The Saga of Snowflake: Can AI and a New CEO Save the Company?
Snowflake shone bright in the last tech epoch. Keeping up with the AI boom falls to Sridhar Ramaswamy, an exacting ex-Googler.
Sridhar Ramaswamy seldom takes a day off. He prefers to work seven days a week, putting in a “non-small number of hours” on the weekend, he said, usually in the morning before his wife, a dentist, wakes up and then again in the afternoon.
“Listen, I’m from India: We’re always very insecure. I can’t sit around and not do anything,” he said. “I like to work.”
As the new CEO of Snowflake, a struggling data and software company that epitomized the last tech era’s cloud boom, he has sought to shake up the place and has dispensed with some parts of the looser management style favored by his predecessor, Frank Slootman. One basic change: Slootman preferred to operate from his Montana house. Ramaswamy has a fourth-floor desk next to other engineers in the company’s San Mateo, Calif., office.
Another difference between the two: Whereas Slootman might’ve tasked an executive with a project and anticipated a summary several months later, Ramaswamy expects his reports to keep him in the loop. “I don’t want to dampen creativity,” he said. “But I also don’t want it to be the case that I disagree with a premise, someone works on it for three weeks and then on slide one, I go, ‘This is a disaster!’”
Ramaswamy’s reply to someone who chafes under the expectations? “For the really recalcitrant ones: Trust is earned,” he said as we hiked through the hills near his South Bay home. “Nobody deserves independence. They earn independence.”
We’d ended up there—on a sunny day that felt hotter than the mid-80s outside—because I’d asked him where he goes to relax and unwind. Quickly, we came to an uncomfortably vertical climb up a few hundred feet involving several dozen wooden steps. In the past, Ramaswamy’s friends, like Sequoia Capital’s Jim Goetz, have joined him on this route; in other moments, Ramaswamy, 57, has enjoyed testing his fitness by doing the climb while breathing only through his nose.
“Just a fun little exercise of ‘Can you be zen?’” he said as we reached the stairs’ end. Still, he conceded, “some friends call it cardiac hill.”
Come to think of it, Snowflake itself is in a bit of an uphill battle at the moment. Lately, its stock has, well, melted. The shares have lost nearly half their
value since the company named Ramaswamy, a former top Google executive, as CEO in February just as it was forecasting disappointing sales growth.
Snowflake was an early pioneer in selling data-warehousing services, where companies store mass amounts of data from multiple sources in the cloud. But, increasingly, it faces stiff competition from rivals like Databricks, Amazon and Google, and it has lagged behind in adding artificial intelligence to its products.
What happens next at Snowflake might serve as a fair bellwether for what’ll happen across Silicon Valley in the immediate future. Many companies find themselves in the same uncomfortable position: needing to satisfy investor—and customer—demands to seize this moment in AI and do so faster than competitors.
The contrast between Snowflake’s present and past adds some dramatic complexity to the situation. This is the first real struggle for Snowflake, which enjoyed a charmed early life and investment from Silicon Valley’s top names, including Sequoia. Sure, Snowflake sells nothing sexy, but the startup had an honest-to-goodness mystique, especially after Slootman led it in a gusher of an IPO just four years ago, the largest ever for a software company. On the stock’s debut, it began trading at $245, more than double its $120 IPO price.
Some further irony to the scenario: Ramaswamy certainly didn’t expect to find himself in the position of leading Snowflake’s comeback. When he sold his AI search startup, Neeva, to Snowflake in 2023, he’d planned to exit Snowflake at the end of six months.
Ramaswamy would’ve been a hot free agent. He’s not a mass-market name like, say, a Benioff, but he did have a long tenure at Google, which ended with running its all-important ads business. The experience put him in good stead with Silicon Valley insiders. Elon Musk, for instance, tried to woo him to join Twitter as CEO before the Neeva sale to Snowflake.
Once at Snowflake, Slootman talked him into its top job, hoping Ramaswamy’s AI background would accelerate Snowflake’s own AI plans.
