Research30 min read
How Much Does It Cost to Run an App? Monthly Costs at 1,000, 10,000 and 100,000 Users (2026)
In short
A line-item cost model for running a business app at 1,000, 10,000 and 100,000 monthly active users, priced only from vendors' own pricing pages checked November 11, 2026: Firebase, Supabase, AWS, Vercel, Amazon SES, Twilio, Sentry, OpenAI and Anthropic. Three reference apps (a booking portal, a photo marketplace and a portal with an AI assistant) with every assumption stated, the bill split by line, Firebase vs Supabase vs AWS for the same app, cost per active user, and the five lines that make a bill jump.
Reviewed by Alpit Patel, Founder and CEO · Updated November 11, 2026
Part of the guide: Software development
On this page 16 sections
- How much does it cost to run an app?
- Key findings
- Method: what we priced and how
- What every line item costs: the vendor prices we used
- Profile A: what a booking or customer portal costs to run
- Profile B: what a marketplace with photos costs to run
- Profile C: what an AI assistant adds to the bill
- Where the money goes at each tier
- Firebase vs Supabase vs AWS: the same app on three backends
- Cost per active user per month
- What makes an app’s running cost jump?
- How to keep the bill low
- How to use this data for your own app
- How we can help
- Changelog for this page
- Sources
Key takeaways
- In our illustrative model, a typical business app costs about $17 to $77 a month to run at 1,000 monthly active users, $181 to $294 at 10,000 and $1,607 to $2,680 at 100,000, using vendors' own prices checked Nov 11, 2026.
- At 100,000 users that works out to about 1.6 to 2.7 cents per active user per month across three app types and three backends, before any mid-size AI model.
- Firebase was cheapest at 1,000 and 10,000 users because its free quotas cover small apps. Supabase was cheapest at 100,000 users in all three profiles, mainly because Pro includes 100,000 auth users and cheap cached egress.
- Text messages beat the backend. One SMS per user a month cost $1,280 at 100,000 users (Twilio's $0.0083 base plus a carrier fee), more than any backend we priced for the same app.
- AI model choice is a 20x decision: 1,000 assistant conversations cost $1.50 on a small model and $30 on a mid model at the same token count. At 100,000 users that is $450 or $9,000 a month.
- Push notifications are free to send: Firebase lists Cloud Messaging as no-cost on both plans. Outside the choice of AI model, moving reminders from SMS to push is the largest saving in our model.
How much does it cost to run an app?
A typical business app costs about $17 to $77 a month to run at 1,000 monthly active users, $181 to $294 at 10,000 and $1,607 to $2,680 at 100,000. Those are illustrative figures from our own model, built only from the vendors’ published prices checked on November 11, 2026, for three reference apps on three backends. A mid-size AI model, heavy image traffic or text messages can move the 100,000-user bill well past that range.
Illustrative, based on our assumptions. Prices from each vendor’s own pricing page, checked Nov 11, 2026. Sources are listed at the end.
This paper shows every line item and the arithmetic behind it, so you can swap in your own usage and get your own number.
Key findings
- Small apps can run for about $20 a month. Our booking portal at 1,000 users paid $0 for its Firebase backend; the bill was the $8.25 a month share of Apple’s $99 developer fee, $13 of text messages and 60 cents of email.
- Every backend lands near 2 cents per user at scale. At 100,000 users the full bill was $0.016 to $0.027 per active user per month across all nine profile and backend combinations, before a mid-size AI model.
- Firebase wins small, Supabase wins at 100,000 users. Firebase was cheapest at 1,000 and 10,000 users in every profile. Supabase was cheapest at 100,000 users in every profile, with a backend bill of $233 against Firebase’s $479 and AWS’s $848 for the portal app.
- Texts cost more than servers. One SMS per user a month was $1,280 at 100,000 users, 69% of the Firebase portal bill. Phone sign-in codes through Twilio Verify cost $0.0583 each against $0.01 for Firebase’s own phone auth SMS.
- Image egress is the marketplace’s biggest backend line. 5.9 TB a month of photo downloads cost $712 on Firebase Cloud Storage at $0.12 per GB, $168 as Supabase cached egress at $0.03 per GB, and fit inside CloudFront’s $200 Business flat-rate plan on AWS.
- AI model choice is a 20x decision. At the same token counts, 1,000 conversations cost $1.50 on OpenAI gpt-6-luna or Claude Haiku 5.5 and $30 on gpt-6.1-sol or Claude Sonnet 5.5. At 100,000 users that is $450 or $9,000 a month.
