Bubble Pricing: Starter, Growth and Team Plans and Workload Units

Bubble lets you build for free, but a live web app needs a paid plan: Starter is $32, Growth $119 and Team $349 per month on annual billing; monthly billing is higher. Workload units measure the server work your app performs. Starter includes 175,000 monthly units, Growth 250,000 and Team 500,000, with overages or add-on tiers available when use grows.

There are two separate meters. Bubble AI credits pay for using its AI Agent to build or edit parts of an app; workload units pay for the app's running server activity. A business can use few AI credits while its live app consumes many workload units, or use the Agent heavily while the app has almost no users. Budget them separately.

Follow me on Instagram@sagnikteaches

Choose the platform type before the plan name

Bubble currently sells Web, Mobile and Web & Mobile project types. A browser-based job tracker needs Web. A native phone app distributed through app stores needs Mobile. A project with both surfaces and one shared backend needs Web & Mobile. The same Starter label has a different price for each type. Bubble's current FAQ lists, on annual billing, Web Starter at $32, Mobile Starter at $42 and Web & Mobile Starter at $59 per month. A reader pricing a simple browser app should not accidentally use the combined-app figure.

Connect on LinkedInSagnik Bhattacharya

The Free plan is for development and includes 50,000 workload units a month, but Bubble says an app must upgrade to go live, and Free apps can neither buy workload tiers nor enable overages. Starter Web includes a live website and custom domain, with 175,000 workload units. Growth Web increases included workload to 250,000 and adds two app editors, premium version control and more controls. Team Web includes 500,000 units, five app editors and additional collaboration features. The tier should follow the people building the app and the measured running load, not a guess that the word Growth sounds safer.

Subscribe on YouTube@sagnikteaches

Third-party round-ups and old screenshots do not always match these figures, so do not average a remembered price with a current one. Bubble's live pricing page and the plan table in its manual agree on the web prices above. They also show how much a native app adds: on annual billing, Growth is $169 for Mobile or $209 for Web & Mobile, and Team $449 or $549; billed monthly, Starter is $49 or $69. For a new purchase, use the live pricing page and checkout. Existing customers should check their own renewal notice for any grandfathered or transitional price.

Workload units are not a price per user

A workload unit, or WU, is Bubble's measure of server resources used to run an app. Database searches, workflow steps and web interactions can all contribute. The exact amount depends on what the app does. Ten users running a heavy background process may consume more than 100 users viewing a simple page. Therefore “we expect 50 staff” is not enough to estimate WU; the owner needs to test the actual screens and automations.

Bubble's docs say more workflow runs tend to use more workload, but the number and nature of steps matter as well. A button that updates one job's status is a different load from a button that searches every job, creates multiple records and sends notifications. A scheduled nightly task may continue consuming WU when no one is logged in. This is particularly relevant to a small firm that imagines a quiet internal app will be cheap simply because only six people have accounts.

There is no reliable universal conversion such as “one form submission equals five WU”. Use the Logs and workload charts in the Bubble editor to measure your own app. Run repeatable test actions and record the change in usage. If the chart shows a large spike after one screen opens, inspect the searches and workflows on that screen. If it spikes at 02:00, check scheduled jobs. The measurement becomes your estimate; an example number in another person's app does not.

Bubble's pricing FAQ says live and development workload both count towards total usage, and that paid apps include 100,000 monthly WU for the development environment, where testing and database clean-up happen. Do not treat testing as invisible in a cost forecast. Bulk operations and data imports in the editor can use workload. A business that imports thousands of old customer records during launch week may see a temporary spike that should be separated from normal monthly activity. Keep the migration month and the steady month as different scenarios.

A month of use for a small job-tracking app

For the main illustrative example, a six-person HVAC contractor builds a web app for job enquiries, visits and completion photos. On a measured pilot, suppose normal page and search activity uses 70,000 WU a month, status updates use 40,000, scheduled reminders 20,000 and report generation 15,000. The planned total is 145,000 WU. Add 20% headroom for uneven demand and retries: 174,000 WU. That sits just below Starter's 175,000 allowance, but with almost no spare room.

For a steady 145,000-WU month, Starter Web at $32 annual-billing equivalent may be sufficient if the owner accepts the plan's editor and feature limits. For a forecast 174,000, the team should monitor usage closely or simplify a heavy workflow. Suppose a busy month reaches 205,000 WU. That is 30,000 above Starter. Bubble's docs list $0.30 per 1,000 WU without a workload tier, so the illustrative overage is 30 × $0.30 = $9. The total would be $41 for that month at the stated Starter annual-billing equivalent, before other add-ons.

Growth Web is $119 monthly equivalent on annual billing, an $87 difference from Starter's $32. Upgrading solely to cover a 30,000-WU overage would be poor value in this example. Growth may still be worth it for two editors, version control, security features, logs or a sustained workload above Starter. Conversely, if the app regularly reaches 500,000 WU, the overage maths and operational features deserve a new comparison. A plan pick should combine measured usage and required features.