So Ramaswamy finds himself currently leading perhaps the most interesting turnaround story unfolding in Silicon Valley. If the exacting Ramaswamy can’t steady Snowflake, his efforts will underscore how daunting it’ll be for other businesses with less experienced leaders to navigate AI, which remains a hugely costly and ultimately unproven technology.
Even if Ramaswamy hasn’t convinced the public markets, billionaire venture capitalist Reid Hoffman, who helped talk Ramaswamy into joining Greylock Partners when he left Google, has plenty of confidence in him. “Sridhar plays to win,” Hoffman told me. “He’ll either change for the game he’s playing—or he’ll change the game he’s playing and win. He doesn’t delude himself.”
While at Google, Ramaswamy had a reputation among his colleagues: He enjoyed challenging their ideas, even ones from then-CEO Eric Schmidt or co-founders Larry Page and Sergey Brin.
“I saw him have heated debates with Larry, with Sergey, with Eric, and he’d say, ‘You’re wrong—here’s the six reasons you’re wrong,’” recalled Patrick Pichette, who served as Google’s chief financial officer. (Said Ramaswamy: “I’ve never been shy about my opinions.”)
Ramaswamy, who was born in southern India and later earned a doctorate in computer science at Brown University, joined Google in April 2003. (Previously, he’d worked at Bell Labs and at Epiphany, a dot-com–era software startup.) As Google went public and started to form into the megagiant we know today, Ramaswamy became a valuable engineer and manager within the company’s drive to bolster its digital ads business.
At Google, Ramaswamy impressed co-workers and his staff with his detached, calculating manner. “He’s like Data from ‘Star Trek’: Compute, compute, compute,” recalled one former Google executive who worked with him. And he excelled at leading during stressful moments—in what Ramaswamy called a “war room” setting, when small groups of engineers would need to fix a problem, including the issue of what to call their efforts.
“Funny thing is, there was a colleague at Google who was offended that I called something a war room: ‘What’s with this martial metaphor shit?’ So to make him happy, I started calling them our ‘basket-weaving efforts,’” Ramaswamy told me. “Look, it’s all about working together: weaving baskets, war rooms—same thing.”
He took over running Google’s ads and commerce team in 2014 with Susan Wojcicki when the ads business totaled around $50 billion in sales. (After an internal fight with Wojcicki, he assumed solo control of the team a year later.) When Ramaswamy left Google in 2018, ads had grown to nearly $120 billion.
His departure came amid personal disillusionment over the economic pressure from online advertising. To reach $100 billion–plus in sales, Ramaswamy had needed to relentlessly push Google ads at users. And he felt the company had gone too far: A breaking point for Ramaswamy, a father of two, came in 2017 when a Times of London investigation revealed Google had sold ads against YouTube videos of children that appealed to pedophiles.
For a time, he joined Hoffman at Greylock, but almost immediately he started working on Neeva, a search startup that would charge users a $5 monthly subscription rather than rely on ads for revenue. “We were going to build a search engine with 40 people that was powered by AI and had no ads in it,” recalled Vivek Raghunathan, who worked with Ramaswamy at Google and left to co-found Neeva with him. They brainstormed their initial plans on walks through Google’s campus. “The mission was audacious,” Raghunathan said.
As Ramaswamy and Raghunathan set out on Neeva, Frank Slootman was settling into Snowflake’s CEO role, then the third person to have the job.
A pair of technology wonks—two Frenchmen, Benoît Dageville and Thierry Cruanes—had founded the company in 2012. They quickly realized they would need help running the company. (“I’m an engineer. I’m not a business guy,” Dageville told me. “I’m not a manager either—I hate managing people.”)
Snowflake’s first CEO was also one of the startup’s major investors, Sutter Hill’s Mike Speiser. He helped Dageville and Cruanes commercialize their idea: data warehouses in the cloud rather than ones held within internal servers. Since a data warehouse is exactly what it sounds like—a digital repository of large amounts of data from various sources—it can overflow with too much stuff and become unruly, just like a physical warehouse with real-life goods. Snowflake’s software made it easier for customers like Rent the Runway and Capital One to study and analyze their company’s data.