- Auth pricing differs more than database pricing. At 100,000 users, sign-in cost $0 on Supabase Pro (100,000 included), $275 on Firebase and $495 on Amazon Cognito’s Lite tier, or $1,350 on Cognito’s default Essentials tier.
Method: what we priced and how
We defined three reference apps, wrote down what one monthly active user (MAU) does in each per month, and priced every line from the vendor’s own pricing page or official price list on November 11, 2026. Every total on this page is illustrative, based on our assumptions. Your app’s usage will differ; we show the assumptions so you can replace them.
The three profiles:
- A. Booking or customer portal. Customers book, pay deposits elsewhere, get confirmations by email and one reminder text a month. Mobile app plus a web portal.
- B. Marketplace with photos and payments. Buyers browse listings with photos, sellers upload them, 10% of users verify a phone number by SMS code each month. Card payment fees are excluded (see the limits below).
- C. Portal with an AI assistant. Profile A plus three assistant conversations per user a month.
| Per active user, per month | A: portal | B: marketplace | C: portal with AI |
|---|---|---|---|
| Database document reads (Firestore) | 400 | 1,500 | 400 |
| Database writes / deletes | 30 / 3 | 40 / 5 | 30 / 3 |
| Database data stored per user | 0.5 MB | 1 MB | 0.5 MB |
| Files stored per user | 2 MB | 4 MB | 2 MB |
| File downloads (egress) | 5 MB, 20 files | 60 MB, 300 images | 5 MB, 20 files |
| File uploads | 2 | 3 | 2 |
| API response egress | 10 MB | 20 MB | 10 MB |
| Web front end egress | 1 MB | 1 MB | 1 MB |
| Cloud or edge function calls | 60 | 100 | 90 |
| API calls on the AWS stack | 200 | 400 | 230 |
| Emails | 6 | 8 | 6 |
| Text messages (one segment each) | 1 | 0.5 | 1 |
| Phone number verifications (SMS code) | 0 | 0.1 | 0 |
| Push notifications | 10 (free) | 10 (free) | 10 (free) |
| AI assistant conversations | 0 | 0 | 3 |
Our assumptions, not measured data from any one app.
The three backends, each priced as a complete stack:
- Firebase (Blaze plan): Cloud Firestore, Firebase Authentication, Cloud Storage, Cloud Functions and Firebase Hosting, us-central1 (Iowa) prices.
- Supabase (Pro plan): Postgres database with a compute size per tier (Micro at 1,000 users, Small at 10,000, Large at 100,000, our sizing assumption), Supabase Auth, Storage with its CDN, Edge Functions, and a Vercel Pro seat for the web front end.
- AWS (US East, N. Virginia): API Gateway HTTP API, Lambda on Arm (256 MB, 100 ms average), RDS for PostgreSQL single-AZ (db.t4g.micro, db.t4g.small, then db.m7g.large, matched to the Supabase memory sizes), S3, CloudFront, Cognito Lite and one NAT gateway so functions inside the database’s VPC can call outside services.
Shared by all three: Amazon SES for email, Twilio for texts, Sentry for error monitoring, Firebase Cloud Messaging and Apple’s push service for notifications, and Apple’s developer fee. For AI we priced OpenAI and Anthropic.
How we did the arithmetic:
- Daily free quotas (Firestore, Firebase phone SMS, Hosting transfer) are counted as 30 days a month, assuming usage is spread evenly.
- Hourly prices are counted at 730 hours a month. We convert MB to GB at 1,024, because these vendors bill in GiB on these lines; the difference is about 7% on storage lines and does not change any conclusion.
- Text messages are priced at Twilio’s $0.0083 US base rate plus a $0.0045 carrier fee, the T-Mobile long code outbound fee and within the AT&T ($0.0035) to Verizon ($0.005) range on the same page.
- CloudFront is priced at the cheaper of its flat-rate plan and pay as you go for that month’s traffic. We assumed 20% of file downloads miss the CDN cache and reach S3.
- AWS’s 12-month free offers (such as API Gateway’s first million calls) and the up to $200 in credits AWS gives new accounts are left out, because they expire. Always-free quotas, such as Lambda’s 1 million requests and 400,000 GB-seconds a month, are included.
What we left out and why:
- Developer time. Fixes, OS updates and store deadlines are a real running cost, usually larger than the infrastructure bill, and we already price them in hours on our app development cost study and the app maintenance cost guide. We do not repeat those tables here.