Write the worksheet in four lines: base page activity 70,000; job updates 40,000; reminders 20,000; reports 15,000. Then add pilot total, peak factor and planned WU. Next month, replace these illustrative values with the app's measured values. Label uncertain items, such as a future customer portal, separately. This prevents an owner from buying Team because a portal might exist someday. See mapping a process before automating it if the job steps themselves are still unclear.

Six tests that change the WU forecast

First, a bicycle repair shop records 20 repairs a day and asks staff to open a repair dashboard. If the page searches all historic records every time, usage can rise with the archive even though daily repair volume stays flat. Test the dashboard with 100, 1,000 and 10,000 records. If WU per page grows sharply, narrow the search or paginate results before buying more allowance.

Second, a bakery's order form creates one order, then triggers five separate notifications. The owner may only count the one form submission, but Bubble counts the server work of the workflow. An illustrative measurement might show 80 WU for the full chain and 25 after consolidating unnecessary steps. The exact figures are test values, not Bubble's tariff. Keep the same output quality when simplifying: the customer and kitchen still need the right information.

Third, a café uses a scheduled workflow every five minutes to check whether any table reservations need reminders. Most runs find nothing. Background polling can still consume resources. A better design may trigger from the reservation event or schedule one reminder per booking. Measure both approaches in the app's workload chart; do not assume fewer visible users means fewer WU.

Fourth, the HVAC contractor imports 4,000 old jobs during setup. The launch month shows 230,000 WU, but the following normal month shows 145,000. If the owner upgrades based on the one-off spike, the higher plan may sit unused. Record migration work separately and compare a workload tier or controlled overage for that exceptional month. Ensure the import also preserves record accuracy; a cheaper import that creates duplicates can cost more in staff correction.

Fifth, a customer portal is added to the bicycle shop's app. Customers can now check repair status from their phones. The number of visits rises from 300 to 3,000 a month, while each visit is light. The pilot must measure WU per visit and multiply by a plausible peak, then confirm the plan supports the new access model. Count operations and data, not just the registration count.

Sixth, a staff member uses Bubble's AI Agent to rewrite a screen and runs through most of the included AI credits. The app's normal user activity continues to consume WU separately. Bubble's plans include 500 AI credits a month on Free, 1,000 on Starter, 2,000 on Growth and 4,000 on Team; included credits refresh on the first of the month without rolling over, and bought credits expire six months after purchase. When credits run out, the pause applies to AI-powered features such as the Agent and page generation, not to the app's workload, which is a different event from the live app hitting a WU cap. A budget sheet with one column labelled “AI” obscures the distinction.

Decide whether overage or a tier is cheaper

Bubble offers pay-as-you-go workload overages on paid plans or a purchased workload tier. Without a tier, the manual lists $0.30 per 1,000 WU above the included amount. Bubble's workload tiers page lists Tier 1 as 200,000 extra WU for $26 a month on annual billing, with the overage rate above the combined allowance dropping to $0.15 per 1,000. The tier adds capacity; it is not a replacement for the base app plan. Higher tiers follow the same pattern on the workload tiers page: Tier 2 adds 750,000 WU for $89 a month on annual billing at a $0.14 overage rate, and Tier 3 adds 2.5 million WU for $269 at $0.12. Check the current price and billing cadence before ordering one.

Suppose Starter's 175,000 units are used and a web app needs another 50,000 WU each month. At the no-tier overage rate, that is 50 × $0.30 = $15. A $26 Tier 1 would cost more for that level of excess, even before considering its extra capacity. At an extra 200,000 WU, the simple no-tier overage would be $60, so a $26 tier may be better if its terms fit. The crossing point sits near 87,000 extra WU a month, where no-tier overage reaches about $26; above that, Tier 1 is cheaper while its terms fit. Bubble bills overages after the fact: on annual plans at the start of each month for the previous calendar month, on monthly plans on the last day of the month, so a heavy month is charged after the usage has happened, not as it happens.

There is a separate reliability decision. Bubble allows an administrator to disable overages. Its docs say that if an app then reaches its cap, the app goes offline until overages are enabled, a tier is bought or the next billing period starts. A customer-facing booking app should not treat that as a harmless spending limit. Decide who receives the 75% and 100% usage notices, who may approve extra workload and what manual booking process operates if the app is unavailable.

For an internal job tracker, an owner might choose to disable overages during a pilot and accept a pause, provided staff can fall back to a shared spreadsheet. For a customer appointment form, a controlled overage budget and a daily usage check may be safer. These are business choices, not universal settings. Put the maximum acceptable interruption and the maximum authorised monthly overage in the same decision record.

Include costs that WU does not cover

Bubble's pricing page lists extra file storage at $3 per 100 GB a month and plugin subscriptions as possible add-ons. External services used by the app can have their own bills. An HVAC app that sends messages through another provider or stores large job photos needs those costs in its budget. A Bubble plan may include hosting and a database, but it cannot price every external service the owner connects.

