Snowflake Charges for the Gas, Not the Car. That's the Trap and the Genius.
Most enterprise software sells seats; Snowflake sells consumption — and it grew revenue per customer for years on that bet. But the same meter that expands also means every efficiency Snowflake ships can shrink its own revenue, a tension now in federal court.
Comes with a free Pricing Power Diagnostic template — plus a worked example for Snowflake.
Most enterprise software companies sell you a seat. You pay a fixed price per user per year, you use the tool a lot or a little, and the bill doesn't move. Snowflake sells you a meter. It charges for the compute you run, the storage you fill, and the data you move — and only for the resources you actually use.1 It is the difference between selling someone a car and selling them the gas. The car company gets paid once. The gas company gets paid every time the customer drives, drives farther, drives more often. Snowflake bet the entire business on the gas.
The official story is that this is simply a more honest, more customer-friendly way to price software — you only pay for what you use, so what's not to like. That framing is half true and quietly incomplete. The consumption model is genuinely a structural advantage for data workloads. It is also a structural liability that no seat-licensing company will ever face — and Snowflake is now defending itself against that liability in federal court.
“A customer-centric, consumption-based business model, only charging customers for the resources they use.”1
Why the meter beats the seat for data work
Data workloads are the worst possible thing to price by the seat, because the value isn't in the headcount — it's in the volume. A bank with fifty analysts can run a thousand times more queries than a startup with the same fifty analysts, and a seat license would charge them the same. Worse, a seat license punishes exactly the customer you want: the one whose data grows, whose questions multiply, whose workloads compound. Snowflake's pricing does the opposite. As a customer's data estate grows, the meter spins faster, and revenue grows with the workload rather than the org chart. That is why the model expanded revenue per customer year after year — and the proof is in a single metric the company is required to disclose.
Net revenue retention above 100% is the consumption model's whole thesis made visible: it means the existing customer base, on net, spends more this year than last — expansion outrunning churn. The figure has cooled from the triple-digit-plus levels of Snowflake's hypergrowth years to roughly 124–126% across recent quarters45, but cooling is not failing. It still means the average dollar in the door last year came back as more than a dollar this year, without anyone signing a new contract. Stack that for five years and you get the headline numbers: product revenue of about $4.47 billion for fiscal 2026, up 29%4, and $1.33 billion in a single quarter that the CEO called the strongest sequential dollar growth in the company's history.5
| Seat license | Snowflake's meter | |
|---|---|---|
| What you pay for | Number of users | Compute, storage, data moved |
| When the customer grows their data | Bill stays flat | Bill rises with the workload |
| Who benefits from heavy use | The customer (free upside) | Both sides |
| What an efficiency gain does to vendor revenue | Nothing | Can reduce it |
Isn't consumption revenue just a coin flip you can't forecast?
The standard objection is that usage-based revenue is unforecastable noise — if customers can use a lot or a little at their own discretion, how does anyone run a public company on it? It's a fair worry, and Snowflake's own filings concede the variability: product revenue is the company's key metric precisely because it is recognized on consumption that is, in its words, inherently variable at customers' discretion.8 But the model is far more disciplined than the caricature. The majority of revenue comes from capacity arrangements — pre-committed dollars billed annually in advance — not ad hoc on-demand usage.3 And the company doesn't forecast blind: its CFO has said it predicts revenue off daily customer consumption data flowing through an internal platform it calls Snowhouse, and has argued for going all in on consumption rather than a hybrid, because a hybrid lets customers arbitrage between models and undermines predictability.6 Remaining performance obligations — contracted dollars not yet recognized — stood at $9.21 billion at the end of Q1 FY2027.5 That is not a coin flip. It is a metered business with a forward order book.
