Calculator
Snowflake cost calculator
Snowflake bills compute by the second, in credits, and the number sneaks up on people. Put in your warehouse size, how long it runs, and your storage, and get a realistic monthly and annual estimate — plus a sense of what a leaner setup would cost. No email required. Below the calculator: how credit pricing works, what drives the bill, and the worked example behind these defaults.
Estimated Snowflake cost
$2,135 / mo
$25,620 per year
Most of that is compute you rent by the second. If this warehouse runs most of the workday, you're paying elastic pricing for a steady workload. A right-sized always-on server, or a self-hosted engine like ClickHouse, DuckDB or Postgres that could handle it, often runs a flat $150–$500/month — and doesn't meter your queries.
A rough estimate to start a conversation, not a quote. Your real number depends on query patterns, concurrency and data size.
1. Check it yourself. An estimate is only as good as the hours you typed in. Measure what your warehouses actually do — the SQL is on that page, and it takes ten minutes.
2. Or send me the export. I'll tell you what I'd investigate first — no call required.
Background: why the bill is mostly overhead.
Found something strange in your Snowflake bill?
Send me the cost export. I'll tell you what I'd investigate first — no call required. If it's interesting, I'll ask before publishing an anonymised teardown.
What to send, and what you get back — a written reply within 3 business days.
Not ready for a call? Take the numbers with you.
I'll email you this breakdown plus the free 18-check teardown — the first section is entirely about where warehouse money actually goes.
The breakdown with your numbers, sent at once. No confirmation step.
What breaks and what it costs — pipelines, warehouse bills, and the failures that only show up in production. A few a month, never padded to hit a schedule. No sequence, no pitch deck. Reply 'stop' once and you're off — it reaches me, not a queue.
Credit consumption per warehouse size is from Snowflake's published figures (each size up doubles the credits per hour). Cost per credit and storage rates vary by edition, region and contract — adjust them to match your account.
How Snowflake pricing works
Snowflake splits the bill in two. Compute is measured in credits: a virtual warehouse consumes credits at a fixed rate for its size, for every second it is awake, and you pay whatever a credit costs on your edition. Storage is billed separately, per terabyte per month, on the compressed size of your data plus whatever time travel and fail-safe are holding.
Two smaller lines finish it. Cloud services — the layer that compiles queries, keeps metadata and serves the result cache — only reaches the invoice when it goes past a daily allowance tied to your warehouse usage, so most teams never see it. Data transfer is charged when data leaves the region or the cloud it lives in. Neither is where a surprising bill comes from. Compute is.
On-demand US list prices are roughly $2 a credit on Standard, $3 on Enterprise and $4 on Business Critical, with storage around $23 per terabyte per month on demand and less if you pre-pay capacity. Those are list prices at the time of writing and they vary by edition, cloud and region — a negotiated contract changes them again. Check Snowflake’s pricing page and your own contract before you rely on a figure. The calculator starts at $3 a credit and $23 a terabyte; change the credit price to match your edition and the estimate moves with it.
Every size step doubles the rate
Warehouse sizes are a doubling ladder. Each step up doubles the credits per hour, which doubles the cost of the same running time. It is worth paying for only when the bigger warehouse finishes the work in half the time — which is true for a large scan and often false for a query that is waiting on something else.
Billing is per second, with a 60-second minimum every time a warehouse resumes. A warehouse that wakes up forty times a day for eight-second queries is billed for forty minutes. And a warehouse keeps billing after the last query until auto-suspend shuts it down — so the suspend setting, not the query, decides most of the compute bill.
What drives the bill
These are the 5 things I find most often when I read a warehouse cost export. None of them is a bad decision anyone made on purpose; elastic pricing just rewards leaving things on.
- A warehouse that never suspends
- Auto-suspend only saves money in gaps longer than the suspend window. A warehouse set to sleep after ten minutes, queried every four, never sleeps — and bills as if it were always on, while the settings page looks correct.
- Dashboards refreshing on a timer
- A scheduled refresh keeps a warehouse awake all day for a report nobody opened this quarter. The cost is not the query; it is that the query arrives often enough to stop the warehouse shutting down.
- Dev, staging and CI on the production meter
- Test suites that spin up a warehouse per run are a real line item, and almost nobody has measured what share of the bill they are.
- A warehouse sized for the worst query
- Each size up doubles the credits per hour. That is only worth paying for when the query finishes in half the time. Often it does not, and the larger warehouse simply costs twice as much.
- Models that rebuild everything, every run
- A full refresh of a table where only the last day changed is the most common piece of compute that buys nothing.