- Card payment fees. They scale with sales, not users. Stripe’s US rate is 2.9% + 30 cents per card payment; our App Store and Google Play fees paper works it into a $9.99 subscription.
- Maps, video, search and other third-party APIs. Each is priced per call by its provider; our API integration cost guide covers how to budget for them.
- Backups beyond the defaults, staging environments, high-availability database replicas, support plans, taxes and committed-use discounts.
- Firestore multi-region prices. Google’s page shows Iowa (us-central1) prices by default. Multi-region locations such as nam5 cost more, and we did not quote them.
What every line item costs: the vendor prices we used
| Line | Firebase (Blaze) | Supabase (Pro) | AWS (US East) |
|---|---|---|---|
| Base monthly fee | $0 | $25, includes $10 of compute credit | $0 (about $36.50 for a NAT gateway and its IP address if functions run inside a VPC) |
| Database | Firestore: $0.03 per 100,000 reads, $0.09 per 100,000 writes, $0.01 per 100,000 deletes, about $0.15 per GiB a month | Compute: Micro $10, Small $15, Medium $60, Large $110 a month; 8 GB disk included, then $0.125 per GB | RDS PostgreSQL: db.t4g.micro $0.016, db.t4g.small $0.032, db.m7g.large $0.168 an hour; gp3 storage $0.115 per GB |
| Free database quota | 50,000 reads, 20,000 writes, 20,000 deletes a day; 1 GiB stored | Inside the plan’s included disk and compute credit | None counted (new-account credits excluded) |
| Sign-in (auth) | 50,000 MAU free, then $0.0055 per MAU to 100,000 and $0.0046 to 1 million | 100,000 MAU included, then $0.00325 per MAU | Cognito: 10,000 MAU free; Lite $0.0055 per MAU, Essentials (the default for new pools) $0.015 |
| Phone sign-in by SMS code | $0.01 per US SMS; first 10 a day free | Sent through a provider you configure (Twilio, MessageBird, Vonage or TextLocal); we priced Twilio Verify | Sent through Amazon SNS (not priced here); we priced Twilio Verify |
| File storage | 5 GB free, then about $0.02 per GB a month | 100 GB included, then $0.0213 per GB | S3 Standard $0.023 per GB |
| File and API egress | 100 GB of file downloads free, then $0.12 per GB; Firestore 10 GiB free, then $0.12 | 250 GB uncached included, then $0.09 per GB; 250 GB cached, then $0.03 | CloudFront flat-rate: Free $0 (100 GB), Pro $15 (50 TB, 10M requests), Business $200 (50 TB, 125M requests); or $0.085 per GB |
| Server functions | 2M calls free, then $0.40 per million, plus compute above the free quota | 2M Edge Function calls included, then $2 per million | Lambda $0.20 per million plus $0.0000133334 per GB-second (Arm); API Gateway $1 per million |
| Free push notifications | Yes FCM is no-cost | Yes send through FCM or APNs directly | Yes send through FCM or APNs directly |
| $0 backend for our 1,000-user portal | Yes inside Blaze no-cost quotas | No Pro needed; the Free plan has 5 GB egress and 1 GB file storage | No Lambda has an always-free quota, but the database is paid from day one |
| Best for | Small and read-light apps; no fixed floor | Postgres apps from about 10,000 users; generous included quotas | Teams already on AWS; media-heavy apps that fit a flat-rate CDN plan |
Prices from Firebase pricing, Firestore pricing, Identity Platform pricing, Cloud Storage pricing, Supabase pricing, Supabase compute, the AWS Price List API for us-east-1 and AWS’s Cognito, CloudFront and VPC pages, all checked Nov 11, 2026. The NAT gateway rate is the $0.045 an hour in AWS’s own VPC pricing example, which uses US East (Ohio).
