By Kevin V. Nguyen Photos and videos by Morgan Ellis Published Sep. 29, 2025 • 6:00am In an office overlooking Menlo Park’s salt ponds and mudflats, founders, investors, and engineers gathered to hear executives from OpenAI, Windsurf, and Glean discuss the key to "scaling AI startups. " The consensus was clear: success demands unwavering belief that only their technology can truly solve customer problems. A decade ago, such a scene might have unfolded in a hotel ballroom or a Sand Hill Road mansion, but today it takes place at Snowflake’s headquarters—an established tech company known for its record-breaking 2020 IPO, the largest software IPO ever. Last week, Snowflake inaugurated its “Silicon Valley AI Hub, ” a 30, 000-square-foot top-floor space within its new headquarters, designed to support companies and developers building the future of AI. Additional plans include launching a rooftop restaurant and bar next year to enhance the hub’s appeal. Over three months, an inaugural batch of 13 startups will operate there, gaining access to coworking spaces, professional video studios, and more—though the most valuable offerings are free access to extensive computing infrastructure and connections to Snowflake’s investor network. The primary condition is that these startups utilize Snowflake’s platform. “These startups are taking over the world, ” said Denise Persson, Snowflake’s chief marketing officer. “They’re moving so fast that big companies can’t keep up. ” Her philosophy: if you can’t beat them, partner early and provide space for startups to thrive. “As a startup, you shouldn’t worry about real estate; focus on perfecting your product. ” Snowflake, a cloud-based data connectivity company, pivoted strongly toward AI following ChatGPT’s 2022 debut. Its chief rival, Databricks—perceived as more AI-centric—raised $16 to $17 billion across four funding rounds in two years. Both companies recently hosted back-to-back conferences at San Francisco’s Moscone Center. Though tech giants like Nvidia, Microsoft, and Amazon heavily bankroll the AI boom, Snowflake’s advantage lies in its large, state-of-the-art space strategically located between San Francisco’s AI hub and Silicon Valley’s core. Warrick Taylor, Snowflake’s VP of global real estate and workplace, revealed the company’s prior plans to relocate from its San Mateo headquarters before the AI surge. With headcount expanding from 1, 000 to 8, 000 between 2020 and 2025, and a hybrid work policy requiring employees to come in three days weekly, Snowflake prioritized a unified campus despite initial thoughts of splitting between North and South Bay locations to avoid fragmenting their culture. Market timing favored Snowflake when Meta, which had pre-leased a 773, 000-square-foot Menlo Park campus it never used, decided to sublease.

Without the pandemic, Snowflake might have needed to develop a new facility. Broker Luke Ogelsby (CBRE), who helped broker the deal, noted Snowflake accelerated while larger tenants shed space. Prior subleasing experience with Meta near Seattle helped Snowflake negotiate efficiently for the Menlo Park campus through direct principal-to-principal communication, bypassing intermediaries, Taylor said. Snowflake now occupies a three-building complex at Menlo Gateway (built by Bohannon Companies in 2019), flanked by two seven-story parking garages. After signing the Bay Area’s largest post-pandemic lease, relocation began in April. Currently, only two buildings are occupied, with plans to expand into the third. While physically expanding its Bay Area presence, Snowflake aims to deepen its AI ambitions. In November, it will launch a suite of AI products, including a tool Persson compares to “ChatGPT but tailored for your organization’s information. ” Over two years, Snowflake partnered with OpenAI and Anthropic to integrate more advanced language models into its offerings. Though six months since Meta’s occupancy ended, the new headquarters still retains minimalist design elements like exposed wiring and pipes, giving a computer lab ambiance. Leading up to the AI hub launch, Taylor’s team rebranded snowy white walls with Snowflake’s blue winter theme, adding stuffed polar bears, penguins, and foxes on each floor. Meeting rooms and cozy nooks were styled like cabins. An unfinished gift shop bears the slogan “Take a little snow home. ” Snowflake plans to host up to 10 weekly events at the AI hub—networking dinners, hackathons, and educational sessions—positioning its headquarters as a vibrant nexus for the AI community. The goal: to engage early with future unicorns. “Now, startups don’t have to begin in a one-bedroom apartment, ” Persson said. “They can say they started on the eighth floor of Building 135. ”