Veteran tech executive Frank Slootman was retired and racing sailboats last year when an old friend reached out. Mike Speiser, managing director at Sutter Hill Ventures, had a huge problem and a huge opportunity to offer him at Snowflake, a startup hoping to revolutionize how businesses store, analyze, and share data.

Seven years after its founding, Snowflake had perfected an amazing new way to run databases on cloud servers, but it was struggling to attract enough big corporate customers.

Slootman, seeing the potential, signed on as CEO, selling some of his sailboats and giving others away. “It sucked me in,” he tells Fortune, quoting the famous line from the third Godfather movie. “Just when I thought I was out, they pull me back in.”

Today, Snowflake is among the hottest of this year's growing crop of tech IPO candidates. This week, it’s expected to go public with a market value of around $30 billion. That would make Snowflake, which will trade under the ticker "SNOW", the most valuable software startup ever to go to public and the fifth most valuable tech startup ever—above Google, Lyft, and Snap, according to Renaissance Capital.

Investors are buzzing about Snowflake, among the leaders in what is known as data warehousing, which helps companies that generate vast amounts of information analyze their data in real time.

Mobile gaming companies, for example, slice and dice the trillions of interactions by their users to learn when to offer advancements for free and when to ask for in-app payments.

Now it's branching out to meet database uses beyond the data warehouse. Hedge funds are using Snowflake’s services to connect to other data sources like retail sales or weather reports —hundreds simultaneously, in some cases—to guide investing decisions. Other companies use Snowflake to track all activity on their networks, and then apply A.I. to identify rogue logins and hacking attacks.

As more companies move to the cloud, Snowflake’s technology, which separates stored data from the computing power needed to analyze it, is gaining momentum. Last year, the company’s revenue jumped 174% to $265 million. It was also the fastest growing business application of 2019 by usage, according to research firm Okta.

Customers tend to increase their spending over time. Year-old customers combined spent 58% more in the first half of 2020 that they had a year earlier, contributing to Snowflake's total sales rising another 133%.

Thanks to the growth, Slootman is hardly the most influential convert to Snowflake’s cause. Two weeks ago, the company announced that in addition to its plans to raise nearly $3 billion for itself in the IPO, Warren Buffett’s Berkshire Hathaway and Marc Benioff’s Salesforce would each also invest $250 million in the company at the same time.

Still, even with its big-name backers, Snowflake doesn't have the cloud database market to itself. Oracle, the king of pre-cloud databases, is increasingly wooing its customers with cloud products while all three major cloud platforms—Amazon, Microsoft, and Google—also have their own cloud database services. They are all joined by legions of other hot startups, like Panoply, Incorta, and Yellowbrick.

The rivals are all racing to serve the many companies seeking better and cheaper ways to store and analyze data in the cloud, says Rebecca Wettemann, CEO and principal analyst at research firm Valoir. “The data warehouse area is arguably the last of the large enterprise tech areas to go cloud,” she says. “Snowflake has big untapped growth potential.”

Oracle, with annual revenue of almost $40 billion and profit of more than $10 billion, has long dominated the database market. But the shift to the cloud offers an opening for Snowflake, says Dan Elman, senior analyst at Nucleus Research. “Oracle isn’t going anywhere, but with so much moving to the cloud, there’s an opportunity for Snowflake to capture a lot of that,” he says.

Two laptops and a white board

Like many Silicon Valley success stories, Snowflake arose from frustrations of an earlier era.

AuDatabase scientists Benoit Dageville and Thierry Cruanes had each toiled at Oracle for more than a decade, helping the software giant eek out incremental improvements in its applications. But by 2012, the pair of French immigrants foresaw cloud computing as a way of not just reducing some of their customers’ problems, but eliminating them altogether. In gust, they left Oracle and started meeting in a small apartment on El Camino Real in San Mateo, Calif., the heart of Silicon Valley.

Brainstorming with just a couple of laptops and a white board, they realized that database users were wasting huge amounts of money buying powerful servers that they only occasionally needed to use fully. Additionally, the amount of data being collected by companies was exploding, but they usually spread it across many storage facilities and kept it in incompatible formats. And a promising open source software tool for handling big data known as Hadoop was hitting the scene, but it had huge challenges producing results as quickly as business executives desired.

The cloud could solve all three problems. Computing power and storage could be increased and reduced on demand. Data could be formatted in more flexible ways. And big data could be processed in real time, avoiding the delays of Hadoop.

They quickly formed Snowflake and recruited Marcin Zukowski, a brilliant developer who had invented a new way to process database analysis requests. Venture capitalist Mike Speiser, one of the earliest backers, came up with the company name in honor of the team’s love of skiing.

“Also, snow comes from the clouds,” Dageville tells Fortune. “It’s a cloud connection. And it's white. If you name your company rain, it’s not as exciting. Snow is pure. Each flake is different.”

For seven years, the company moved forward, but by early 2019 some problems had become evident to board members like Sutter Hill’s Speiser. Sales had stalled among large customers. Rollouts of new feature and products seemed uncoordinated. And the kinds of core corporate functions needed to take the company public, like legal and HR, needed to be revamped.

That’s where Slootman came in.

“Have to shift gears”

The 61-year old Dutch executive’s first CEO job was at an early-stage startup called Data Domain that made specialized storage hardware. After joining almost at the start in 2003, Slootman helped lead it through an IPO in 2007 and then, in 2009, a $2.4 billion sale to EMC, now part of Dell Technologies.

