Editorial note. Snowflake per-credit figures are the commonly cited on-demand list rates for AWS US regions; Snowflake negotiates capacity per customer, so treat them as indicative. Datadog rates are from datadoghq.com/pricing. Both as of September 2026; neither company is affiliated with Sonic AI.
Snowflake's bill is the model every "consumption-based" enterprise contract is copied from: usage is metered in an abstract unit (a credit), the unit has a price that depends on the edition, and the customer chooses between paying list price as they go or pre-purchasing a year of capacity at a discount. Once a B2B contract crosses a few hundred thousand dollars a year it stops being a price list and becomes this kind of commitment — and there is a second kind, Datadog's, that looks identical on a quote and bills completely differently, covered further down.
What Snowflake charges
| Component | On demand | Capacity (pre-purchased) |
|---|---|---|
| Compute — Standard | ≈ $2.00 / credit | Discounted per-credit rate on an annual commitment |
| Compute — Enterprise | ≈ $3.00 / credit | Large commitments cited at 25–45% off |
| Compute — Business Critical | ≈ $4.00 / credit | Same structure |
| Storage | per TB-month; ≈ $23 under capacity on AWS US East, higher on demand | Lower rate under contract |
Credits are consumed per second of warehouse runtime scaled by size; on demand is invoiced monthly in arrears, capacity is invoiced up front and drawn down.
The model underneath — and the other kind of commitment
Prepaid drawdown (Snowflake capacity) is a usage slab at the scale of a year: a committed price buys a quota of units, a running balance carries across months, usage is priced at the contracted rate and deducted, and overage beyond the quota has its own rate. The customer gets one invoice and then eleven statements. Revenue is recognised as the balance is consumed, not when the invoice is paid.
Minimum commitment (Datadog annual) is a floor with a per-unit overage: the committed quantity at the committed rate is billed each month in advance, usage above it at the on-demand rate in arrears, and each covered meter has its own commitment and its own pair of rates. The customer gets a monthly invoice with two rates for the same unit on it.
What a month looks like
Enterprise, 1,200 credits used in September, capacity contracted at $2.40 for 12,000 credits/year:
| On demand | Capacity | |
|---|---|---|
| Invoice at term start | — | $28,800 (12,000 × $2.40) |
| September | 1,200 × $3.00 = $3,600 invoiced | 1,200 × $2.40 = $2,880 drawn; balance 9,600 → 8,400 |
| Term end | — | Unused balance expires; usage past 12,000 billed per contract |
Who else sells commitments — Datadog, and the minimum-with-overage version
Snowflake's capacity is a prepaid drawdown. Datadog is the other enterprise commitment, and sales teams use the word "commit" for both: promise a quantity of hosts and events, pay for it monthly whether you use it or not, and pay a higher on-demand rate for anything above. Same quote page, different documents, different accounting. Databricks and AWS Savings Plans are Snowflake-shaped; New Relic, Segment and most annual SaaS contracts with an overage clause are Datadog-shaped.
Datadog
| Product | Unit | Annual commitment | On demand |
|---|---|---|---|
| Infrastructure — Pro | per host / month | $15 | $18 |
| Infrastructure — Enterprise | per host / month | $23 | $27 |
| Logs — ingestion | per GB | $0.10 | $0.10 |
| Logs — indexed, 15-day | per million events | $1.70 | $2.55 |
On-demand rates run roughly 20–33% above committed ones; billable host counts derive from usage during the month.
Datadog — Enterprise, committed to 200 hosts and 50M indexed events, ran 236 hosts and 61M events:
| Line | Qty | Rate | Amount |
|---|---|---|---|
| Infrastructure — committed, September (in advance) | 200 hosts | $23.00 | $4,600.00 |
| Infrastructure — above commitment (in arrears) | 36 hosts | $27.00 | $972.00 |
| Logs indexed — committed | 50M events | $1.70 | $85.00 |
| Logs indexed — above commitment | 11M events | $2.55 | $28.05 |
| Logs ingested | 820 GB | $0.10 | $82.00 |
| September | $5,767.05 |
Two rates for the same unit on the same invoice, split at the commitment. A system with one rate per product either under-bills the overage or over-bills the base.
Where commitments go wrong
- Calling both "commit". A drawdown produces a balance and statements; a minimum produces a floor and true-ups. The contract has to say which, because the documents and the journal differ.
- Contracted rates per customer. Every capacity customer has their own per-credit price. The catalog holds list; the contract holds the override; the statement must show the override.
- Commitments per meter, not per contract. Datadog commits hosts, events and GB separately. A shortfall on one does not offset an excess on another unless the contract says it does.
- Level meters need a rule. "Hosts this month" comes from a high-water mark over hourly readings. Define it; the customer will ask.
- Annual promise, monthly invoice. The floor has to be scaled to the invoice period and the shortfall or excess carried the way the contract says.
- Cash up front, revenue over time. Billed and recognised diverge on purpose; the journal must show both.
Billing this shape on Sonic AI
Snowflake capacity is a usage slab on Sonic with an annual commitment period: the slab carries the committed price and units, the billing schedule carries the running balance across months, and usage above the quota bills at the slab's overage rate. The commitment invoices in advance; monthly documents show consumption against the balance; revenue recognition reads consumption, so the waterfall shows the commitment releasing as it is used. Datadog annual is a committed line billed in advance plus a linear overage line in arrears on the same meter, grouped under a minimum commitment so the engine knows the floor and computes the true-up per period — each committed meter with its own pair of rates, the annual floor scaled to each monthly invoice. Host counts and credits arrive through a usage template or a Postgres sync.
See usage & seats, revenue recognition, minimum commitment billing and billed is not recognised.
If you price like Snowflake
- Is it prepaid and drawn down, or a minimum with a true-up? Different models, different documents.
- Where does the contracted rate live, and does the statement show it?
- Is the commitment per meter or pooled?
- Does revenue recognition follow consumption rather than the up-front invoice?
Bring a capacity or committed contract to a demo — we will schedule the floor, the drawdown and the journal from the PDF.
Similar pricing models
Prepaid capacity that usage draws down is the model behind Databricks, AWS Savings Plans and enterprise OpenAI agreements; a committed quantity with on-demand overage — Datadog (above), New Relic, Segment — is a minimum commitment. Committed annual volume on a per-unit rate is the enterprise tier of Twilio; an annual seat commitment with an uplift is Salesforce.
Related terms: minimum commitment, overage, deferred revenue, billed vs recognised.