Apify is killing rentals. If you run Actors all day, renting was the cheaper deal. Here is the math

You run Actors on a schedule, from a pipeline, or from another Actor. You paid a flat rental plus compute and the bill was flat. On October 1 Apify retires rental, and every row you pull becomes a line on the invoice. If your gut says that costs more, it does, from about 7,411 results a month.

Disclosure: we build Bytewells, a marketplace that kept rental pricing, so we have an interest here. That is why the method is public and the exceptions are in the article.

What Apify is changing

1. Late September 2026. Developers schedule new pricing for their rental Actors. Apify gives 14 days' notice on each change.
2. October 1, 2026. Rental pricing is retired. Actors that haven't been moved become free and earn their developers $0. If your pipeline depends on one, nobody is paid to maintain it.
3. From about October 10, 2026. The first pay-per-event prices take effect for developers who switched in late September. Your first bills at the new prices follow.

Your instinct, tested

Each rental Actor has a break-even: the monthly results above which pay per event costs more than the old rental plus the compute you paid. We know each Actor's rental and per-result price, but not its compute, so the chart uses the median compute per result from our runs (0.0018 compute units) for every Actor. The spread you see comes from their prices. The chart bins the 291 Actors that announced a per-result price. The median rental was $20 a month.

Half the Actors break even between 3,314 and 15,039 results a month, which is 2 to 8 runs a day at 65 results a run. Compute is charged per run, so smaller runs push the break-even up (see micro-runs). For 13 of the 291, the per-result price does not cover that median compute, so renting never wins (8 to 23 depending on the compute figure in the table below). An Actor that uses more compute per result than the median has a higher break-even, one that uses less has a lower one.

Break-even by plan

Each row uses that plan's per-result price for every Actor and that plan's compute price. A bigger plan lowers both sides, so the median stays between 6,709 and 8,117. That table has no plan fees and no choice of plan. The next section adds both.

What the plan discounts actually give you

Pay per event is priced per plan tier, and the tier comes with your plan. The plan fee is prepaid usage: you can spend it on Store Actors and compute, and you pay as you go above it.

Then we measured the discount each Actor gives per tier, against its own Free-tier price. Two groups: the 291 Actors you would move to, and the 100 most used pay-per-event Actors in the Store.

• Half of Actors give no discount at all. Only 46% to 53% of the Actors you would move to discount on any tier, so their median discount is 0% on Starter and 8% on Business.
• The ones that do discount, step it. Among them the average is 14%, 22% and 31% on the three plans. The top 100 Actors discount more, with a median of 36% on Business.
• The discount is not the whole price. The top 100 also list lower to begin with, $2.6 against $2.99 per 1,000.

The medians alone would suggest plans do nothing. The averages suggest a 7% to 16% saving. Neither is your Actor, so the chart and the calculator below take the discount profile as an input.

Where it hurts

1. It still climbs with every row

Same median Actor ($20 rental, $2.99 per 1,000 results at the Free-tier price, median compute), each side on its cheapest Apify plan. The pay-per-event line climbs in steps: a plan's fee is prepaid usage, so a month costs at least the fee, and a bigger plan only wins once its discount pays for the jump. The table treats this Actor as your whole bill. If one Scale or Business plan already covers many of your Actors, the fee is shared between them and each Actor's steps come later than shown.

Discounts bend the curve and do not flatten it. At 189,000 results a month the multiple is 5.7× with average discounts and 5.0× with the top 100's, and at 1 million it is about 8×. Counting plan fees and discounts, the median Actor's break-even is about 8,293 results a month with average discounts (7,658 with none, 8,629 with the top 100's), against 7,411 for the median of per-Actor break-evens on Starter. Below that, pay per event is the cheaper bill.

