Snowflake Comes Out of Stealth — Separating Storage from Compute Rewrites the DWH

On 21 October 2014, the San Mateo startup Snowflake Computing came out of stealth to announce the Snowflake Elastic Data Warehouse, along with US$26 million in total funding—a Series B led by Redpoint Ventures, joined by Sutter Hill Ventures (which had provided seed funding and led the Series A) and Wing Ventures. General availability came eight months later, on 23 June 2015.

Two years after Amazon Redshift (announced 2012, GA February 2013) had opened the door to the cloud DWH, Snowflake walked through it carrying its own design principle: separate storage from compute. Six years later it would mount an IPO worth roughly US$70.4 billion at the first day's close—the largest software IPO to that date.

Founders — Veterans of Oracle and Vectorwise

Snowflake's three founders sat at the heart of the recent history of commercial DBMS.

Benoit Dageville, French-born, was an Oracle architect who worked for years on Oracle Database's parallel execution and self-tuning machinery. Thierry Cruanes, also ex-Oracle, worked on the query optimiser. Marcin Żukowski led the vectorised-query-execution research that produced MonetDB/X100 at CWI and its commercial form, Vectorwise. The 2016 SIGMOD paper The Snowflake Elastic Data Warehouse carries all three as its lead authors.

In 2012 the three founded Snowflake Computing in San Mateo. The concept was a clean redesign of the data warehouse for the cloud era. No legacy RDB architecture to carry forward; design from scratch on AWS S3 and EC2. Former Microsoft executive Bob Muglia became CEO in June 2014, and stealth ended four months later. In the October 2014 release Muglia framed the opportunity by saying there had been shockingly little innovation in data warehousing in a decade.

The Core of the Design — Complete Separation of Storage and Compute

Snowflake's central innovation is the multi-cluster, shared-data architecture, a three-tier structure.

Storage layer. All data is held in Amazon S3 (later Azure Blob and GCS as well), in compressed columnar format. Compute is detached; storage is effectively unlimited, cheap, and 11-nines durable.

Compute layer. Queries run in "virtual warehouses" that can be started and stopped independently of storage. Sizes are selectable, and multiple warehouses can run in parallel against the same data—a BI warehouse, an ETL warehouse, and a data-science warehouse can each have their own compute resources without interfering with each other's performance.

Cloud services layer. A shared layer for authentication, metadata, query optimisation, and transaction management.

The three-tier split was decisive. Early Redshift fixed the ratio of storage to compute per node: add storage and you add compute, and vice versa. Snowflake removed that constraint, delivering an elasticity in which storage scales cheaply and indefinitely on S3 while compute spins up only when needed, billed by the second.

And storage is referenced as shared data across virtual warehouses. Different users, teams, or regions can hit the same physical data with separate compute, which is why the design is called "multi-cluster, shared-data".

2014-2020 — Quiet Rapid Growth

At the stealth exit in October 2014 Snowflake was still a little-known startup. The service ran only on AWS, and the press release claimed a 90 per cent lower cost than on-premises data warehouses—the vendor's own figure, not an independent benchmark.

Enterprise adoption thickened through the late 2010s. Customers named in the 2020 IPO prospectus include Capital One (whose subscription agreement dates from 30 June 2017), Adobe, Sony, Nielsen, McKesson, DoorDash, Instacart, Dropbox, and Akamai. As of 31 July 2020 Snowflake had 3,117 customers, up from 1,547 a year earlier, including 146 of the Fortune 500 and seven of the Fortune 10, with net revenue retention above 150 per cent.

Escaping cloud lock-in became a design differentiator of its own. Microsoft Azure support went to preview in July 2018 and GA that September; the Google Cloud partnership was announced in June 2019 and reached GA in February 2020. Customers had a DWH that was not tied to any single cloud provider.

2020 IPO — The Largest Software IPO to That Date

On 16 September 2020, Snowflake listed on the NYSE as "SNOW". The prospectus set the initial public offering price at US$120.00 per share for 28,000,000 Class A shares—above the $100-110 range set days earlier, itself raised from $75-85. The stock opened at $245 and closed the first day at US$253.93, up more than 111 per cent, valuing the company at roughly US$70.4 billion at the close. CNBC called it the largest software IPO ever.

The prospectus also records that Salesforce Ventures LLC and Berkshire Hathaway Inc. each agreed to buy US$250 million of Class A stock at the IPO price in a concurrent private placement; Berkshire additionally took 4.04 million shares in a secondary transaction.

The CEO at the time was Frank Slootman, formerly of ServiceNow, who took the job in 2019. He was succeeded in February 2024 by Sridhar Ramaswamy, previously a Google advertising executive. Slootman is chairman of the board and co-founder Dageville a director, per the signature page of the FY2026 Form 10-K.

Today — US$4.68 Billion in Revenue, the DWH for the AI Era

The figures come from the Form 10-K. FY2024 (year ended January 2024) revenue was US$2.81 billion, 36 per cent above the prior year's $2.07 billion; FY2025 was $3.63 billion; FY2026 (year ended 31 January 2026) was US$4.68 billion, of which product revenue was $4.47 billion. As of 31 January 2026 Snowflake had 13,328 customers (10,996 a year earlier), 733 customers with trailing-twelve-month product revenue above US$1 million (576 a year earlier), and 790 of the Forbes Global 2000, who contributed roughly 43 per cent of revenue. Net revenue retention was 125 per cent.

From 2023, Snowflake has been adding AI features fast: Snowpark (Python/Scala data processing), Snowpark Container Services (run arbitrary containers), Cortex (call LLM inference from SQL), Iceberg Tables (open table-format support). The competition with Databricks—data warehouse versus data lakehouse—has become the industry's biggest front line.

From Purchase Order to API Call

The October 2014 stealth exit was the moment the cloud-native DWH design pattern went public. Where Redshift transposed the on-premises DWH architecture onto the cloud, Snowflake answered a different question: what does a DWH designed for the cloud from scratch look like? That answer—separating storage from compute—has propagated to Databricks, ClickHouse Cloud, OSS data platforms built on Apache Iceberg, and even OLTP-side systems like Aurora, Neon, and PlanetScale.

If Redshift began renting a DWH by the hour, Snowflake pushed toward paying for compute only while it runs. October 2014 is a node in the decade-plus story in which data infrastructure went from a hardware purchase order to an API call.