The trap nobody who sells seats ever has to face
Here is the part the customer-friendly story leaves out, and it is the sharpest edge of the whole bet. When you charge for consumption, every improvement that makes your platform faster, leaner, or cheaper to run is a knife pointed at your own revenue line. Make a query 30% more efficient and you have just cut 30% off what that query bills. A seat-licensing company never faces this — Salesforce gets paid the same whether your reps work the CRM hard or barely log in. Snowflake's incentives are structurally tangled: it must keep building efficiency to keep customers loyal, while knowing that efficiency directly suppresses the meter it lives on. That is not a hypothetical. It is the core of an active securities class action, which alleges Snowflake failed to disclose that its own product improvements — specifically Iceberg Tables and revised tiered storage pricing — reduced customer consumption and therefore revenue. When those headwinds surfaced, the suit notes, the stock fell roughly 18% in a single day.7
Usage-based pricing aligns you with the customer better than any seat license can — they pay for value, you grow when they grow. But it quietly creates a conflict no flat-fee vendor ever has to manage: your product team's job is to make the thing more efficient, and efficiency is exactly what shrinks the bill. The most dangerous version of this isn't the revenue hit itself — it's the disclosure trap. Build a genuine improvement, ship it because customers love it, watch consumption dip, and you may have just created a fact you were legally obligated to warn investors about before it showed up in the numbers. The meter that aligns you with customers can misalign you with your own shareholders. Price on consumption only if you're prepared to narrate every efficiency gain as a potential headwind — out loud, in advance.
Whatever the court decides about disclosure, the strategic shape of the bet is already clear. Snowflake chose the model that grows with its customers' success rather than their headcount — and that choice has expanded revenue per customer year after year, exactly as advertised. But it also bound the company to a permanent, unnatural discipline: it must keep making its own platform cheaper to use, and then keep telling its investors that doing so might cost it money. A seat company sells the car and walks away. Snowflake sells the gas, and so it has to care, forever, how efficiently the engine burns it — and confess, every quarter, that a better engine is the thing it most has to fear.
Pricing Power Diagnostic
A scored diagnostic of pricing power: brand pull, switching costs, substitutes, and how critical the product is to the buyer. Each dimension rated 1-5 so you can see, at a glance, whether a price rise sticks or sends customers running. Blank to grade your own offer; filled as the worked example scoring a story's business on its real ability to charge more.
The worked example unlocks with a subscription. See plans →
Sources
Where this comes from — the filings, records, and reporting behind it.
- 1Snowflake's S-1 explicitly describes its pricing as 'a customer-centric, consumption-based business model, only charging customers for the resources they use.'
- 2Snowflake's IPO priced at $120.00 per share; the company issued 32,200,000 shares (including 4,200,000 upon exercise of the overallotment option) and received net proceeds of $3.7 billion after underwriting discounts.
- 3Snowflake's FY2021 10-K states customers choose consumption under either capacity arrangements (majority of revenue, billed annually in advance) or on-demand arrangements (charged monthly in arrears), making pure 'unpredictability' a partial myth.
- 4For full-year fiscal 2026 (ended January 31, 2026), Snowflake reported product revenue of approximately $4.47 billion, up 29% year-over-year; Q4 FY2026 product revenue was $1.23 billion, up 30% YoY; NRR was 125%; and RPO was $9.77 billion, up 42% YoY.
- 5For Q1 FY2027 (ended April 30, 2026), Snowflake reported product revenue of $1.33 billion, up 34% YoY — 'the strongest sequential dollar growth in our history' per CEO Sridhar Ramaswamy — with NRR of 126% and RPO of $9.21 billion, up 38% YoY.
- 6Snowflake CFO Michael Scarpelli stated the company forecasts revenue using daily customer consumption data via an internal platform called 'Snowhouse,' and advised going 'all in on the consumption model rather than doing a hybrid' because hybrids allow customer arbitrage that undermines revenue predictability.
- 7A securities class action lawsuit alleges that Snowflake failed to disclose that product efficiency improvements — including Iceberg Tables and revised tiered storage pricing — negatively affected customer consumption and revenues, with a lead plaintiff deadline of April 27, 2026; the stock fell 18.14% in a single day when the headwinds were disclosed.
- 8Snowflake's product revenue metric is explicitly defined by the company as its key metric because revenue is recognized based on platform consumption — 'inherently variable at our customers' discretion' — and not based on contract duration, with revenue primarily derived from compute, storage, and data transfer.