When Snowflake Computing was based 10 years in the past, the massive knowledge market appeared a lot completely different than it does at the moment. Momentum was constructing behind one thing known as Hadoop, whereas cloud computing was considered with suspicion. Regardless of these headwinds, the Snowflake founders caught to their preliminary imaginative and prescient, and finally performed a serious function in flipping the massive knowledge script and discovering large success alongside the best way. However the place will this success result in over the subsequent decade?
The primary two years of Snowflake Computing’s existence was spent in stealth mode. Co-founders Benoit Dageville, Thierry Cruanes, and Marcin Zukowski had been all knowledge warehousing veterans who beforehand labored at Oracle, IBM, and Actian, and they also had an insider’s view of the boundaries of knowledge warehouses. By working an information warehouse within the cloud and separating compute from storage, they believed, they might overcome these limits.
Snowflake’s first entry into the general public eye was modest sufficient. On October 21, 2014, the corporate concurrently got here out of stealth and introduced a Collection B spherical of financing led by Redpoint price $26 million, which isn’t a big quantity by at the moment’s enterprise capital requirements. Former Microsoft government Bob Muglia, who was tapped to be its first CEO, unveiled its first industrial providing, dubbed the Elastic Knowledge Warehouse, which ran completely on Amazon Net Companies and have become usually obtainable in 2015.
The subsequent a number of years had been spent in a heads down mode, iterating on that first launch of the Snowflake knowledge warehouse working on AWS, and banging the drum for its model of huge knowledge processing within the cloud. The corporate took residence first place within the Strata + Hadoop World Startup Showcase in 2015, and raised a $45 million Collection C spherical later that 12 months (it will subsequently be expanded to $76 million). It partnered with BI distributors like Tableau, Looker, and MicroStrategy; railed in opposition to failed huge knowledge initiatives; and promoted its cost-savings in opposition to different nascent cloud distributors.
A feisty Muglia took photographs at rivals, together with Hadoop. The startup’s CEO tore the open supply software program undertaking to shreds in a 2017 Datanami interview, earlier than the Hadoop implosion appeared imminent. “I can’t discover a joyful Hadoop buyer. It’s type of so simple as that,” he stated. “It’s very clear to me, technologically, that it’s not the know-how base the world will probably be constructed on going ahead.”
Momentum began constructing for Snowflake in 2017 with a pair of latest options. The primary newly launched product was Snowpipe, a steady knowledge ingestion functionality, and the second is the start of knowledge sharing.
That 12 months, Snowflake laid “the fundamental underlying constructing block that enables two separate Snowflake accounts to collaborate over shared knowledge belongings in a significant, safe, and well-governed method,” says Torsten Grabs, Snowflake’s director of product administration for knowledge lake, knowledge pipelines, and knowledge science.
The corporate’s buyer progress signaled confidence to potential traders, and in April 2017, the corporate accomplished a $100 million Collection D spherical, bringing the fledgling agency’s whole funding to $205 million.
Buoyed by stable progress, Snowflake expanded to Microsoft Azure, which opened the corporate to many extra firms that run on that cloud infrastructure. By late 2018, firm surpassed the 1,000-customer mark, barely three years into its existence. It additionally began getting extra severe about the way it companions by way of the launch of the Companion Join program, “which permits customers to spin up a Snowflake account by way of a companions expertise and vice versa,” Grabs says. Snowflake closed two large founding rounds in 2018, together with a Collection E price $263 million in January and a $450 million spherical in October. At that time, it was valued at $3.5 bilion.
2019 would mark a interval of transition for the corporate, which was nonetheless known as Snowflake Computing and was nonetheless primarily based in San Mateo, California. On the product entrance, it expanded into Google Cloud. It additionally launched the Snowflake Knowledge Market, in addition to the cross-cloud constructing blocks that may finally be often known as Snowgrid. On the enterprise entrance, Muglia was changed by Frank Slootman, a former ServiceNow government, because the CEO in Might 2019.
The COVID-19 pandemic was a gut-punch to many firm’s plans in early 2020, however Snowflake appeared to roll proper together with the punches. The corporate launched a number of new merchandise at its annual person convention, together with the supply of Snowsight, a brand new GUI designed to let customers get nearer to the information. It additionally launched Snowpark, which might give customers the flexibility to work with Snowflake knowledge in a language apart from SQL. Lastly, in a nod to the significance of companions, it launched a brand new formal partnership program, dubbed Snowflake Companion Join.
