Apify compute units explained: pricing, examples and total cost
A compute unit measures allocated memory over time. Use it to budget a pay-per-usage Actor, then account for the other charges and the prepaid balance in your plan.
Last updated Nov 15, 2026
Prices checked November 15, 2026. Examples below are hypothetical, in USD, with monthly billing and no tax, discounts or add-ons. They are arithmetic examples, not scraping benchmarks.
How to calculate Apify compute units
Apify defines one CU as 1,024 MB of allocated memory running for one hour. Use the memory assigned to the Actor, not just the memory it happened to use. For repeated runs at the same allocation:
CU = (allocated memory in MB / 1024) × (run minutes / 60) × number of runs
Compute charge = CU × your plan’s USD per CUTwo GB for 30 minutes and four GB for 15 minutes both produce one CU. Giving an Actor more memory only reduces this bill if the runtime falls enough. A failed or retried request can still consume runtime; CU billing does not promise a fixed number of successful pages. See Apify's resource and CU definition.
Current monthly plans and CU rates
| Plan | Subscription | Prepaid usage | USD per CU |
|---|---|---|---|
| Free | $0.00 | $5.00 | $0.20 |
| Starter | $19.00 | $19.00 | $0.20 |
| Scale | $199.00 | $199.00 | $0.16 |
| Business | $999.00 | $999.00 | $0.13 |
Source: Apify pricing. Enterprise pricing is custom. Prepaid usage is a shared dollar balance, not a separate allocation for each service.
Three worked examples on Starter
Assume each workload repeats within one billing cycle, no other billable usage, and a usage limit that permits the spend. These examples isolate compute to show the difference between a usage charge and a subscription payment.
| Memory | Minutes / run | Runs / cycle | CU | Compute charge | Starter subtotal |
|---|---|---|---|---|---|
| 512 MB | 20 | 30 | 5 | $1.00 | $19.00 |
| 2,048 MB | 15 | 120 | 60 | $12.00 | $19.00 |
| 4,096 MB | 45 | 60 | 180 | $36.00 | $36.00 |
Scroll the example table sideways to see the charges and subtotal.
In the last row, 4 × 0.75 × 60 = 180 CU. At $0.20/CU, the compute charge is $36. The $19 subscription covers the first $19, leaving $17 of overage: $36 in total, rather than $19 + $36. In the first two rows, the subscription payment is still $19.
For an unchanged paid monthly plan, a simplified subtotal is the subscription plus any eligible usage above its prepaid balance. Free has no paid overage; paid-plan overage requires a sufficient usage limit. Unused credits expire. See the subscription and overage rules.
Check the Actor's pricing model first
- Pay per usage: budget the platform resources the Actor consumes.
- Pay per event: price the events listed by its developer, which may include starting a run or returning a result. Most include platform usage in that price.
- Events plus usage: some Actors charge both. Add separate resource costs only when the Actor's pricing says they apply.
Read the Actor's Pricing tab before applying a CU estimate. Dataset operations after a run can still incur charges across these models. Apify documents these distinctions here.
Turn a test run into a budget
CU is only one part of platform usage: proxies, data transfer and storage can add costs. Start with a small, representative run and inspect its resource breakdown. Record both spend and usable results; a cheap run that returns duplicates or incomplete records is not an equivalent workload. Repeat the sample for materially different target sites.
For a rough extrapolation, divide the sample's applicable usage cost by usable results, then multiply by the results you need. Treat that as a planning estimate: retries, site mix and run size can change the ratio. Add other activity in the same billing cycle before applying the shared prepaid balance once.
Check Billing → Current period for the actual cycle and its service breakdown. Historical usage is grouped by calendar month, which may differ from that cycle; invoices are the reference for past charges. See Apify Billing.
How this differs from our page-cost calculator
ScrapeSelect's Apify model estimates CU from page attempts and a rendering profile. Its current assumptions are 0.8 CU per 1,000 HTTP attempts and 6 CU per 1,000 browser attempts, with premium-proxy traffic added where applicable. These are modeling inputs, not Apify's per-page tariffs or measurements of your Actor.
That model does not itemize Store event charges, storage or separate data-transfer fees. Use it for an initial comparison, then replace assumed throughput with a representative run before budgeting a real Actor.