The shared services, the same on every stack:
| Service | Price used | Free tier | Source, checked Nov 11, 2026 |
|---|---|---|---|
| Email, Amazon SES | $0.10 per 1,000 emails | Up to $200 of AWS credits for new accounts, not counted | Amazon SES pricing |
| Text message, Twilio US | $0.0083 per segment plus carrier fees ($0.0035 to $0.005 outbound on long codes) | None | Twilio US SMS pricing |
| SMS sign-in code, Twilio Verify | $0.05 per successful verification plus $0.0083 per SMS | None | Twilio Verify pricing |
| Push, Firebase Cloud Messaging | $0 | No-cost on both plans | Firebase pricing |
| Error monitoring, Sentry | Team $26 a month billed annually ($29 monthly), 50,000 errors | Developer plan: 1 user, 5,000 errors | Sentry pricing |
| Web front end, Vercel | Pro $20 a month per deploying seat, 1 TB transfer and 10M CDN requests included | Hobby is $0 but for personal, non-commercial use only | Vercel pricing |
| Apple Developer Program | $99 a year, counted as $8.25 a month | Fee waivers for eligible nonprofits, education and government | Apple Developer: What’s included |
Google Play’s one-time $25 registration is not in the monthly totals. Store commissions by country and payment route are in our App Store and Google Play fees by country research.
Profile A: what a booking or customer portal costs to run
At 1,000 users the portal costs $22 a month on Firebase and about $67 to $72 on Supabase or AWS, because those two have fixed floors. At 100,000 users it costs $1,607 to $2,223, and most of that is text messages.
| Line item, Profile A | 1,000 MAU | 10,000 MAU | 100,000 MAU |
|---|---|---|---|
| Email (Amazon SES) | $0.60 | $6.00 | $60 |
| Text reminders (Twilio, 1 per user) | $13 | $128 | $1,280 |
| Push notifications | $0 | $0 | $0 |
| Error monitoring (Sentry) | $0 | $26 | $26 |
| Apple Developer Program, per month | $8.25 | $8.25 | $8.25 |
| Backend and hosting on Firebase | $0 | $12 | $479 |
| Backend and hosting on Supabase | $45 | $50 | $233 |
| Backend and hosting on AWS | $51 | $80 | $848 |
| Total with Firebase | $22 | $181 | $1,853 |
| Total with Supabase | $67 | $218 | $1,607 |
| Total with AWS | $72 | $248 | $2,223 |
Illustrative, based on our assumptions. Vendor prices checked Nov 11, 2026. Rounded to the dollar above $10, so lines may not add exactly.
Inside the backend lines at 100,000 users:
- Firebase, $479: Authentication $275 (50,000 users above the free 50,000 at $0.0055), Firestore egress $116, Cloud Storage $52, Firestore reads and writes $14, Hosting $13, stored data $7, functions $1.60.
- Supabase, $233: Large compute $100 after the $10 credit, uncached egress $65, Pro plan $25, Vercel $20, Edge Functions $8, cached egress $7, disk $5, file storage $2. Auth is $0 because Pro includes 100,000 users.
- AWS, $848: Cognito Lite $495, CloudFront $158 (pay as you go was cheaper than the $200 flat plan here), RDS $128, NAT gateway $36.50, API Gateway $20, S3 $6, Lambda $5.
Profile B: what a marketplace with photos costs to run
The marketplace is the profile where the backends split furthest apart, because photo downloads are priced very differently on each. At 100,000 users the bill was $1,837 on Supabase, $2,153 on Firebase and $2,271 on AWS.
| Line item, Profile B | 1,000 MAU | 10,000 MAU | 100,000 MAU |
|---|---|---|---|
| Email (Amazon SES) | $0.80 | $8.00 | $80 |
| Text notifications (Twilio, 0.5 per user) | $6.40 | $64 | $640 |
| Push notifications | $0 | $0 | $0 |
| Error monitoring (Sentry) | $0 | $26 | $26 |
| Apple Developer Program, per month | $8.25 | $8.25 | $8.25 |
| Backend and hosting on Firebase, incl. phone sign-in | $1.24 | $95 | $1,398 |
| Backend and hosting on Supabase, incl. phone sign-in | $51 | $119 | $1,083 |
| Backend and hosting on AWS, incl. phone sign-in | $57 | $141 | $1,517 |
| Total with Firebase | $17 | $201 | $2,153 |
| Total with Supabase | $66 | $225 | $1,837 |
| Total with AWS | $72 | $248 | $2,271 |
Illustrative, based on our assumptions. Vendor prices checked Nov 11, 2026.
