AI robotics companies love San Francisco. They're just too big to stay

By Kevin V. Nguyen Photography by Sean Fan Published Aug. 23, 2026 at 6:00am Share Share Copy link to this article Email (opens in new tab) Twitter (opens in new tab) Bluesky (opens in new tab) Facebook (opens in new tab) LinkedIn (opens in new tab) Telegram (opens in new tab) WhatsApp (opens in new tab) Reddit (opens in new tab) Close share menu There are $ 0 comments on this story Published Aug. 23, 2026 at 6:00am For nearly eight years, San Francisco was the perfect playground for Bright Machines to hone its technology and business model. The robotics company got its start in a SoMa loft, initially aiming to develop tech for other manufacturers’ factories. But as the AI boom created an insatiable demand for specialized chips and servers, Bright Machines pivoted to becoming a manufacturer itself. In 2023, the company moved into a 10,000-square-foot industrial building in the Inner Mission where it could build and test its robots, which assemble hardware for data centers. Because its machines could produce custom designs faster than traditional hardware manufacturers — some take weeks to reconfigure an assembly line and retrain staff — Bright Machines shipped finished products out of its hybrid office-and-workshop, just up the street from Dandelion Chocolate. ADVERTISEMENT “There is a good energy and vibe to this area,” said CEO Sviat Dulianinov. “Especially for the young engineers. They can walk outside and encounter so many other people working on similar things.” As AI applications push further into the physical world , the office market around Bright Machines’ headquarters, with its proximity to the central freeways, has become a hotbed for similar companies experimenting in the space. Firms like Mira Murati’s Thinking Machines Lab, Weights & Biases (acquired by CoreWeave), and Physical Intelligence are steps from Dulianinov and his team. That kind of scrappy setup worked when Bright Machines was still proving its technology. But now, bigger orders are coming in. Staffers in sales, operations, finance, and product management are being hired every month. There’s no room to put additional desks. CEO Sviat Dulianinov sees the move to the Peninsula as an opportunity to recruit more talent. So Bright Machines will leave San Francisco this fall for Burlingame, where it has leased a space five times larger. “We would have loved to stay here,” Dulianinov said. “But now we need to focus more on production.” This pattern will play out across the AI robotics scene if investment dollars continue to flow as they are, as companies that thrive in San Francisco’s dense startup ecosystem find that large-scale industrial space and infrastructure are hard to come by. Take Jeff Bezos’ Project Prometheus. The stealth AI startup, which has an office downtown, spent the early part of this year hunting for industrial space , with a requirement of approximately 100,000 square feet. It struck out in February on a warehouse on 23rd Street in Dogpatch, then signed a lease at the old American Steel Blocks site in West Oakland. Related Behemoths of the boom: Ranking SF’s biggest AI companies by office size (opens in new tab) An SF startup is secretly testing robots in Airbnbs, and trashing them, lawsuit claims (opens in new tab) Still, the hype you’re hearing about robots in San Francisco is real. The amount of commercial space leased by robotics companies this year alone surpassed the cumulative total from 2020 to 2025 — driven primarily by the growth of autonomous vehicle companies, according to JLL. Most of these companies required R&D and industrial space, but a third of the leases signed included offices. However, the biggest robotics deals are funneling toward the East Bay or Silicon Valley — which stretches down to Morgan Hill, where Archer Aviation last month leased more than 500,000 square feet (opens in new tab) at the newly built Cochrane Technology Center. Of all AI robotics-related deals JLL has tracked since 2017, San Francisco accounted for only about 11% of the total square footage leased in the Bay Area. Related OpenAI expands to the South Bay by leasing massive Mountain View campus (opens in new tab) Amazon’s landlord pitches SF’s second-tallest tower as part of massive Caltrain buildout (opens in new tab) Bright Machines intends to work and manufacture out of its 16th Street headquarters until it hands over the keys to the landlord. Some workers lament leaving San Francisco, where they enjoy the dense, walkable neighborhoods. Others, like Dulianinov, who commutes from the Peninsula, see the impending move as an opportunity to recruit a wider range of talent, especially in engineering. It’s been a dizzying journey to growth. Bright Machines attempted to go public via SPAC in 2021 before pulling out of the deal (opens in new tab) and raising more venture capital. Now, as the build-out of “AI infrastructure” becomes a mandate of the country’s biggest investors (opens in new tab) , the company is in position to become a cog in the industry, especially as more advanced chip manufacturing gets on-shored. “Everyone wants to deploy new servers [into data centers] faster,” said Fiaz Mohamed, chief growth officer at Bright Machines. “The technology is getting pushed like it’s never been before, and our customers are having to refresh their products more often.” Assembling chips and servers for data centers require precise placements. Bright Machines’ San Francisco office is tight on space. Tori Colthurst, robot perception engineer, says she’s keeping an open mind about the office move. Bright Machines develops and ships AI-enabled robots but also sells a data platform and design software that allow customers to more swiftly operationalize new designs in their factories. The robot arms have a wide range of precise movements and use cameras and infrared energy to scan for defects. The company also makes and ships custom products to order at its own factory lines in Burlingame and Guadalajara, Mexico. Mohamed said Bright Machines’ robots are deployed at more than 150 sites worldwide. “The game for us now is all about scaling,” he said. “We’ve proved that we can adapt to new product requirements fast. But can we do that across hundreds, maybe even thousands of lines? It’s really hard.” Copy link to this article (opens in new tab) (opens in new tab) (opens in new tab) (opens in new tab)