Supabase Pricing in 2026: Every Plan, Real Costs and Hidden Fees

Supabase does not charge per seat: Pro is $25/month for the organization, Team is $599/month, and Enterprise publishes no list price. Dashboard members are unlimited, so ten engineers do not multiply the subscription. The bill that moves is compute: one instance per project, and the included credit covers a single Micro instance.

Prices below come from Supabase's pricing page and platform docs, checked September 24, 2026. Each organization holds one plan, and every project is its own Postgres server, with auth, storage, realtime, and edge functions on that same instance, while compute is metered by the hour on top of the plan fee. The product choice against Firestore is in Supabase vs Firebase, and this page is the invoice.

Toolradar data: the managed databases ranking evaluated 23 tools, updated September 2026. The writeup on this site is the Supabase review, and the directory record is Supabase on Toolradar. A catalog-side pass at the same invoice is Supabase pricing on Toolradar.

Plan	Price	Best for	What changes the bill
Free	$0	A prototype that can sleep	Pause after a week idle, 2 active projects
Pro	$25/mo	One production project on Micro	Quotas, then overage if the spend cap is off
Team	$599/mo	SOC 2, longer logs, project roles	Same usage quotas as Pro, no spend cap
Enterprise	Custom quote	Uptime SLA, private Slack, PrivateLink	Sales, not checkout

Free is a real Postgres box that goes to sleep

Free is a working project, not a demo login. You get unlimited API requests, 50,000 monthly active users, 500 MB of database, 5 GB of egress, 5 GB of cached egress, and 1 GB of file storage. The instance is shared CPU with 500 MB of RAM, smaller than Micro on a paid org, with no card, no daily backups, no email support, and a 50 MB file upload cap.

The leash is inactivity: Supabase pauses a free project after one week without use. You can pause it yourself, and paused projects do not count toward the cap, while two projects may be active. That cap is counted across every Owner and Admin, including people who already used a free-project allowance elsewhere, so a second owner can block a launch.

Exceeding a quota long enough brings fair-use restrictions: the project can pause, the database can go read-only, or API calls can return 402. Paid and free projects cannot share an organization, so a database that must stay awake and a scratch project that should stay free are two orgs until you transfer one. Compute on Free is the Nano size at no hourly charge, and Supabase says those free resources can change.

Pro is an organization fee, then a project meter

Pro is the self-serve plan, charged up front once a month, and the fee is not a seat and it is not a ceiling. Quotas on the pricing page are 100,000 monthly active users, 8 GB of disk per project, 250 GB of egress, 250 GB of cached egress, and 100 GB of file storage. Email support, 7-day backups, 7-day logs, and unbranded auth email start here, API requests stay unlimited, and uploads can go to 500 GB.

Past those quotas the unit prices stay small until the app is not:

Meter	Included on Pro	After the quota
Monthly active users	100,000	$0.00325 per MAU
Disk	8 GB per project	$0.125 per GB
Egress	250 GB	$0.09 per GB
Cached egress	250 GB	$0.03 per GB
File storage	100 GB	$0.0213 per GB
Edge Function invocations	2 million	$2 per million
Realtime messages	5 million	$2.50 per million
Realtime peak connections	500	$10 per 1,000

Cached egress is a storage CDN hit and the cheaper bandwidth line. Uncached egress is one pool for database, auth, functions, realtime, the pooler, and log drains. The spend cap, on by default, blocks overage on those rows instead of billing them, and turning it off is how the unit prices reach the card.

One Micro project stays inside the plan fee, because that fee includes a monthly compute credit equal to one Micro instance. Supabase's own example of two Micro projects lands at $35/month after that credit is applied once. The billing FAQ puts three Micro projects at $45/month. There is no official annual discount, a credit top-up can cover future months, and it will not pay an invoice that is already overdue.

Compute is billed per project, by the hour

Every project is a virtual machine running Postgres, and you pay for the size you selected whether or not the database is busy. Paused projects drop off the meter, and a partial hour is billed as a full hour. Hourly rates are in the compute doc, and a 730-hour cycle is the month used in Supabase's invoice examples. The monthly list below is the one on that pricing page.

Size	Monthly list	Shape	Direct / pooler connections
Micro	$10	Shared, 1 GB RAM	60 / 200
Small	$15	Shared, 2 GB RAM	90 / 400
Medium	$60	Shared, 4 GB RAM	120 / 600
Large	$110	Dedicated, 2 vCPU, 8 GB	160 / 800
XL	$210	Dedicated, 4 vCPU, 16 GB	240 / 1,000
2XL	$410	Dedicated, 8 vCPU, 32 GB	380 / 1,500

Sizes continue through 16XL, and anything larger is a sales conversation. Micro through Medium are shared, and Large is the first dedicated size, which is the upgrade when connection caps or noisy neighbors are the problem.