The backend lines in full, because this is the comparison most people searching “Firebase vs Supabase cost” want to see:
| Backend line, Profile B | 1,000 MAU | 10,000 MAU | 100,000 MAU |
|---|---|---|---|
| Firebase | |||
| Firestore reads, writes, deletes | $0 | $4.05 | $48 |
| Firestore stored data | $0 | $1.31 | $14 |
| Firestore network egress | $1.14 | $22 | $233 |
| Cloud Storage (photos stored and downloaded) | $0.10 | $60 | $712 |
| Cloud Functions | $0 | $0 | $5.85 |
| Authentication | $0 | $0 | $275 |
| Phone sign-in SMS ($0.01 each) | $0 | $7.00 | $97 |
| Hosting | $0 | $0 | $13 |
| Firebase subtotal | $1.24 | $95 | $1,398 |
| Supabase | |||
| Pro plan | $25 | $25 | $25 |
| Compute after $10 credit (Micro, Small, Large) | $0 | $5.00 | $100 |
| Database disk above 8 GB | $0 | $0.22 | $11 |
| Uncached egress (API) | $0 | $0 | $153 |
| Cached egress (photos through the CDN) | $0 | $10 | $168 |
| File storage above 100 GB | $0 | $0 | $6.19 |
| Edge Functions | $0 | $0 | $16 |
| Authentication | $0 | $0 | $0 |
| Phone sign-in (Twilio Verify) | $5.83 | $58 | $583 |
| Web front end (Vercel Pro) | $20 | $20 | $20 |
| Supabase subtotal | $51 | $119 | $1,083 |
| AWS | |||
| API Gateway (HTTP API) | $0.40 | $4.00 | $40 |
| Lambda | $0 | $0.60 | $16 |
| RDS PostgreSQL and storage (t4g.micro, t4g.small, m7g.large) | $14 | $26 | $134 |
| S3 storage and requests | $0.13 | $1.29 | $13 |
| CloudFront (flat-rate Free, Pro, Business) | $0 | $15 | $200 |
| Cognito Lite | $0 | $0 | $495 |
| NAT gateway and its public IPv4 address | $36 | $36 | $36 |
| Phone sign-in (Twilio Verify) | $5.83 | $58 | $583 |
| AWS subtotal | $57 | $141 | $1,517 |
Illustrative, based on our assumptions. Vendor prices checked Nov 11, 2026.
Two lines decide this comparison. Photo egress: 5.9 TB a month at 100,000 users costs $712 through Firebase Cloud Storage, $168 as Supabase cached egress and nothing extra inside CloudFront’s $200 Business plan, which allows 50 TB. And phone sign-in: Firebase’s own phone auth charges $0.01 per US text, while Supabase and our AWS stack send codes through a provider such as Twilio Verify at $0.0583 per successful code, so 10,000 verifications a month cost $97 against $583.
Profile C: what an AI assistant adds to the bill
On a small model, the assistant is about a fifth of the bill. On a mid model, it becomes the bill. At 100,000 users and three conversations each, the AI line is $450 a month on a small model and $9,000 on a mid model.
| Line item, Profile C | 1,000 MAU | 10,000 MAU | 100,000 MAU |
|---|---|---|---|
| Email (Amazon SES) | $0.60 | $6.00 | $60 |
| Text reminders (Twilio) | $13 | $128 | $1,280 |
| Error monitoring (Sentry) | $0 | $26 | $26 |
| Apple Developer Program, per month | $8.25 | $8.25 | $8.25 |
| AI assistant, small model | $4.50 | $45 | $450 |
| Backend and hosting on Firebase | $0 | $12 | $482 |
| Backend and hosting on Supabase | $45 | $50 | $239 |
| Backend and hosting on AWS | $51 | $80 | $856 |
| Total with Firebase | $26 | $226 | $2,306 |
| Total with Supabase | $71 | $263 | $2,063 |
| Total with AWS | $77 | $294 | $2,680 |
| AI line on a mid model instead | $90 | $900 | $9,000 |
| Total with Supabase and a mid model | $157 | $1,118 | $10,613 |
Illustrative, based on our assumptions. Vendor prices checked Nov 11, 2026.
How we priced one AI conversation
We assumed a conversation of five exchanges where each turn resends a system prompt, some retrieved account data and the growing history: 10,000 input tokens and 1,000 output tokens in total. That is an assumption, not a measurement; your own prompt design can move it a lot in either direction.
| Model | Input per 1M tokens | Output per 1M tokens | Per conversation | Per 1,000 conversations | 100,000 users x 3 a month |
|---|---|---|---|---|---|
| OpenAI gpt-6-luna (small) | $0.10 | $0.50 | $0.0015 | $1.50 | $450 |
| Claude Haiku 5.5 (small, prompts up to 100,000 tokens) | $0.10 | $0.50 | $0.0015 | $1.50 | $450 |
| OpenAI gpt-6.1-terra | $1.00 | $5.00 | $0.015 | $15 | $4,500 |
| OpenAI gpt-6.1-sol (mid) | $2.00 | $10.00 | $0.03 | $30 | $9,000 |
| Claude Sonnet 5.5 (mid) | $2.00 | $10.00 | $0.03 | $30 | $9,000 |
Prices from OpenAI API pricing and Anthropic Claude pricing, standard tier, checked Nov 11, 2026. Illustrative token counts, our assumption.
