AI Fuels a San Francisco Real Estate Rebound
San Francisco’s commercial real estate market may be on the verge of a comeback, fueled by the burgeoning growth of artificial intelligence. A recent report indicates that the city, now considered a hub for generative AI, is witnessing increased demand for office space from companies in this rapidly expanding sector.
According to a Cushman & Wakefield report, AI companies collectively leased 2.6 million square feet of office space in San Francisco between 2022 and 2024. The real estate brokerage anticipates this figure to triple by the end of 2026. Robert Sammons, a senior director of research at Cushman, noted that this surge in demand will significantly aid San Francisco’s recovery from the record-high office vacancy rates experienced last year. The brokerage projects that, thanks to the AI sector, the market will see more space leased than vacated this year and next, marking a reversal of the trend since the COVID-19 pandemic.
“San Francisco… has always been such a boom-and-bust market, more so than a lot of other places,” Sammons observed. “And it feels as if we’re in the early stages of another boom.”
Other real estate firms share this optimistic outlook. JLL, for example, projects that AI companies will lease an additional 7 million square feet in San Francisco between the start of this year and the end of 2030.
Cushman’s forecast is based on industry trends and the existing demand from generative AI companies. Last year alone, these companies leased 1.15 million square feet in San Francisco, accounting for 15% of the total office space leased in the city. Key players in this expansion include OpenAI, Anthropic, and Databricks.
Databricks recently secured 150,000 square feet at One Sansome Street, with plans to double its footprint in the coming years. Meanwhile, OpenAI inaugurated its new headquarters in Mission Bay, subleasing space from Uber in 2023.
According to JLL’s research, AI companies signed 90 leases last year, up from 36 in 2023.
Sammons believes that considering the number of generative AI companies in San Francisco and the substantial funding they’ve secured, even more deals are expected in the coming months.
Globally, venture firms invested $56.2 billion in generative AI companies last year. Of this, 52% went to San Francisco based companies, with Bay Area companies accounting for 82% of the funding total.
The Bay Area is home to 825 generative AI companies, with 466 based in San Francisco, according to the Cushman report. In contrast, the New York metro area, the second-largest center for this technology globally, has 237 such businesses.
Variety in Demand
Within the city, there is a diverse range of companies, according to Sammons. “These are companies that could still be in someone’s apartment all the way up to the big players,” he said.
These companies have varied space requirements depending on size and outlook.
Smaller startups often seek the best sublease deals, prioritizing space that’s already finished. “They want that ‘plug-and-play’ space, where they can just move right in and get started,” Sammons stated, noting they are less inclined to wait for or fund construction.
These smaller companies typically look for 3,000 to 10,000 square feet. Spaces in older brick or timber buildings in the Mission or South of Market areas, or areas like Jackson Square or the Northern Waterfront, are popular choices. They are also considering traditional mid- or high-rise office buildings in the central business district.
Chris Pham, a senior analyst at JLL, highlights Mission Bay as another popular location, partly because OpenAI is located there, and because many venture firms have offices in that area. Alexander Quinn, JLL’s senior director of research for Northern California, adds that the area offers easy access to Interstate 280 and Caltrain.
Many such smaller AI startups are targeting deals of three years or less, according to Sammons. Many are still determining future scale, or aim to be acquired by larger companies. Although shorter than traditional leases, landlords are generally accommodating, he said. In fact, 63% of all leases signed by generative AI companies last year were for less than three years, according to Cushman’s report. Only 11% were for over five years.
Another real estate company, CBRE, sees similar trends among smaller AI startups. Colin Yasukochi, who leads the firm’s research on tech industry real estate demand, mentions that many are seeking downtown spaces, aiming for two-to-five year deals, and often prefer move-in-ready properties. They are willing to pay higher rents for these spaces, understanding that build-out costs would be even higher.
“Their [venture-capital] funders don’t necessarily want them to spend a lot of money on real estate,” Yasukochi said. “They want them to spend money on developing their technology.”
Quinn observed that the demand from smaller AI companies is influencing the overall market. Approximately two-thirds of the leases signed by AI companies last year were from those signing their first office deals or moving into newer, larger spaces. Pham noted that around 40% of the AI companies that signed deals last year were either seed-stage or Series A companies, indicating they are in their early stages.
“That’s a pretty remarkable [statistic],” Quinn said. “It shows just the amount of firm formation that’s occurring within The City of San Francisco, the ecosystem that exists here and the growth of that market and expectations of AI.”
In contrast, Sammons notes that larger players in the generative AI market generally prefer recently built buildings, aiming to be the sole tenants if possible. Yasukochi adds that these larger companies are more likely to sign seven-year deals.
Impact Beyond Real Estate
Beyond the real estate sector, the demand from generative AI companies benefits San Francisco’s office market and economy in several ways, Sammons says. He notes that while the high vacancy rates were partly caused by tech companies embracing remote work during the pandemic, AI companies’ employees tend to be in the office more often, fostering collaboration.
Furthermore, the demand for space from these companies is generating office demand for other businesses, both within and outside of the tech sector. This includes companies that provide services to generative AI businesses and non-technology companies that have launched AI divisions, as Pham points out. With about 20% of all workers with AI expertise residing in the Bay Area, companies are eager to be located near this talent, according to Quinn.
Sammons reflects on the earlier narrative that the pandemic had ‘killed’ San Francisco’s office market. The increased leasing activity by generative AI companies is beginning to change this perception. “San Francisco is not dead, clearly,” he asserted. “There is activity again. Tenants want to be here.” He concludes that “Everything is feeling better.”
