Modern Pricing & UBB
February 26, 2026 ubb

Snowflake's Consumption Model Is Both the Best and Worst Idea in SaaS

Snowflake's IPO in September 2020 was the largest software IPO in history at that point — $3.4 billion raised at a $33 billion valuation. The business was growing 120% year-over-year. The pricing model was pure consumption: you buy compute credits, you run queries, the credits drain. No seats. No tiers. No negotiating which features you get. Just usage. Wall Street called it a revelation. Every SaaS company in Silicon Valley started quietly asking if they should switch to credits.

Then 2023 happened. And Snowflake taught the other half of the lesson.

The Genius of Compute Credits

Snowflake's credit system is elegant. You pay for virtual warehouses — compute clusters that run your SQL queries. The bigger the warehouse, the more credits it burns per hour. A small (XS) warehouse costs 1 credit per hour. A 4XL burns 128. You can spin them up and down instantly, auto-suspend them when idle, and right-size for each workload.

What this model got right was aligning cost with value. A startup running a few dashboards pays pennies. A Fortune 500 running continuous data pipelines across petabytes pays millions. The same product, the same pricing model, infinite scalability across customer size. You never need a separate enterprise contract structure — large customers just consume more.

The growth mechanics this enables are extraordinary. Snowflake's net revenue retention (NRR) consistently ran above 150% through its early public years. That means existing customers, without any new product purchases or upsells, were spending 50% more year-over-year — purely because their data workloads grew. That's what usage-based pricing looks like when product-market fit meets compounding consumption. It is genuinely one of the best business model designs in the history of enterprise software.

The 2023 Problem

In August 2023, Snowflake cut its full-year product revenue guidance. The stock dropped 20% in a day. The cause: customers had gotten better at using Snowflake. They were optimizing queries, clustering tables, using materialized views, and generally doing the engineering work to consume fewer credits for the same analytical output. Snowflake had spent years making their product easier to use efficiently — and now customers were using it efficiently.

This is the paradox at the heart of pure consumption models. If your product helps customers do more with less, your revenue shrinks per unit of value delivered. Snowflake's Q3 FY2024 earnings call was illuminating: CFO Mike Scarpelli noted that "customers are optimizing their spend" as a primary factor in the guidance revision. He was describing rational customer behavior as a headwind. That's not a bug in Snowflake's model — it's a structural feature of pure consumption pricing that every UBP company needs to internalize.

"Customers are being very thoughtful about their Snowflake consumption. We are seeing customers optimize their workloads, which is impacting near-term growth." — Snowflake Q3 FY2024 earnings call

What Every UBP Company Should Learn From This

Lesson 1: Pure consumption models need consumption growth floors

Snowflake's fix was to accelerate new workload adoption — more customers running more distinct use cases, so optimization of one workload doesn't crater revenue. They've invested heavily in Cortex (AI/ML features), Snowpark (Python workloads), and Streamlit (app development) specifically to generate new credit consumption surfaces. The lesson: if your pricing is purely consumption-based, your growth strategy must constantly expand what customers consume, not just deepen existing workloads.

Lesson 2: Committed spend is the hedge

Snowflake has moved significant enterprise revenue onto committed spend contracts — minimum annual spends that customers pre-purchase at a discount. This creates a revenue floor that pure consumption doesn't provide. The committed spend model preserves the usage-based upside (customers who exceed their commit pay overage) while eliminating the optimization risk (if they under-consume, the revenue is already recognized). Most mature UBP companies eventually arrive here.

Lesson 3: NRR above 130% is unsustainable long-term

Snowflake's NRR above 150% was always going to normalize. That level of consumption growth requires customers to be in the early stages of data adoption — moving from nothing to something. Once customers are mature Snowflake users running optimized workloads, NRR settles toward 110-120%. Model this in your projections. The first three years of a usage-based company look extraordinary. Years four through seven are when the model actually matures, and Wall Street's expectations often don't account for that normalization.

Snowflake remains one of the most impressive pricing model stories in enterprise software history. The credit system that took them to a $33 billion IPO is still working. They're now a $15 billion revenue-run-rate business growing at 29%. They figured out the optimization problem, expanded into AI, and maintained best-in-class NRR. But they also taught every usage-based SaaS company a lesson in 2023 that no business school case study could: when your customers get smarter, pure consumption pricing makes them your adversary. Design accordingly.

Sources

← Usage-Based Billing · All posts