Three notes before you use these numbers:
- Prompt caching cuts the input side. Cached input is $0.10 per million on both gpt-6.1-sol and Sonnet 5.5, against $2 uncached. If 70% of each conversation’s input were cache hits, the mid-model cost would fall from $30 to about $16.70 per 1,000 conversations, before the small premium both vendors charge to write the cache.
- Tokens are not the same size across vendors. Anthropic says the tokenizer used by Claude 4.7 and later models produces about 30% more tokens for the same text than its previous tokenizer. Count tokens with each vendor’s own tools before comparing.
- Claude Haiku 5.5 has two prices. The $0.10 and $0.50 rates apply to prompts up to 100,000 tokens; longer prompts are $0.50 and $2.50.
Where the money goes at each tier
On the Firebase stack, the backend is 0% to 26% of the bill for the portal apps. Text messages are the largest share at 10,000 and 100,000 users in Profiles A and C, while the marketplace’s photo traffic makes its backend the largest share at 100,000.
| Profile and tier (Firebase stack) | Backend and hosting | SMS and sign-in codes | AI (small model) | Monitoring and Apple fee | Total | |
|---|---|---|---|---|---|---|
| A, 1,000 MAU | $0 | $13 | $0.60 | $0 | $8.25 | $22 |
| A, 10,000 MAU | $12 | $128 | $6.00 | $0 | $34 | $181 |
| A, 100,000 MAU | $479 | $1,280 | $60 | $0 | $34 | $1,853 |
| B, 1,000 MAU | $1.24 | $6.40 | $0.80 | $0 | $8.25 | $17 |
| B, 10,000 MAU | $88 | $71 | $8.00 | $0 | $34 | $201 |
| B, 100,000 MAU | $1,301 | $737 | $80 | $0 | $34 | $2,153 |
| C, 1,000 MAU | $0 | $13 | $0.60 | $4.50 | $8.25 | $26 |
| C, 10,000 MAU | $12 | $128 | $6.00 | $45 | $34 | $226 |
| C, 100,000 MAU | $482 | $1,280 | $60 | $450 | $34 | $2,306 |
Illustrative, based on our assumptions. In Profile B, phone sign-in SMS ($7 and $97) moves from backend to the SMS column. Vendor prices checked Nov 11, 2026.
Firebase vs Supabase vs AWS: the same app on three backends
At 100,000 users Supabase had the lowest backend bill in all three profiles and AWS the highest. The AWS gap is mostly sign-in: Cognito’s Lite tier charges $0.0055 per user above 10,000, so 100,000 users cost $495, against $275 on Firebase and $0 on Supabase Pro.
| Backend bill at 100,000 MAU | Firebase | Supabase | AWS |
|---|---|---|---|
| A: portal | $479 | $233 | $848 |
| B: marketplace | $1,398 | $1,083 | $1,517 |
| C: portal with AI | $482 | $239 | $856 |
Illustrative, based on our assumptions. Vendor prices checked Nov 11, 2026.
At small scale the order flips. Firebase had no fixed fee in our model, so at 1,000 users its full marketplace bill was $17 against $66 on Supabase and $72 on AWS. Supabase’s floor is the $25 Pro plan plus a $20 Vercel seat (Vercel’s free Hobby plan is for personal, non-commercial use). AWS’s floor is a small RDS instance and a NAT gateway. The chart below shows the full bill per active user for the marketplace, so the three tiers sit on one scale.
| Profile B, full bill per MAU per month | Firebase | Supabase | AWS |
|---|---|---|---|
| 1,000 MAU | $0.017 | $0.066 | $0.072 |
| 10,000 MAU | $0.020 | $0.022 | $0.025 |
| 100,000 MAU | $0.022 | $0.018 | $0.023 |
Illustrative, based on our assumptions. Vendor prices checked Nov 11, 2026.