And who may overlook Snowflake’s huge debut on the New York Inventory Change below the ticker image SNOW? That September IPO raised $3.4 billion (giving the corporate a $33 billion valuation) and was dubbed by the mainstream press as “the biggest ever IPO for a software program firm” (though it’s a cloud providers supplier). The corporate additionally doubled its buyer base, from a rely of 1,550 in July 2019 to three,100 in July 2020.
2021 was one other busy 12 months for the newly public firm, with the launch of Snowpark for Python, which remains to be in public preview. Snowflake additionally expanded assist for unstructured knowledge, which is extra vital for the forms of AI use circumstances the place Python can be used (whereas conventional SQL queries run on structured, tabular knowledge). Snowflake additionally began permitting members of the Snowflake Market to monetize their knowledge. It additionally launched knowledge clear rooms, in addition to the primary two vertical clouds, for media and monetary providers. We are able to’t overlook the temporary time in early 2021 when Snowflake recognized as “headquarterless,” earlier than Slootman settled down in Bozeman, Montana.
Snowflake’s evolution continued in 2022, with a number of notable unveiling at its June person shindig, together with Unistore, its first storage repository for transactions; enhancements to Snowpipe for streaming knowledge pipelines; and a personal preview of its new knowledge utility framework primarily based on its $800-million acquisition of Streamlit in March. The corporate additionally introduced assist for Apache Iceberg, the open desk format that’s gaining momentum; launched new vertical clouds for healthcare and life sciences, in addition to for retail; and likewise debuted a brand new safety providing.
Snowflake at the moment has 6,000 clients, and with a market capitalization in extra of $55 billion, is taken into account one of many cloud giants, a title it must share with one other post-Hadoop breakout huge knowledge star, Databricks. The corporate introduced in additional than $1.2 billion in income in fiscal 2022, but it’s struggling to please Wall Road, which has pushed its inventory value right down to about $180 per share, lower than half of its November 2021 all-time excessive. And whereas there are grumblings about surprising prices from clients, clearly, the variety of clients Snowflake has exhibits that it’s doing one thing proper.
Not content material to offer clients with instantaneous entry to limitless SQL compute on large knowledge units in an information lake surroundings, Snowflake is taking part in the massive knowledge lengthy recreation and positioning itself for the subsequent huge factor. For Grabs, who joined the corporate in 2017, it’s much less of a shift away from conventional knowledge warehousing than a continuation of the corporate’s unique path.
“To me, it doesn’t really feel as such of a dramatic shift of the place we had been initially, as a result of already to start with, Benoit and Thierry had been consider Snowflake as an information lake providing,” he says. “They had been intentional about eager about Hadoop as one other huge knowledge processing platform that Snowflake from the very get go ought to compete properly with.”
Hadoop was the massive competitor in these early days, and Snowflake spent simply as a lot time doing Hadoop replacements as greenfield enterprise knowledge warehouse installations, Grabs says. The truth that Snowflake has soared whereas Hadoop has fallen is certainly pertinent to this dialog. “We’re the higher Hadoop,” Grabs quips.
However the place will the corporate go subsequent? The corporate has put stakes within the floor in a number of areas instantly adjoining to the world of superior analytics, together with AI, streaming knowledge, knowledge purposes, converged OLAP/OLTP, knowledge clear rooms, and vertical knowledge clouds. Which of those will outline Snowflake 10 years from now?
That reply shouldn’t be clear, however one factor is: The corporate gained’t be sitting nonetheless. “We’ve to innovate every day,” Grabs says. “We can not sit on the laurels of what was finished prior to now.”
Grabs likes to remind clients that, yearly, the window of time they must course of new knowledge and decide is getting smaller and smaller. As that window goes down, the amount of knowledge goes up, and latency calls for get tighter and tighter. These are a few of the enterprise challenges which can be driving plenty of funding in streaming knowledge analytics and real-time databases. Snowflake can also be monitoring this problem, and in search of methods to maintain clients on prime of it.
“We’re getting very inventive about completely different storage layouts and the way we bodily signify storage. We’re under no circumstances a columnar retailer solely,” Grabs says. “It’s additionally motivating our funding into materialized tables, dynamic tables that primarily replace as new knowledge arrives and likewise fairly frankly for hybrid tables, with our Unistore workloads, which provides a special latency profile, response-time profile, than what an everyday Snowflake desk does.”