The credit applies to compute only, and it does not roll over, so it will not pay for a read replica, a branch, disk, egress, or IPv4. One project left on Small for a full cycle totals $30 in Supabase's published example, after the credit. Three Small projects in that doc total $60 after the credit, because two of the three instances are uncovered. A mid-cycle resize bills each size for the hours it ran, settled at the next reset, with no charge on the day you click.

Nano is the free-plan size, and on a paid org a leftover Nano bills at the Micro rate. You cannot launch a new Nano once the org is paid, and size changes are not automatic, so move a leftover Nano to Micro when you can take the restart.

Branching, the alternative to a second project, is $0.01344 per branch per hour, the same hourly rate as Micro, and credits do not touch it. At that rate, 730 hours comes to $9.81, on top of a primary whose credit is already spent, and deleting the branch stops the meter.

Team buys compliance, not a larger quota

Team is the same organization shape at a much higher subscription. The pricing-page comparison gives it the same disk, egress, storage, MAU, function, and realtime quotas as Pro, at the same overage rates, so the quotas do not grow. What changes is SOC 2 and ISO 27001, project-scoped and read-only roles, priority email with an SLA, 14-day backups, 28-day logs, and platform audit logs. Dashboard SSO says to contact them, HIPAA is a paid add-on with no printed fee, PrivateLink is included, and an uptime SLA is not on this plan.

The spend cap does not exist on Team, and that switch is Pro-only, so quota overages bill. Team still gets the same one-project compute credit, and the add-ons below bill either way. Stay on Pro until a security review asks for the report, the role split, or the longer log window. Ten developers still do not multiply the subscription, and ten projects will.

Enterprise is a quote because the contract is the product

The page lists a designated support manager, an uptime SLA, PrivateLink, round-the-clock support, a private Slack channel, and help with security questionnaires. Disk, egress, MAU, storage, functions, and realtime are marked custom. Point-in-time recovery can pass 28 days, and log retention is 90 days.

Nothing on the public page sits next to the Team subscription except a sales form, so buy this tier for the SLA or a quota Team will not raise. A company that only needs SOC 2 should price Team first. Invoices stay in USD, card is the default, and a larger upfront payment is a support ticket rather than a checkout option.

The spend cap ignores the lines people opt into

The spend cap is on by default on Pro, which is why a buggy loop does not become an egress invoice. While it is on, crossing a covered quota stops that usage instead of charging it, and past a grace period the fair-use restrictions follow. The billing FAQ says a later overrun can skip a fresh grace period if you already blew the quota once. Covered meters include disk size, egress, edge invocations, logs, MAU, realtime, storage size, and image transformations.

Compute is absent from that list, and so are branching, read-replica compute, custom domains, extra disk IOPS, extra throughput, IPv4, log drains, phone MFA, and point-in-time recovery. Those count as choices you turned on, so a cap that still looks enabled will not absorb a Large instance, a replica, or a branch that ran all month. The rules for that split are in the cost control doc.

Add-on	Price	Spend cap
Point-in-time recovery	$100 per month per 7 days of retention	No
Custom domain	$10 per domain per month per project	No
Phone MFA	$75 per month for the first project, then $10 per month per extra project	No
Log drains	$60 per drain per month, plus $0.20 per million events and $0.09 per GB	No
IPv4	$0.0055 per hour, listed as $4 per month	No
Image transformations	100 origin images, then $5 per 1,000	Yes, the overage

General-purpose disk includes 8 GB, 3,000 IOPS, and 125 MB/s, extra IOPS are $0.024 each, and extra throughput is $0.095 per MB/s. High-performance disk is a different product at $0.195 per GB, with IOPS at $0.119, and it is not the included 8 GB. A read replica mirrors the primary's compute, disk, IOPS, throughput, and IPv4, and replica compute sits outside the credit and the cap.

The IPv4 doc shows a full month of the add-on as its own line, and every replica gets an address when the primary has one. SAML for your own users, separate from dashboard SSO, includes 50 monthly active users and then $0.015 per MAU, while social login does not use that rate.

What ten developers pay next to Firebase, Neon, and PlanetScale

None of these four bills a seat, so the useful question is which meter runs when the database stays on. Prices were checked on each vendor's pages on September 24, 2026. Product fit, apart from this invoice, is the backend-as-a-service list and the Postgres hosting list.