Two cautions on reading the comparison. First, the database sizing is ours: we put Supabase on Large compute and AWS on a db.m7g.large (both 2 vCPUs and 8 GB of memory) at 100,000 users, and a busier or lighter app would need a different size on either. Firebase has no instance to size, so its database cost moves only with reads, writes and storage. Second, the AWS stack is the most configurable: choosing Cognito’s default Essentials tier instead of Lite would add $855 a month at 100,000 users, and keeping functions out of a VPC would remove the NAT gateway. Whether you need a server of your own at all is covered in do I need a backend for my app.
Cost per active user per month
| Profile and backend | 1,000 MAU | 10,000 MAU | 100,000 MAU |
|---|---|---|---|
| A: portal, Firebase | $0.022 | $0.018 | $0.019 |
| A: portal, Supabase | $0.067 | $0.022 | $0.016 |
| A: portal, AWS | $0.072 | $0.025 | $0.022 |
| B: marketplace, Firebase | $0.017 | $0.020 | $0.022 |
| B: marketplace, Supabase | $0.066 | $0.022 | $0.018 |
| B: marketplace, AWS | $0.072 | $0.025 | $0.023 |
| C: portal with AI (small model), Firebase | $0.026 | $0.023 | $0.023 |
| C: portal with AI (small model), Supabase | $0.071 | $0.026 | $0.021 |
| C: portal with AI (small model), AWS | $0.077 | $0.029 | $0.027 |
| C: portal with AI (mid model), Firebase | $0.112 | $0.108 | $0.109 |
| C: portal with AI (mid model), Supabase | $0.157 | $0.112 | $0.106 |
| C: portal with AI (mid model), AWS | $0.162 | $0.115 | $0.112 |
Full monthly bill, including shared lines, divided by monthly active users. Illustrative, based on our assumptions. Vendor prices checked Nov 11, 2026.
The per-user cost falls as fixed fees spread out, then flattens near 2 cents because almost every remaining line is billed per use. A mid-size AI model is the exception: on its own it costs $0.09 per user a month at three conversations each.
What makes an app’s running cost jump?
Five lines grow faster than the user count, and each one can outrun the backend.
| 100,000 messages in a month | Unit price | Monthly cost |
|---|---|---|
| SMS sign-in codes, Twilio Verify | $0.05 per success + $0.0083 per SMS | $5,830 |
| Text messages, Twilio US | $0.0083 + about $0.0045 carrier fee | $1,280 |
| Firebase phone sign-in SMS, US | $0.01, first 10 a day free | $997 |
| Email, Amazon SES | $0.10 per 1,000 | $10 |
| Push, Firebase Cloud Messaging | No-cost | $0 |
Vendor prices checked Nov 11, 2026. The carrier fee is T-Mobile’s fee on Twilio’s page, which sits inside the AT&T to Verizon range.
- Text messages and SMS codes. A text costs about 128 times an email in this table. SMS is the most expensive way to send a reminder, receipt or code. Twilio also counts SMS per 160-character segment, so a long text can cost two or three times the base rate.
- Egress from images and video. File downloads are billed per GB on most backends. The marketplace’s 60 MB of photos per user a month became 5.9 TB at 100,000 users and $712 on Firebase Cloud Storage. Unresized phone photos or autoplay video can multiply that.
- AI tokens. A mid model is 20 times the price of a small one for the same conversation, and long conversations resend their history each turn, so input tokens grow faster than the number of messages.
- Over-fetching database reads. Firestore bills each document read. In Profile B, reads cost about $45 a month at 100,000 users. A screen that loads five times as many documents as it shows, or a live listener that re-reads a whole list on every change, would push that to about $225.
- Sign-in pricing tiers. Auth is free at small scale on all three stacks and then is not. Above 50,000 users Firebase charges $0.0055 per user; above 10,000 Cognito charges $0.0055 on Lite or $0.015 on Essentials; Supabase Pro includes 100,000 then charges $0.00325.
How to keep the bill low
- Send reminders by push, not SMS. In Profile A, moving the one monthly text to push would cut the 100,000-user bill by $1,280, from $1,853 to $573 on Firebase.
- Use passwordless email, Sign in with Apple or Google, or passkeys instead of SMS codes where you can. If you need phone verification, compare Firebase’s $0.01 per SMS with a Verify product at $0.0583 per success, and use rate limits and CAPTCHA on the sending endpoint, as Supabase’s own phone login guide recommends.
