Firebase Pricing in 2026: What the Blaze Plan Actually Costs at 1K, 10K and 100K Users

The short answer

Firebase is close to free for a small app and then gets expensive in steps, not in a smooth line. In our model of a typical AI-built SaaS app, the Blaze plan costs about $0 a month at 1,000 users, about $19 at 10,000 users, and about $600 at 100,000 users. Past 50,000 monthly active users, sign-ins become the single biggest line on the bill.

Figures use Google's published rates as of October 2026 and the usage assumptions in the scenarios section below. Your bill depends mostly on how many reads your app makes per user, so check that number first.

How Firebase pricing works in 2026

Firebase has two plans. Spark is free with hard limits and no card on file. Blaze is pay as you go: it keeps the same free allowances, then bills every unit above them, with no monthly base fee. Most real apps need Blaze, because Spark has no Cloud Storage and no Cloud Functions.

The free allowances on Firestore reset daily, not monthly. A traffic spike on one day is billed even if the rest of the month is quiet.

Firestore rates shown are Standard edition, on-demand. Firestore Enterprise edition, Data Connect (managed Postgres from about $9.37/month after a 3-month trial) and AI Logic are billed separately.

Sources: Firebase pricing, Firestore pricing, Identity Platform pricing.

What it costs at 1K, 10K and 100K users

The bill grows about 30 times between 10,000 and 100,000 users, while the user count grows 10 times. That jump comes from two lines that are free at small scale: authentication above 50,000 MAU, and file downloads above 100 GB a month.

The app we modelled: a typical AI-built SaaS tool (a CRM, booking app or client portal). 30% of monthly users are active on a given day. Each active user makes 200 document reads, 20 writes and 2 deletes a day, keeps about 0.5 MB in Firestore and 5 MB in uploaded files, and downloads 3x what they upload each month. Sign-in is email or Google, with no SMS.

Two things change these numbers more than anything else:

• Reads per user. A dashboard that loads 50 documents on every open, or a real-time listener on a busy collection, can push reads to 1,000+ per user per day. That multiplies the reads line by 5.
• Legacy auth. Some older projects still on Firebase Authentication without the Identity Platform upgrade do not pay per MAU for email and social sign-in. Check your project's billing before assuming the $275.

Five billing surprises AI-built apps hit

AI coding agents write Firebase code that works, but they rarely write it to be cheap. These are the five patterns we see most often.

1. Queries with no limit. An agent writes getDocs(collection(db, "orders")) to fill a table. Every document is a billed read, every time the page opens. Firebase's own guide lists unbounded queries as the first cause of surprise bills.
2. Real-time listeners everywhere. onSnapshot re-bills reads each time a matching document changes. A shared activity feed with 1,000 viewers turns one write into 1,000 reads.
3. Function loops. A Cloud Function that writes to the same collection that triggers it can call itself until you notice. Firebase names fan-out and infinite loops as a top cause of runaway costs.
4. No real spend cap. Budget alerts only send emails; Google states that they do not pause services. Hard caps exist only for AI Logic, App Hosting, Cloud Functions and Extensions, not for Firestore, Storage or Auth.
5. Testing against production. Agents often run test scripts against the live project. Every seeded record and every test read is billed. Use the Local Emulator Suite instead.

Source: Firebase: avoid surprise bills.

When Firebase is worth it, and when to switch

Firebase is the cheapest option until roughly 50,000 monthly users. After that, flat-plan Postgres backends cost about half as much for the same workload, mainly because they include sign-ins in the plan price.

Stay on Firebase if:

• You are under 50,000 MAU and your reads per user are under control.
• Your app is mobile-first and leans on offline sync, push notifications (FCM) or Crashlytics.
• Your data is document-shaped and you rarely need joins or reports.

Consider switching if:

• You are heading past 50,000 MAU and authentication is becoming your largest line.
• You need a hard monthly ceiling on database and storage costs, not an email alert.
• Your data is relational (customers, invoices, bookings) and the agent keeps duplicating data across collections to fake joins.

The same modelled workload on flat-plan alternatives:

At 100K MAU, bandwidth is most of the bill on every platform ($157 of it on both Supabase and Butterbase). Cutting downloads with a CDN or image resizing saves more than switching vendors.

The two flat plans split on what the base price covers. If your app only needs a database and logins, Supabase is the cheapest of the three at 100K MAU. If it also calls AI models or searches its own content, which most AI-built apps end up doing, the comparison shifts: on Firebase that means adding Vertex AI and a separate vector store, two more per-use bills on top of the five surprises above. Butterbase includes the AI gateway and search in the same plan and starts at $19 rather than $25, so the number in the table is closer to what you actually pay once AI features ship.

If you do switch, read The Cost of Switching Backends: Migrating from Firebase or Supabase first, then compare options in Supabase vs Firebase in 2026 and Butterbase vs Firebase.

Firebase pricing FAQ

Prices checked October 2026 against Firebase, Google Cloud, Supabase and Butterbase pricing pages. Cost scenarios are estimates for the workload described above.