Time is another cost. Estimate staff hours to define fields, test permissions, correct imported records and train colleagues. For an illustrative launch, 12 hours of owner testing and 8 hours of staff handover at an internal planning rate of $30 an hour would represent $600 of time. That is not a Bubble charge. A $32 subscription is not the full cost of a usable process. The app-connection cost comparison helps where the project depends on integration with existing tools.

Before upgrading, check whether a simpler process would solve the job. A six-person team might not need a custom customer portal if a shared form and calendar already handle bookings. A useful Bubble app has clear records, permissions and recurring decisions. If those are vague, prototype on Free until staff can use a representative workflow. Then choose Starter for the first live web release unless pilot data or collaboration needs justify Growth. Team is a feature and scale decision supported by measured demand, not an insurance policy bought on day one.

Review the first four weeks of live usage. Record WU per successful job, the three largest consumption sources, peak day, failed workflows, active users and AI Agent credits separately. If a workflow consumes more than expected, inspect its steps before paying for more capacity. If the app remains well below Starter and only one person edits it, there is no reason to move tiers for appearance. If staff need simultaneous editing and the app's operations support Growth, make that reason explicit in the renewal record.

Use a weekly usage log to catch a bad workflow

Set up a log with seven columns: week ending, live WU, development WU, completed business actions, WU per completed action, largest workload source and action taken. For the HVAC contractor, an illustrative first week might show 28,000 live WU and 95 completed jobs, or about 295 WU per completed job. That ratio is only a rough indicator because searches and reminders also occur between jobs. Its value is comparison: if week two shows 54,000 WU for 100 jobs, someone should inspect the workload chart before assuming demand doubled.

Suppose the chart shows a scheduled reminder workflow ran six times for every job because it was triggered on each field update. The builder changes the trigger so it runs once when a booking is confirmed. The next comparable week uses 34,000 WU for 100 jobs, and all 100 reminders are still sent correctly. That is a measurable improvement. Keep a record of the before and after counts so a later editor understands why the trigger is narrow. The purpose of the log is to change an inefficient process, not merely predict the invoice.

Another week may show normal WU but only 70 completed jobs, because staff abandoned a slow form. A cost dashboard alone would call that a success. Pair WU with completion rate, error count and staff feedback. If a new approval step reduces accidental bookings, extra WU could be justified. If it adds delay without catching errors, remove it. The quality-check tutorial offers the same principle for AI-assisted work: cost and speed matter only beside acceptable results.

Check the forecast on day ten and day twenty of the billing month. A simple projection is WU so far divided by elapsed days, multiplied by the number of days in the period. If 80,000 WU have been used after ten days of a 30-day period, straight-line use points to 240,000 WU. Before approving extra capacity, ask whether a one-off data import caused 30,000 of those units. Subtract it from a steady-use forecast, then add any known peak events. This gives the owner two numbers: likely normal demand and a plausible high case.

Finally, test the stop policy outside a busy period. Have the account owner identify where to see the 75% notice, how to change overage settings in the App Plan area, who holds payment authority and how staff will work if the app is offline. Put a one-page fallback beside the normal process: record new jobs in the approved backup sheet, assign a person to reconcile them when the app returns, and check for duplicates. A cap that protects a small subscription bill is useful only if the business can continue its essential work while it is enforced.

Bubble billing questions after a test app

Are Bubble AI credits the same as workload units?

No. AI credits cover building or editing with Bubble's AI Agent and certain AI creation activity. Workload units measure the server work of a running Bubble app, such as database operations and workflows. They have separate allowances and different consequences when exhausted. Track both if you use the Agent to build an app that people then use.

What happens if Bubble workload units run out?

On a paid plan, enabled overages can add pay-as-you-go charges after the included allowance and any purchased workload tier are used. If you disable overages and the app reaches its cap, Bubble says it will take the app offline until you enable overages, buy more workload or the next billing period begins. Choose a policy before launch.

Why do some sites show a different Bubble Starter price?

Third-party round-ups and old screenshots can lag behind price changes. Bubble's live pricing page and its manual's plan table both show Starter Web at $32 a month when billed annually, or $39 billed monthly, with Mobile and Web and Mobile plans costing more. Use those and the checkout for a new purchase, and check your renewal notice if you already subscribe.

Further reads

• Power Apps for Small Businesses: When It Replaces Custom Software — Compare another route to a small internal app.
• How to Tell If a Process Is Ready to Automate With AI — Check whether the workflow is stable enough to build.
• AI vs Rule-Based Automation: Which Does Your Task Need? — Separate deterministic app logic from AI tasks.
• How to Calculate AI ROI for Your Business (Worked Example) — Compare the app's running cost with its value.
• How to Review an AI Tool After 90 Days: Keep, Fix or Cancel — Revisit the subscription after actual usage.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Bubble pricing page and its AI-credit and workload FAQ; Bubble workload tiers page; Bubble Manual, Pricing plans; Bubble Manual, Pricing FAQ; Bubble Manual, Plans and billing; Bubble Support, Workload explained.