The multiple depends heavily on compute per result, which we don't know for your Actor. At 189,000 results a month with average discounts it is 11.8× at light compute (0.0006 CU a result), 5.7× at the median and 3.4× at heavy compute (0.0034). At the compute of runs with 4 to 10 results (0.0049) it is 2.5×. That last case is closest to the heavy account above, whose median run returns 4 results.

2. Your spikes are billed row by row

Heavy use is bursty. In one heavy account's run history, 51% of runs started in a single hour of the day. Under rental a batch that doubles adds its compute. Under pay per event it adds every extra row at the list price, which for the median Actor is about 8 times what that row's compute would have cost at the median compute per result, and about 3 times at the compute of runs with 4 to 10 results.

3. Your Actors go quiet if nobody moves them

An Actor that isn't moved by October 1 becomes free and its developer earns $0. Free is not the same as maintained. If a scraper breaks in November, the person who would fix it is unpaid. If one of yours is in that spot, get it listed.

4. The stack multiplies it

Power users don't run one Actor. The account whose week of runs we could read used 23 different Actors, 7 of them other developers' pay-per-event Actors. Every Actor that moves adds its own per-result line. The break-even is per Actor, and so is the bill.

5. Discounts are for large accounts, and you have to ask

Apify can agree custom discounts with large accounts. That is a negotiation, not a plan tier, and it does not show up in list prices. If you are many times above break-even, have that conversation first. The rest of this guide is for when it doesn't get you far enough.

The exception: micro-runs

Compute is charged per run, and pay per event is charged per result, so run size decides who wins. Pooling runs from all three partner accounts (316 Actors), compute per run rises about 3 times, from 0.025 to 0.079 compute units, between the smallest runs and the largest. Results per run rise far more, so compute per result falls. Break-even for the median Actor:

From 4 results a run up, renting wins at a volume between about 6,817 and 10,050 results a month. At 1 to 3 results a run, compute alone ($0.0040 a result) costs more than the median per-result price ($0.0030), so renting doesn't win there. It is no better at Bytewells' $0.15 per compute unit: $0.0030 a result. If your pipeline fires one-result runs, check the Actor's own numbers before you move. One heavy account's median Actor returned 4 results a run, which is in the range where renting still wins: $0.0119 per run at list price, against about $0.0068 of compute plus a share of a flat rental.

You are not the only one above the line

Plot each Actor's average renter against that Actor's own break-even:

At the median of 65 results a run, the average renter is above break-even for 34% of the 238 Actors with usage data, and those Actors hold 42% of the users. Count by runs and the answer barely depends on what you assume:

The share of Actors above the line goes from 11% to 41% depending on results per run. The share of runs stays between 89% and 94%, because the runs are on the Actors with heavy renters.

• The 238 Actors had 1,533 users and 5.1 million runs in 30 days. The median Actor had 2 users.
• The busiest Actor is 63% of the runs, with 4 users. Share of runs above break-even: 94% overall, 83% without that Actor, 71% without the five busiest.

Most renters are light and will pay less after October 1. A few users in a few Actors do most of the running. If that is you, you're the market rental was built for, and you can see what you'd save. We can't count how many of you there are from public data, only where the usage is.

What heavy accounts actually run

Three partner Apify accounts shared data for heavy-use analysis. We use results, run counts and start times only, and none is named.

• No run in account A's week of history was started by a person or Apify's scheduler: 100% came from the API or another Actor.
• 51% of its runs started in one hour of the day, which looks like batch jobs.
• The median Actor returned 4 results a run, and 93% of runs returned five or fewer.

Three accounts are not a market. They show what the top of the distribution looks like: automated, recurring, and 25 to about 2,000 times the break-even. Per-user runs on the Store follow the same shape, with a median of 46 runs a month and 1,816 at the 90th percentile.

Work out your own bill

It starts from the heavy account's pattern, small runs at about 200,000 results a month with the compute of 4 to 10-result runs. Switch to the median Actor, or put in your Actor's price, your runs, your results per run and your compute. It tells you which is cheaper and where your break-even is.