In 2014, Speiser turned over the CEO spot to Bob Muglia, a 23-year Microsoft veteran. That went well, and Snowflake’s valuation shot up from $75 million the year Muglia joined to nearly $4 billion four years later. But as Snowflake’s investors began eyeing a possible IPO, the company’s board recruited Slootman to take Muglia’s place.
Slootman’s speciality was IPOs—readying a company for a listing and guiding it through its initial years as a public firm. He’d enjoyed great success at this while running two other software companies, Data Domain and ServiceNow.
Slootman had expected to retire after ServiceNow but got coaxed into a final run with Snowflake. He made some straightforward changes, like hiring a new CFO, and by the time he set out to create a pandemic-era virtual road show for the IPO, he had impressive figures to wow investors. In fact, Snowflake’s revenue climbed to over $260 million in the company’s fiscal year ending in January 2020, nearly triple from a year earlier.
When Snowflake’s board initially persuaded Slootman to take the position, he’d told the directors to expect him to keep the job for no more than five years. So who could take his spot?
Just as Slootman’s fifth year at Snowflake neared, Ramaswamy sensed he and Raghunathan had largely taken their audacious mission as far as it could go. They’d made an excellent product, Ramaswamy felt, but found limited consumer demand for a subscription-based search engine.
“We weren’t quite seeing hockey-stick growth at that point,” he admitted. “And a lot of folks on my team I had recruited from pretty plum jobs at Google and other places. They’d been living on just a salary.” He wanted them to share in the rewards of an exit sooner than later.
When Ramaswamy began to shop Neeva around, he weighed several acquirers before settling on Snowflake, sensing he and his team would fit in best there. Slootman initially spoke to Ramaswamy about acquiring Neeva on a Thursday. By Monday, they had signed terms.
If Slootman had one eye on the door at Snowflake, so did Ramaswamy. During the negotiations, Ramaswamy made it clear to Slootman that he intended to leave by the end of the year.
As Slootman crunched the numbers for Neeva, his CFO, Mike Scarpelli, encouraged him to think about whether he could convince Ramaswamy to stay. “I told Frank at the time, ‘Sridhar could be your successor—you need to spend some time with him,’” Scarpelli recalled.
Before officially joining Snowflake, Ramaswamy and Raghunathan shut down Neeva entirely: Snowflake had no interest in a consumer search engine. They wanted Ramaswamy and his team for their “technical chops,” as Ramaswamy put it. Snowflake wanted the new team to build AI tools.
In his first few months, Ramaswamy operated with no direct reports but nonetheless hoped to ignite Snowflake’s AI efforts. “I’ve joked to friends: It was an exercise in soft power to try to persuade a company of 7,500 people that change was needed,” he said.
As the six-month mark neared, Ramaswamy went on a weeklong trip to Europe with Slootman. Mostly, they went to meet with customers, whose loyalty to Snowflake's products impressed Ramaswamy. “Honestly, that’s what convinced me there was a future here,” he said.
At a dinner in London overlooking the Thames, Slootman, who by then had come to realize how much AI would matter to Snowflake’s next CEO, pitched Ramaswamy on stepping in to replace him.
Later, Ramaswamy called his mother in India to tell her the news. She was thrilled. “I come from a pretty humble family,” he said, “so absolutely I think my mom was more excited about me becoming CEO than, honestly, I was.”
In February, Slootman announced the change. (Slootman, who wouldn’t comment for this story, remains chair.) With the CEO swap coming on top of a lackluster sales forecast, Wall Street panicked. Snowflake shares slid nearly 20% in a day to under $190. And they’ve continued to slide for the last several months, recently down to around $110.