In 2011, Slootman was recruited to run ServiceNow, at the time a small software developer focused on the help desk market. He saw the potential for ServiceNow’s suite of software, which included messaging and event tracking, to appeal to a much wider audience. He bolstered the sales staff and set out after new customers in finance, healthcare, and other industries. It worked. Taking the company public in 2012 at a valuation of about $2 billion, ServiceNow today has a stock market value of almost $90 billion.

Stepping down as CEO in 2017, Slootman followed his passion into the world of regatta sailboat racing known as Pac52 for the high-tech 52-foot long sailing vessels involved. Slootman describes it as the “NASCAR of sailing” because of the high speeds approaching 25 knots. His boat, named Invisible Hand, won the “Transpac” race from Long Beach, Calif., to Hawaii that year.

But after many more races and two years at sea, he was ready to hear Speiser’s pitch to go back to the office. “Two years straight, that’s a lot of sailing,” Slootman says now. “I got my fill a bit.”

The issues facing Slootman at Snowflake were similar in some ways to his prior companies, unique in others.

Snowflake started out going after just one part of the database market, the data warehouses that stored big data and fed business analytics apps. But even before Slootman joined, the company realized it had the ability to attack a much larger market and offer all kinds of database functions. It has even moved beyond database apps to create a marketplace for using the underlying sets of data as a sort of super-powered Bloomberg Terminal.

Instead of just storing their own data, customers can pay to access data from others and integrate it into their spreadsheets and machine learning apps, for example. One of the most popular data sets right now contains detailed tracking of COVID-19 outbreaks across the U.S.

To attack the new markets, Slootman has bulked up Snowflake’s sales staff and recruited sales teams that specialize in landing large accounts. Since he took over, the company’s overall workforce has doubled to over 2,000.

At the same time, Slootman promoted co-founder Dageville to chief product officer while eliminating four separate positions that previously had oversight of various offerings. “There were too many cooks in the kitchen,” the CEO says.

And to prepare for going public, Slootman recruited some executives he’d worked with at past jobs. At the top of the list is Snowflake’s new CFO, Michael Scarpelli, who joined a year ago after working with Slootman at ServiceNow, EMC, and Data Domain.

While the expansion has cost a lot of money—Snowflake lost $171 million in the first half of 2020—Slootman hopes to get the company’s out of the red, though without getting too specific.

“You cannot gas the company and throw unlimited resources at problems and hope for the best, which a lot of startups do and this one did as well,” he says. “Eventually you’re going to have to shift gears.”

Answering the impossible question

So far, the moves seems to be working. As of the end of July, Snowflake had 56 customers that spent at least $1 million over the preceding 12 months, more than double the 22 it had in July 2019.

One large customer is Office Depot, which replaced four different data systems, including a data warehouse and a data backup service with Snowflake. It’s halved the cost of supporting the systems and sped up the time to build new applications, says Andrew Parry, vice president for application development and support at Office Depot. Beyond speeding up typical applications, Snowflake allows the company to answer “data questions that were once impossible," Parry adds.

Even some of 2020's other IPO candidates are Snowflake customers. Food delivery service DoorDash says moving its data to Snowflake in early 2019 has doubled the average speed of a data request at half the cost of its old system.

But the competition is not letting up. Oracle just reported that its quarterly cloud services and licensing revenue grew 2% to almost $7 billion. And the always pugnacious competitor has plenty to say about its sexier, younger rival founded by former Oracle employees.

“Snowflake took a number of shortcuts to bring its single-purpose product to market quickly, but filling those gaps is going to be a much harder job,” Jenny Tsai-Smith, a vice president for product management at Oracle, tells Fortune. “Snowflake is a data warehouse-only technology. Customers don’t want multiple specialized databases as that leads to data fragmentation, management burdens, and security issues.”

Snowflake may also have a hidden vulnerability. Neil Carson, the CEO of rival data startup Yellowbrick, admits that Snowflake’s software is a “brilliant innovation.” But he points out that Snowflake is completely dependent on running its apps via the cloud platforms of Amazon, Microsoft, and Google, also its biggest competitors. That means the three rivals could raise the price of running Snowflake, Carson argues. Yellowbrick makes its own proprietary hardware systems to run database programs and sells them to customers directly or places them in cloud data centers for rent.

Snowflake is “going to get this continual massive direct competition from their landlords, basically,” Carson says. “It's going to be tough later.”

Then there is the issue of valuation. With trailing 12-month revenue of just over $400 million, Snowflake’s IPO value is climbing dangerously close to an almost certain kiss of death in prior stock market history: going public at more than 100-times sales. Perhaps the last company to do so was Castlight Health in 2014, which subsequently lost over 95% of its value, but most prior examples occurred during the Internet bubble in 1999 and 2000 and ended up disappearing altogether.

Still, none of those companies had anywhere near the size and visibility of Snowflake (Castlight’s revenue was just $13 million at its IPO), not to mention the imprimatur of investors Buffett and Benioff. While $30 billion is a considerable market cap for a newly public company, business software-oriented Workday is worth nearly $50 billion and VMware almost $60 billion. At the high end, Oracle has a market cap of $183 billion, Salesforce is worth $$225 billion, and Adobe is over $230 billion.

And though cloud computing was once somewhat controversial, Buffett’s interest now shows it has become “a part of the traditional investor's canon,” Valoir analyst Wettemann notes.

CEO Slootman takes nothing for granted, however. “We have formidable competition, three of the four largest companies in the world, which makes this an interesting contest,” he says. “But if you’re going to be anywhere in the world of software, this is a heck of a cool place to be.”