To find your Actor's pricing, open https://api.apify.com/v2/acts/username~actor-name and read the pricingInfos array: the current rental price and any scheduled pay-per-event price with its start date. No token needed. If nothing is scheduled, the Actor becomes free on October 1.

Your options, in order

1. Far above break-even: ask Apify for a custom discount

Apify can agree custom discounts with large accounts. Prices in this guide are list prices, so a negotiated rate can change the answer. If you are many times above break-even, start here.

2. Above it: ask the developer for a flat option

Many developers would rather keep a flat price for a heavy user than lose one. Ask for a volume tier or a flat monthly price. Developers can list on more than one marketplace, so they have somewhere else to offer it.

3. Above it: rent where rental survived

If the Actor runs your business and you're above break-even at 4 or more results a run, use a marketplace where the developer can still charge a flat monthly price. That is the Apify rental alternative we're building, next, and you can see what you'd save and we'll reach out to their developers.

4. Below it, or one-result runs: stay on pay per event

Under the break-even, or with 1 to 3 results a run, pay per event is the cheaper option. Check the numbers and stay put.

The Apify rental alternative we're building

Bytewells is an independent Actor marketplace that speaks the Apify API. Developers list Actors at a flat monthly price, keep 90%* of it, and get paid to keep the Actor working. Actors built with the Apify SDK run without code changes, and Apify integrations work after you change the base URL.

You pay the developer's price plus the compute your runs use, at $0.15 per compute unit from prepaid credit. There is no platform subscription and no per-result charge. Here is how the two compare:

Where it stands today. Most Actors are not listed yet, and we'd rather tell you that than promise a date. It works like this: you see what you'd save and pick your Actors, we reach out to the developers and help them add those Actors to our store, and you migrate with a one-line change in your code. Details of how renting works are in how it works for customers.

Methodology

Every figure above comes from public Apify data and heavy-use analysis of three partner Apify accounts (anonymous). This is how each was measured, so you can check it or argue with it.

Data

• Rental Actors and prices. Apify's public Store API, pulled September 26, 2026. 306 rental Actors had scheduled a move to pay per event. 291 of them have both a rental price and a per-result price. Those 291 are the sample for every break-even figure.
• Price per result. For each Actor, the event that charges per result: a dataset-item event, or a primary event titled as a result. Monthly passes, start fees, add-ons and hourly events are excluded, and anything above $1 a result is dropped. Plans map to Apify's discount tiers as Free, Starter = Bronze, Scale = Silver, Business = Gold. Apify's docs give the Starter example; Scale and Silver, Business and Gold is our reading.
• Discounts by tier. Each Actor lists a price per event for Free, Bronze, Silver and Gold. Its discount on a tier is 1 - tier price / Free price for the per-result event. We dropped Actors whose prices rise with the tier. We report the median, the mean, the share that discount at all, and the mean among those that do, for the 291 Actors moving, the 100 most used pay-per-event Actors, and the whole Store (9,201 Actors with a per-result price).
• Plans. Fees, prepaid usage and compute prices from apify.com/pricing (Sep 2026), monthly billing.
• Usage. Runs and users over the last 30 days, from Apify's public Actor statistics. Runs per user is runs divided by users, per Actor. 238 of the 291 Actors have usage data.
• Compute per result. A sample of 465 runs of 217 different Actors (311,076 results). We capped runs per Actor so no single Actor dominates, took results per run from dataset writes, dropped runs under 10 results because start-up cost inflates them, and counted each Actor once at its median. The result is 0.0018 compute units per result (middle half 0.0006 to 0.0034) and 65 results per run. Pooling all three accounts' Actors moved the median by about 5%.
• Compute by run size. The same runs plus all three accounts' Actors (316 in total), grouped by results per run without the 10-result cut, median across Actors in each group.
• Three partner accounts. The owners shared read access. We read billed dataset writes by month where history exists (up to 24 billing cycles), retained runs (7 to 31 days depending on the account), and for the newest 30 runs of each Actor the results and compute. The table above shows the two accounts with the longest billing history; the third contributed the run sample for compute (217 Actors, 465 runs). Everything is aggregated and no account, Actor or run is identified. We use no dollar figures from any of them.