Ramaswamy believes the changes he’s made at Snowflake can eventually restore investor confidence—“It’s nothing where we can wave a magic wand and have things get better instantly,” he said—and over the past year, Snowflake has released a raft of AI features.
Those features probably aren’t extraordinary enough to change the world—or destroy it. (“I don’t feel I’m qualified to comment on how fast the technology can get to where it’s scary,” Ramaswamy said.) Rather, many are AI basics.
Snowflake, for example, has introduced a chatbot for data search and has released proprietary large language models so customers can build their own AI apps. A more complex feature can convert what’s called unstructured data (prose, images, video) into structured data (spreadsheets, databases, Excel-type files). A chatbot can parse the latter much more easily.
Eventually, someone using Snowflake software might see nothing more than a single text field: Plug in a query involving the data, and the answer appears below. “It’ll be a way for users to talk to their data using natural language,” said Raghunathan, who is now a Snowflake vice president. “People will talk to their structured data and say, ‘Why are same-store sales in Texas down today?’”
Within the sales department, Ramaswamy has pushed staff away from “the whale-hunting game,” he said, eschewing Snowflake’s past prioritization for landing big deals over amassing a significant bedrock of clients that bring in $100,000 to $2 million in annual revenue.
Ramaswamy has urged staff across the company to adopt “a spirit of urgency and continuous self-improvement,” he said. “Snowflake had that in part, but it was not running through the entire company as a way of living.” This has involved, for instance, asking engineers to consider how they can cut down on the time required for tasks that would take “tens of minutes to minutes.”
As Snowflake continues on its turnaround, Ramaswamy has made it clear to his direct reports that they should expect him to deeply examine their proposals. “One thing that was somewhat surprising—and still is—to internal folks is my demand and desire to understand,” he said.
It’s a change from the Slootman days. “Frank probably made faster decisions and shot more from the hip,” said Scarpelli, the CFO. “Sridhar’s a little bit more data driven—that’s probably the engineer in him—and doesn’t mind getting into the details. I cannot believe the volume of information that he reads and can retain. It’s mind boggling.”
I heard much the same from Denise Persson, the chief marketing officer. “He’s very hands-on: I don’t think anyone else on the planet would have this commitment to being hands-on,” Persson said. “He’s relentless.”
“Frank’s obviously a legendary CEO—I’m much more in the thick of it,” Ramaswamy explained. “I want to look at steps along the way and not wait for a month or a quarter.”
Even with all that planning, Ramaswamy has hit some notable speed bumps. In May, Snowflake had to scrap plans to buy Reka AI, a startup that makes LLMs. (Ramaswamy wouldn’t comment on exactly why the deal fell through.) Around the same time, it had to publicly announce that hackers had accessed some Snowflake customer accounts; Ramaswamy says those accounts weren’t using safety features like two-factor authentication correctly.
As much as Ramaswamy evangelizes about AI, he also seemed compelled to assure me that “it’s not as if everybody now works on AI” at Snowflake. It seemed like a rare admission from a CEO at this moment—that other parts of the business matter as much as AI, and that eventually Silicon Valley’s obsession with AI will ebb.
Ramaswamy expects many of the changes he has made—the greater rigor, the improved speed—to serve the company well no matter what era dawns next. “There’s a permeation process going on,” he said.
When Ramaswamy has made time on his calendar for leisure recently, he has been preparing for a mile-plus swim from Alcatraz Island to San Francisco that he hopes to do this fall.
To train, he previously had a coach critique his form and study it by swimming under Ramaswamy as he swam. Recently, he joined a Pacific Heights swim club. “The first time I went, they gave me some helpful advice: ‘You’ll be in a little bit of a shock, but in five minutes, your face will be numb. You won’t feel a thing. You’ll be fine,’” he recalled.
Ramaswamy only learned to swim as an adult, and only after wearing out his knees as an obsessive marathoner. “I was a huge runner: Zen for me was a 20-mile run,” he said. “There’s nothing like running in 34-degree rain that can make you feel what cold is like.”
