Definition
Consumption-Based Pricing: A pricing model where the buyer pays per unit of resource consumed rather than per user or per server. Units vary by category: compute seconds on serverless, terabytes scanned on warehouse queries, DBU on Databricks, credits on Snowflake, tokens on AI APIs. Pricing is metered and billed in arrears, with optional committed spend tiers that buy discount in exchange for capacity reservation.
Consumption pricing won the cloud era because it aligned vendor revenue to customer workload. AWS built the model into its DNA from 2006 onward. Snowflake and Databricks rebuilt the data warehouse market on the same metric, separating compute from storage. Google Cloud and Microsoft Azure followed. Today consumption is the default for compute, storage, data, AI inference, and a growing share of application logic.
The commit mechanic is where most enterprise negotiation power sits. AWS EDPs trade 1 to 5 year spend commitments for 5 to 25 percent across service discount, with the deepest discounts at the 5 year tier and above 100 million dollars annual commit. Snowflake capacity contracts use the same logic in credit terms. The trade off is forecast risk: commit too high and the buyer pays for unused capacity; commit too low and the buyer pays on demand rates on the overage. Benchmark practice is to commit at 60 to 70 percent of expected baseline and let overage land at on demand or burst tiers.
Benchmark your consumption commits
We benchmark consumption commit discount levels across 2,300 cloud and data contracts. Send us your forecast and we return commit sizing intelligence in 48 hours.
The three biggest risks
The three biggest risks in consumption contracts are forecast error, runaway workloads, and renewal uplift. Forecast error shows up as unused commits that expire worthless at end of term, since most cloud commits do not roll over. Runaway workloads come from misconfigured queries, infinite loops, or AI agents calling themselves recursively. Renewal uplift hits once the buyer's consumption pattern is locked in: the vendor knows the floor, knows the trajectory, and prices the renewal accordingly. FinOps practice mitigates all three through tagging, budget alerts, query timeouts, and aggressive renewal positioning that benchmarks against alternative providers.
For related vocabulary, see the per token pricing definition, the per transaction pricing definition, the AWS EDP definition, the Google Cloud CUD definition, and the MACC definition. The glossary hub covers the broader pricing vocabulary. For cloud commit benchmark data, see the cloud infrastructure benchmark.
Frequently asked questions
What is consumption-based pricing?
Consumption based pricing charges buyers per unit of resource consumed. The unit varies by service: compute seconds for AWS Lambda, GB scanned for BigQuery, credits for Snowflake, DBU for Databricks. There is no fixed seat fee. Buyers pay only for what they use.
How do AWS EDPs and Snowflake commits work?
AWS Enterprise Discount Programs trade 1 to 5 year spend commitments for 5 to 25 percent off list across services. Snowflake capacity contracts commit dollars in advance for 15 to 35 percent off on demand pricing. Databricks commits use the same mechanic with DBU prepay at 20 to 40 percent off list.
What are the risks of consumption pricing?
Three primary risks: forecast error where commits go unused at end of term, runaway costs from misconfigured workloads or unbounded queries, and price uplift at renewal once consumption habits are entrenched. The mitigation is tagging, FinOps tooling, and committed spend that aligns to 60 to 70 percent of expected baseline.
Benchmark your contract
Send us the proposal or renewal. We return discount, term, and unit price intelligence in 48 hours.