Formulas

• Break-even = rental price ÷ (price per result − compute per result × price per compute unit). If the price does not cover compute, the Actor never breaks even. There are 13 such Actors among the 291.
• Bill on a plan = the larger of the plan fee and the usage at that plan's prices. The Free plan is allowed only while usage is $5 or less. Each side of a comparison takes the cheapest plan for that volume.
• Rental plus compute today = rental price + results × compute per result × the plan's price per compute unit, as usage on that plan.
• Pay per event after October 1 = results × the Free-tier price per result × (1 - the plan's discount), as usage on that plan. Compute is included in that price. The chart and tables use the average discounts of the Actors moving (7%, 11%, 16%); the top-100 case uses that group's median (10%, 22%, 36%).
• Bytewells estimate = rental price + results × compute per result × $0.15.
• Above the line = an Actor whose runs per user × assumed results per run is more than its break-even. Shares are counted by Actor, by users, and by runs.

Why medians

Prices, compute and usage are all right-skewed: a few extreme Actors and users pull an average far from what most people see. Mean compute per result is about 1.7 times the median, and mean price per 1,000 results is several times the median. We use medians for the defaults.

Limits

• Compute is the sample median, not each Actor's own. We have the rental and per-result price of each of the 291 Actors, but no compute measurements for them. Every break-even uses the same compute per result, so the spread in the chart comes from prices only. The table above shows how the median moves with the compute figure, and the calculator takes yours.
• List prices only. Apify can agree custom discounts with large accounts. None are in these figures, and we can't see them.
• Discounts vary by Actor. About half of the Actors moving give none, and the ones that do range widely, so one Actor can sit far from the median or the mean. The chart shows two assumptions, not a forecast.
• One Actor as the whole bill. A plan fee covers all your Actors and compute. We treat the median Actor as if it were the only thing you pay for. If you run several, the fee is shared and the steps come earlier for each.
• Plan simplifications. We ignore annual billing (10% off), assume unspent prepaid usage is lost, and treat the Free plan as unavailable above $5. The mapping of plans to discount tiers is our reading of Apify's docs.
• Scenarios use the median Actor. Your Actor's rental, price per result and compute differ; the calculator takes your numbers.
• Results per run is assumed when public data has none. We used the median from our sample for every Actor. The share of Actors above the line depends on it (chart above); the share of runs does not.
• Concentrated usage. One Actor with 4 users is 63% of all runs, and the median Actor has 2 users. Run shares are quoted with and without the busiest Actors. The average renter is a mean, and a few very heavy users pull it up.
• Two accounts. They show what heavy use looks like, not how common it is. Billed dataset writes exclude results from pay-per-event Actors, which the developer bills, so volume is a lower bound. One account's run list covers a week and the other's is too short to count runs.
• Compute per result varies by Actor and by run size. Cutting the sample other sensible ways moves the median between about 0.0012 and 0.0028, which shifts the median break-even by less than 10%.
• Proxy and storage costs are not included. Under rental you paid proxy on top; under pay per event the developer usually prices it in. That understates "rental plus compute today" a little.
• Prices are list prices on one day, and developers can change them. Usage is a 30-day snapshot.
• The Bytewells figure assumes the Actor is listed there at the same rental price. Developers set their own prices, and not every Actor is listed yet.

FAQ

Sources: public Apify Store data pulled September 26, 2026, a sample of Actor runs and three partner accounts' heavy-use data (see the methodology), and apify.com/pricing (Sep 2026). Figures may change. Apify and Apify Store are trademarks of Apify Technologies s.r.o. Bytewells is not affiliated with Apify. Last updated September 30, 2026.