran brokers remain slightly unclear on what some of the AI firms actually do and are sometimes confounded by their particular workplace habits.
“One of the execs that I spoke to recently said they described their culture as 9-9-6. I’m thinking it’s some kind of mythology,” said CBRE’s Mary Ann Tighe in December onstage at New York University’s Schack Institute of Real Estate Capital Markets Conference. “He’s like, ‘No, no. It’s 9 a.m. to 9 p.m. six days a week.”
In-office industry
That productivity pressure is feeding directly into office use.
At Two Trees’ The Refinery at Domino, in Williamsburg, Brooklyn, AI or AI-related tenants occupy about half the 460,000-square-foot building. Five or six of those companies have an employee checking in seven days a week, Zahler said. Ninety percent of tenants have someone coming to the building five days.
The intensity is also showing up in AI firms’ rapid expansion, with some doubling or tripling their space at The Refinery in a matter of months. The speed isn’t limited to real estate. Zahler said hiring velocity is unlike anything she has seen before, with companies even recruiting during building tours.
“I’ve been on tours where someone has to step away because they’re dealing with hiring candidates as we’re walking,” Zahler said.
The demand is beginning to diminish the surplus of available space, especially in San Francisco, where the vacancy rate has hovered above 30 percent since about 2020.
Still, landlords are swayed by the potential upside, explaining the scramble to meet the tenants’ needs.
BXP’s Salesforce tower is a trophy space, 98 percent leased, with some of the best views of San Francisco Bay.
But Christine Yeun, vice president of leasing, said even there the firm is considering upgrades to appeal to the AI tenant.
“Our competition is spending $40 million on full-floor amenity centers, with podcast rooms and climbing walls,” she said.
The AI industry’s growth creates an incentive to accommodate tenants early, even at short lease durations, in the hopes of capturing the next Anthropic — which leased just 7,000 square feet in 2021 and now has 1 million in San Francisco alone — before it scales.
That kind of growth could be repeated. Sierra AI, founded by OpenAI’s chairman and former Salesforce co-CEO Bret Taylor, doubled its footprint in San Francisco’s SoMa neighborhood in the beginning of 2025, taking about 80,000 square feet in a BXP building. By the end of the year, Sierra wanted triple that, and the company moved out, taking 300,000 square feet at McCarthy Cook & Company’s 185 Berry Street in the China Basin neighborhood, next to the San Francisco Giants’ Oracle Park and Tishman Speyer’s Mission Park.
When BXP leased 70,000 square feet to Decagon AI, at 680 Folsom in SoMa, Yeun was prepared for the question. “On the tour, they were asking what we’d have available in three years,” she said. “They’re negotiating up front that they want the rights to space in the building.”
If a landlord can accommodate that kind of growth, tenants may stick around after their flexible term.
“From a landlord’s perspective, if you can accommodate a growing AI business, there’s a higher likelihood that they’re going to look to expand within the building,” Marans of Savills said. “The AI boom is very much wanted by landlords. I think there’s a deep belief in the industry as a whole and its staying power.”
Recent Comments