Buyer	Supabase	Firebase	Neon	PlanetScale
Prototype, no card	Free, pauses after a week	Spark, no-cost quotas	Free, 100 CU-hours per project	No free plan
One small database, always on	Pro fee, Micro covered by the credit	Blaze, no plan fee, then usage	0.25 CU for 730 hours is $19.35 before storage	Single-node PS-5 in us-east-1 is $5/mo
Auth, files, and functions	Included on the org plan	Separate product meters	On the same Neon plans	Database only
Highly available production	A larger size, or a replica you pay for	Cloud SQL after the trial, from $9.37/mo	Scale, $0.222 per CU-hour	HA PS-5 in us-east-1 is $15/mo
Ten people, one project	Still one org fee	Still no seat line	Unlimited members, usage only	Still the cluster price

Firebase Spark needs no card, and with Identity Platform, monthly active users are no-cost through 50,000. Firestore's standard edition includes 1 GiB stored and 10 GiB of egress a month, then Google Cloud rates the Firebase page does not reprint as one number. Blaze keeps those quotas and adds a billing account. Cloud Functions include 2 million invocations, then $0.40 per million, and phone auth is per SMS, on a rate card separate from the plan table.

SQL Connect, the Postgres path, is a three-month trial and then starts as low as $9.37/month, and that figure is Cloud SQL, not a discount on Firestore. The Firebase review covers the product, and the directory page is Firebase on Toolradar.

Neon Launch and Scale have no monthly minimum, and Launch compute is $0.106 per CU-hour, which suits a database that is allowed to sleep. Neon describes a CU as about 4 GB of RAM, so 0.25 CU is about 1 GB, close to Micro. Left on for 730 hours, that is 182.5 CU-hours, which comes to $19.35 before storage. Paid storage is $0.35 per GB-month, public egress includes 500 GB per project, then $0.10 per GB, and an extra branch past the allowance is $1.50 per branch-month.

Suspended compute is not billed, and invoices under $0.50 are not collected. Free is permanent: 100 CU-hours, 0.5 GB, and 5 GB of egress per project, and hitting compute or egress suspends the database until the next period instead of charging the card. The Neon review is the product page, and the directory record is Neon on Toolradar.

PlanetScale's plans doc says the old free plan is gone, and single-node Postgres starts at $5/mo. The Postgres pricing doc lists that PS-5, in us-east-1, at 1/16 vCPU and 512 MB of RAM, and the highly available PS-5 there is three times the single-node sticker. Network-attached storage includes 10 GB, then $0.125 per GB-month in us-east-1, the same disk rate Supabase charges past 8 GB.

A non-HA PS-5 includes 10 GB of public egress, a production branch includes 100 GB, and further egress in us-east-1 is $0.06 per GB. Each Postgres development branch is billed as its own single-node cluster, and SSO on the Base plan is $199/month. Auth, files, and functions stay on another invoice, and the directory record is PlanetScale on Toolradar.

Project count, instance size, and whether the database may sleep are what move the invoice on all four.

Fees that never appear on the plan card

The subscription is the line you approved, and compute is the line that arrives anyway. The spend cap skips compute, so a resize, a second project, or a replica changes the invoice while the cap still looks on. The plan fee is charged at the start of the cycle, usage including compute is charged at the end, and a quiet week does not shrink an instance that stayed running.

Disk size sits inside the cap, so a database past 8 GB stalls instead of silently adding gigabytes at the overage rate. Turning the cap off to finish a migration also unblocks egress, MAU, and storage, and those meters move as a set.

Social login sits inside the MAU quota, SAML past 50 users is the $0.015 rate, and phone MFA is the $75 project add-on, outside the cap. Image transformations include 100 origin images, and the per-thousand fee in the table is the rest. Taxes are separate from the plan, and the billing FAQ says sales tax, VAT, and GST apply where Supabase is registered, based on the billing address, with collection scheduled to finish by June 30, 2026. A tax ID is what flips an eligible business invoice to reverse charge.

Charges are in USD and can appear to come from Singapore, the payment entity, so a card that blocks foreign merchants fails renewal. An overdue invoice pauses projects and drops the org to Free until it is paid. Supabase does not refund unused credits, canceling mints credits for unused subscription time, usage you already consumed can still be invoiced, and top-ups cannot pay that outstanding bill.

Docker and the CLI skip the subscription, and Studio ships in that Docker setup, while the machine, disk, and bandwidth are then yours, which is a VPS hosting decision rather than a plan decision. A Next.js app on Vercel is a separate invoice and does not shrink this compute line.

