On an office floor overlooking the salt ponds and mudflats of Menlo Park, a room full of founders, investors, and engineers packed in to listen to execs at OpenAI, Windsurf, and Glean opine on the secret to “scaling AI startups.”
The lesson — as was preordained — was to be steadfast in one’s conviction that only their technology can solve a customer’s pain point.
A decade ago, this scene may have taken place in a hotel ballroom or a Sand Hill Road mansion — but today, it’s transpiring at the offices of an already established tech company, Snowflake, which famously went public in 2020 in the “biggest software IPO ever.”
Last week, Snowflake cut the ribbon on its “Silicon Valley AI Hub”— a 30,000-square-foot top-floor space in its new headquarters designed for companies and developers chasing the future of artificial intelligence. There’s more in the works.
Next year, it plans to add a rooftop restaurant and bar to make the space more of a destination.
Over the course of three months, an inaugural class of 13 startups will set up shop there where they’ll have access to coworking space, professional video studios, and snacks. But more important than the advanced sparkling water machine is the free massive computing infrastructure and proximity to Snowflake’s network of investors.
All they have to do is use its platform.
“These startups are taking over the world,” said Denise Persson, chief marketing officer of Snowflake. “They’re moving so fast that the big companies can’t keep up.”
If you can’t beat them, then buy them. Or at least befriend them early enough and give them the space to thrive. “As a startup, you don’t want to worry about real estate,” she said. “All you should care about is getting your product in order.”
Snowflake, which helps companies connect large sets of data over the cloud, pivoted hard into AI after the release of ChatGPT in 2022. During this time, its top competitor, Databricks — regarded as the more AI-centric company of the two for now — also gained serious ground, raising four rounds of funding totaling between $16 and $17 billion in two years. In June, both companies even hosted their own conferences at the Moscone Center just a week apart.
With giants like Nvidia, Microsoft, and Amazon spending obscene amounts of money backstopping the AI boom, Snowflake may be arriving a bit later to the party. But its secret weapon? Having the largest, newest space located between the AI frontier of San Francisco and the backbone of Silicon Valley.
Warrick Taylor, vice president of global real estate and workplace at Snowflake, said the company had been looking to move on from its previous San Mateo headquarters in earnest prior to the AI boom since the IPO set the stage for hyper growth. Between 2020 and 2025, the company’s headcount went from 1,000 to 8,000 workers, most of whom are required to come into office at least three days a week.
“Because [office space in the] Peninsula was constrained, we initially thought about splitting our workforce across the North and South Bay,” Taylor said. “But we thought it was important for our culture that we pursue a single campus.”
Market timing worked perfectly in Snowflake’s favor. Meta had pre-leased and built out an entire 773,000 square foot campus in Menlo Park it never occupied. Had the pandemic never happened, Snowflake would likely have had to partner with a developer to build an entirely new office of its own if it wanted to stay in San Mateo County.
“We were very fortunate,” said broker Luke Ogelsby of CBRE, whose team represented Snowflake in the eventual sublease of the entire Menlo Park campus. “We were hitting the accelerator just as bigger tenants were disposing of space.”
Negotiations were aided by the fact that Snowflake struck a deal with Meta to sublease another unused building near Seattle in 2024 before inquiring about Menlo Park. “The relationship we established with Meta from the previous negotiation was instrumental,” Taylor said. “When sticky issues came up, I was able to quickly get on a call, principal to principal, without having to weave in third-party intermediaries.”
At Menlo Gateway — built by office developer Bohannon Companies in 2019 — Snowflake now controls a three-building complex flanked by two seven-story parking garages on each side. After signing what is still the Bay Area’s largest post-pandemic office lease, the company started relocating its workers in April. For now, employees only fit into two of the three buildings. The rest, they will eventually grow into, Taylor said.
While staking a larger claim physically in the Bay Area, the company is also hoping to lay down a marker in the AI race for the long haul. In November, Snowflake will unveil a new suite of AI products, including one Persson called “ChatGPT but for your organization’s information.” In the past two years, the company has partnered with both OpenAI and Anthropic to integrate more of both language models into Snowflake’s products.
It’s been six months, but much of Meta’s minimalist designs still lingers, primarily in the form of exposed wiring and pipes on the ceiling, giving the floors a slight computer lab feel.
In the weeks leading up to the launch of its AI hub, Taylor’s workplace team raced to cover the white walls with Snowflake’s blue-hued winter-themed branding. Now there are stuffed polar bears, penguins, and foxes on each floor. Meeting rooms and nooks were made to resemble cozy cabins. The still-unfinished gift shop has the tagline “Take a little snow home” inscribed on the wall.
She said the company anticipates hosting up to 10 events a week at the AI hub, including networking dinners, hackathons, and educational sessions. By making its headquarters a popular gathering space for the AI community now, Snowflake hopes to get in early with the next generation of unicorns.
“Now, they no longer have to start in a one-bedroom apartment,” she said. “Why not say you started on the eighth floor of Building 135?”