The first one has its own calculator: what a warehouse that never suspends costs you works out the idle share from your auto-suspend setting and how often the warehouse is queried. The rest are the cost section of the Data Platform Teardown — 18 questions from the paid audit, published in full, with what a bad answer sounds like.
A worked example, step by step
This is the calculator's own starting position, worked by hand so you can see where the number comes from: one Medium warehouse, awake 8 hours a day, 22 working days a month, 1 TB of storage, at $3 a credit — the Enterprise list price, which is the assumption to change first if you are on a different edition.
- 1A Medium warehouse burns 4 credits an hour.
- 24 × 8 hours × 22 days = 704 credits a month.
- 3704 credits × $3 = $2,112 of compute.
- 41 TB × $23 = $23 of storage.
- 5Total: $2,135 a month, or $25,620 a year.
Storage is $23 of a $2,135 bill. Everything else is compute you rent by the second — which is why the hours you typed in matter far more than the size of your data, and why the cheapest fix is almost always fewer awake hours rather than a smaller warehouse.
What to do with the number
The point of this calculator is not to tell you Snowflake is wrong. Sometimes the elastic, zero-ops, multi-team warehouse is exactly what you should pay for. The point is to make the number visible before you decide — and to notice when a steady workload is sitting on pricing designed for spiky ones.
If that is you, a lot of it may fit comfortably on a leaner engine. I have written up both comparisons with the trade-offs, not just the price: DuckDB vs Snowflake for workloads one machine could hold, and ClickHouse vs Snowflake when you need a server and real concurrency. If DuckDB is the direction, the DuckDB fit assessment answers whether your workload actually suits it before you bet a migration on the answer.
Want the real number for your workload rather than an estimate? The Data Platform Audit is the paid version of this page. A fixed-price teardown of your data stack: 12–20 hours of senior work over one week — a ranked fix-list, a cost model comparing your spend to a lean rebuild, and a small proof-of-concept on your own stack. $3,000, and if day one shows no credible opportunity, you pay nothing.
Common questions
- How much does Snowflake cost?
- It depends almost entirely on how many hours your warehouses run. Compute is billed in credits: a Medium warehouse burns 4 credits an hour, so 8 hours a day over 22 days is 704 credits. At $3 a credit that is $2,112 a month of compute, plus $23 per terabyte per month of storage. Put your own numbers into the calculator on this page for an estimate.
- What is a Snowflake credit and what does one cost?
- A credit is Snowflake’s unit of compute. A running warehouse consumes credits at a fixed rate for its size, billed per second with a 60-second minimum each time it resumes. On-demand US list prices are roughly $2 a credit on Standard, $3 on Enterprise and $4 on Business Critical. Those are list prices at the time of writing and they vary by edition, cloud and region — check Snowflake’s pricing page and your own contract.
- How much does Snowflake storage cost per TB?
- Around $23 per terabyte per month for on-demand storage in US regions, and less if you pre-pay capacity. Storage is usually the small half of the bill: in the example on this page, 1 TB costs $23 a month against $2,112 of compute. If compute is more than about 70% of your bill, you are paying elastic, per-second pricing for what is really a steady workload.
- What does a Snowflake warehouse cost per hour?
- The credits per hour double at every size step: X-Small 1, Small 2, Medium 4, Large 8, X-Large 16, 2X-Large 32, 3X-Large 64, 4X-Large 128. Multiply by your credit price to get an hourly rate — at $3 a credit, an X-Small costs $3 an hour and a 4X-Large $384 an hour.
- How much does a single Snowflake query cost?
- Snowflake does not price a query; it prices the time a warehouse is awake. A query that runs for 30 seconds on a Medium warehouse still bills the 60-second minimum if it woke the warehouse, and costs nothing extra if the warehouse was already running for something else. That is why per-query costing misleads and why the useful unit is warehouse hours, not queries.
- How do I estimate Snowflake costs before I commit?
- Estimate the hours, not the queries. Decide how many warehouses you will run, how many hours a day each is genuinely awake — including the idle time before auto-suspend — and how much data you will store. Put those into the calculator on this page. Then check the estimate against a month of QUERY_HISTORY once you are live; the gap is nearly always idle time nobody counted.
- Why is my Snowflake bill higher than my estimate?
- Because the estimate counted query time and the bill counts warehouse time. Idle-but-awake warehouses, dashboards refreshing on a timer, dev and CI sharing the production meter, and full table rebuilds all bill compute that nobody asked for. Those four are where most of the difference lives, and all four are visible in your own query history.