Snowflake (2012-2024): from cloud-data-warehouse startup to $2.7 billion product revenue and Berkshire Hathaway’s first IPO investment in 60 years

Why the architectural insight mattered

The data warehouse category in the early 2010s was dominated by on-premises incumbents (Oracle, Teradata, IBM Netezza, Microsoft, Pivotal Greenplum) with architectures that tightly coupled compute and storage. When customers needed more compute capacity they had to add servers that included both processing and storage; conversely, when customers needed more storage they had to add servers that included both. The architectural coupling produced inefficient resource allocation and made capacity planning rigid.

Snowflake’s founders had spent years inside Oracle and other database-engineering organizations and saw the architectural inefficiency directly. Their thesis was that a cloud-native data warehouse could separate compute (virtual warehouses that could be sized independently) from storage (a single shared object-storage layer on S3 or equivalent), producing dramatically better elasticity, pricing flexibility, and operational simplicity. The product that shipped in 2015 was the first commercial cloud-native data warehouse with this architecture, and it produced rapid early customer adoption because the architectural benefits were directly observable in customers’ bills.

The IPO and the Berkshire investment

Snowflake’s September 2020 IPO was a remarkable event for several reasons. The IPO price was set at $120/share, valuing the company at approximately $33 billion at the offering. The stock more than doubled on the first day of trading, closing at $253.93 and giving the company a closing market cap of approximately $70 billion. The IPO was the largest software IPO ever at the time. The most-discussed element was the concurrent private placements: Berkshire Hathaway and Salesforce Ventures each invested approximately $250 million at the IPO price. Berkshire’s investment was notable because Warren Buffett had publicly resisted IPO investments for his entire career; Snowflake was the first IPO-period investment Berkshire had made since the 1956 Ford IPO 64 years earlier.

The Berkshire investment signaled mainstream-financial-conservative-investor validation of the cloud-data-platform thesis, and the IPO pricing reflected what may have been peak-cycle software valuation enthusiasm. Through 2021-2022 the stock traded down materially from the post-IPO peak as software multiples compressed across the sector. Through 2023-2024 the stock has stabilized and partially recovered as the operating fundamentals continued to outperform expectations.

The product-led adoption model and customer expansion

Snowflake’s commercial motion through 2018-2024 has been a hybrid product-led-and-sales-led model. The product was usable in self-serve trial mode (free $400 of credit, then pay-per-use pricing), which allowed individual data engineers and analysts to test the platform without enterprise procurement involvement. The consumption-based pricing model (customers pay for compute and storage they actually use, with no minimum commitments at the lowest tiers) reduced the activation friction further. The product-led adoption surfaced enterprise sales opportunities organically.

The customer-expansion economics have been unusually strong. Net revenue retention peaked at 178% in 2020 (meaning the average existing-customer dollar grew by 78% in trailing twelve months even net of churn) and has remained above 130% through 2024 — among the highest sustained NRR figures of any enterprise software company. The number of customers generating over $1 million in trailing-twelve-month product revenue has grown from a handful at IPO to 461 by fiscal 2024. Forbes Global 2000 penetration is now 691 customers (~35% of the Global 2000). The combination of high NRR plus high large-customer count is what produces the durable growth-and-economics profile.

How RGM thinks about cloud-native architectural disruption

When clients in infrastructure software ask about how cloud-native architectural insights produce platform-disruption opportunities, the Snowflake case is the structural example. Three structural lessons. First, the architectural insight has to produce directly-observable customer benefits, not just engineering elegance. Snowflake’s separation of compute and storage produced lower customer bills, faster query times, and simpler operations — benefits the customer could verify within weeks of trial. Architectural insights that require customers to internalize multi-year benefits typically do not produce the same adoption velocity. Second, the consumption-based pricing model aligns vendor and customer incentives in a way that the seat-based or capacity-based pricing of incumbents does not. Customers pay only for what they use, which reduces activation risk; vendors get paid more as customers succeed, which aligns long-term incentives. The CRR/NRR strength of Snowflake reflects the consumption-based pricing dynamic. Third, the IPO timing for cloud-native infrastructure software has been generally favorable in the 2017-2021 window, with Snowflake among the most successful. Subsequent capital-markets compression has reduced the IPO-pricing advantage but the underlying business-model strength of cloud-native infrastructure remains.

The pattern is generalizable to other cloud-native infrastructure categories (Databricks, MongoDB, Confluent, Datadog, HashiCorp). The structural conditions that produced Snowflake’s outcome — clear architectural insight, consumption-based pricing, product-led activation followed by sales-led expansion — are present in many of these adjacent companies. We tell clients in infrastructure-software categories that the combination of architectural differentiation and consumption-based pricing produces the highest-quality business models in modern enterprise SaaS, and that the long-term winners typically build on this combination.

Frequently asked questions

Related