Devshot data: the audience file updated September 20, 2026 counts 28,000 developer subscribers, and the open rate is 30%. They read a backend bill the way they read a cloud bill, so a second project plus a spend cap left on is the month-two surprise. Plan changes for this list land in the Devshot brief.

Which plan is worth paying for

For one person, start on Free, accept the pause, and move to Pro when the project must stay awake or the database passes 500 MB. Keep Micro until connections or CPU say otherwise, and add a second project only when you can name the invoice line. A branch left up all month costs about a Micro and does not receive the credit, so delete it when the pull request merges.

For a team, Pro is the right subscription until a review names SOC 2, ISO 27001, or a read-only role, and Team does not raise the quotas. Turn the spend cap off only for the meter you meant to exceed, and check the upcoming invoice that week, because the cap is not a per-meter budget. If the database should sleep, price Neon before a larger Supabase size, and if auth lives elsewhere, price PlanetScale before a replica.

If the rest of the stack is still open, the serverless platforms roundup is the wider set, and the Firebase comparison above is the direct one.

How we checked these prices

Louis Corneloup, founder of Dupple and Toolradar, wrote this breakdown. We read Supabase's pricing page, the billing FAQ, the compute doc, the cost control doc, and the IPv4 doc on September 24, 2026.

Firebase figures come from firebase.google.com/pricing, Neon figures come from neon.com/pricing, and PlanetScale figures come from the plans doc and the Postgres pricing doc, us-east-1 column. The 23-tool count is Toolradar's managed databases ranking that month, and how this site publishes is on the Devshot about page.

FAQ

How much does Supabase cost in 2026?

There is no per-seat price: Free is $0, with two active projects and a pause after a week idle, Pro is $25/month per organization, and Team is $599/month per organization. Enterprise publishes no list price, compute is extra per project, a credit covers one Micro instance, and quota overages bill on Team or when the Pro spend cap is off, and invoices are in USD.

Does Supabase charge per seat?

No. Team members are unlimited on every plan, so ten engineers on one project pay the same subscription as one engineer. What scales is the number of active projects, the compute size of each, replicas, branches, and any quota you cross with the cap off.

Is Supabase free?

Yes, with limits. Free includes 50,000 monthly active users, 500 MB of database, 5 GB of egress, and 1 GB of file storage, with no card. Projects pause after one week idle, only two may be active, and Owners and Admins share that allowance. Daily backups and email support are paid, and Docker and the CLI avoid the subscription while the server under them does not.

What is included in Supabase Pro?

The org subscription includes a compute credit equal to one Micro project, 100,000 MAU, 8 GB of disk per project, 250 GB of egress, 250 GB of cached egress, 100 GB of file storage, 2 million edge invocations, email support, and 7-day backups. Unit rates apply when the spend cap is off. A second project, a larger instance, point-in-time recovery, a custom domain, and IPv4 sit outside that subscription.

Does the Supabase spend cap stop a large bill?

It stops part of the bill: on Pro it is on by default and it blocks extra disk, egress, functions, MAU, realtime, and storage size. It does not cap compute, branches, read replicas, IPv4, point-in-time recovery, custom domains, extra IOPS, log drains, or phone MFA, and Team has no switch. A replica or a size change can still hit the invoice while the cap looks enabled.

How much is Supabase for two projects?

Supabase's pricing page puts a Pro org with two Micro projects at $35/month after the credit, and three Micro projects are $45/month in the billing FAQ. A branch instead of a second project is $0.01344 per hour, and the credit does not apply, so a branch that stays up all month is close to another Micro. A free project and a paid project cannot live in the same org.

Is Supabase cheaper than Firebase?

Both have a no-card plan, and Firebase will not pause you for a quiet week. For one always-on database around 1 GB of RAM, Pro is the org fee with Micro covered, while Blaze has no plan fee and then meters Firestore, functions, and bandwidth on their own. Neon at 0.25 CU for a full month is $19.35 before storage, which can undercut Pro when auth and files are not on this invoice. PlanetScale's single-node entry is the us-east-1 sticker in the table, with no auth product.

Can I pay Supabase yearly?

There is no published annual plan, and the billing FAQ says annual subscriptions are not officially supported. The workaround they describe is a credit top-up so the card is not charged every month. Those credits pay future invoices, they do not pay one that is already outstanding, and unused credits are not refunded.

Cite this: Devshot, "Supabase Pricing in 2026: Every Plan, Real Costs and Hidden Fees", September 2026.

The daily Devshot brief is where plan changes for this stack show up before they show up on an invoice. Subscribe at devshot.co.

Some offers on this page may be paid placements or contain affiliate links.