- Resize images on upload and serve them through a CDN. On Supabase, cached egress is $0.03 per GB against $0.09 uncached. On AWS, a flat-rate CloudFront plan covers up to 50 TB for $15 or $200 a month.
- Route most AI turns to a small model and send only the hard ones to a mid model. Cache the system prompt and fixed context, which both OpenAI and Anthropic bill at a fraction of the normal input price.
- Paginate and limit queries. Load the 20 documents a screen shows, not the 200 in the collection, and avoid listeners on large lists.
- Pick the auth tier on purpose. On AWS, check whether your user pool is on Lite or Essentials before you reach 10,000 users.
- Right-size the database every quarter. A compute size picked at launch is often too big for a year and then too small in a month.
How to use this data for your own app
Start with the profile closest to yours and change the per-user assumptions in the method table to your own: how many screens a user opens, how many photos they load, how many texts and emails you send. Multiply by your expected users at launch, at one year and at your target, and price each line with the vendor rates above. If you have a developer or agency, ask them for exactly that: a line-item estimate at 1,000, 10,000 and 100,000 users, with the assumptions written down. A quote that says “hosting: $200 a month” with no usage behind it is a guess.
Then put the running cost next to the people cost. At 100,000 users every app in our model ran for $1,607 to $2,680 a month before a mid-size AI model, which is often less than a few days of developer time for fixes and updates. Our software maintenance and support page explains how that ongoing work is scoped, and our app cost estimator covers the build itself.
Choosing a backend: if you expect to stay under about 10,000 users for the first year, the backend with no fixed fee (Firebase in our model) is cheapest. At 100,000 users, the largest tier we priced, Supabase’s included auth and egress quotas made it cheapest in every profile. Pick AWS when your team already runs it, when you need its compliance or networking options, or when heavy media fits inside a flat-rate CloudFront plan.
How we can help
We are aibrevo, founded in 2021 in San Francisco by Alpit Patel, with a 16-person in-house engineering team led by our CTO, Shivam. Our app and web app builds run from $7,000 for a focused first version to about $25,000-$30,000 for many features, user roles, payments and integrations, with larger builds quoted on scope. After a free 30-minute call we send a fixed-price quote in writing for project builds, and we also work hourly. You own the code, and your app is published under your own Apple and Google accounts. Send us a message if you want your own app priced line by line.
Changelog for this page
- November 11, 2026: published. Three reference profiles priced at 1,000, 10,000 and 100,000 monthly active users from Firebase, Google Cloud, Supabase, Vercel, AWS (Price List API for us-east-1 and pricing pages), Amazon SES, Twilio, Sentry, OpenAI, Anthropic and Apple pages, all checked November 11, 2026.
Sources
- Firebase, Pricing: Spark and Blaze quotas for Firestore, Authentication, Cloud Storage, Cloud Functions, Hosting and Cloud Messaging, checked Nov 11, 2026.
- Google Cloud, Firestore Standard edition pricing: Iowa (us-central1) reads, writes, deletes, storage, free quota and network egress, checked Nov 11, 2026.
- Google Cloud, Identity Platform pricing: MAU tiers and US phone authentication SMS price, checked Nov 11, 2026.
- Google Cloud, Cloud Storage pricing: regional Standard storage, Class A and B operations, data transfer to the internet, checked Nov 11, 2026.
- Google Cloud, Cloud Functions pricing, which now shows Cloud Run pricing: request-based CPU and memory rates in us-central1, checked Nov 11, 2026.
- Supabase, Pricing, Compute and disk, Egress and Phone login, checked Nov 11, 2026.
- Vercel, Pricing, checked Nov 11, 2026.
- Amazon Web Services, AWS Price List API offer files for us-east-1: Lambda, API Gateway, RDS (including Aurora PostgreSQL), S3, SES and Cognito (publication dates Oct 29 to Nov 9, 2026), and the CloudFront global offer file (Oct 19, 2026), read Nov 11, 2026.
- Amazon Web Services, Lambda pricing, API Gateway pricing, Cognito pricing, CloudFront pricing, Amazon SES pricing and VPC pricing, checked Nov 11, 2026.
- Twilio, US SMS pricing and Verify pricing, checked Nov 11, 2026.
- Sentry, Pricing, checked Nov 11, 2026.
- OpenAI, API pricing, standard tier, checked Nov 11, 2026.
- Anthropic, Claude pricing, checked Nov 11, 2026.
- Apple Developer, What’s included, checked Nov 11, 2026.