# Fly.io Pricing in 2026: Plans, Real Costs and Alternatives
> Full breakdown of Fly.io pricing in 2026. Machines, Postgres, free tier changes, real app costs, and why teams are switching to Kuberns as an alternative.
- **Author**: manav-dobariya
- **Published**: 2026-04-28
- **Modified**: 2026-04-28
- **Category**: Alternatives
- **URL**: https://kuberns.com/blogs/flyio-pricing/
---
Fly.io pricing looks straightforward on the surface. Then you add a database, configure a second service, and factor in volumes and bandwidth, and the real cost of running a production app becomes less obvious.
On top of that, Fly.io removed its free tier in 2024. Developers who were running apps for free now pay for every Machine and every GB of storage. And unlike simpler platforms, every app on Fly.io needs a Dockerfile and a `fly.toml` config file before it can deploy at all.
This guide breaks down exactly what Fly.io costs in 2026, where developers hit unexpected charges, and why teams are switching to [Kuberns](https://kuberns.com) as a smarter alternative.
## Fly.io Pricing at a Glance
Fly.io does not have fixed monthly plans. You pay for what you use, billed per second based on the resources your Machines consume. There is no workspace seat fee, but every resource including Machines, volumes, and bandwidth is metered.
| Resource | Pricing |
|---|---|
| Shared CPU Machine (256 MB) | ~$1.94/month always-on |
| Shared CPU Machine (512 MB) | ~$3.19/month always-on |
| Shared CPU Machine (1 GB) | ~$5.70/month always-on |
| Performance CPU Machine (2 GB) | ~$99/month always-on |
| Volume storage | $0.15/GB/month |
| Outbound bandwidth | $0.02/GB after first 100 GB/month |
| Fly Postgres (shared-cpu-1x, 256 MB + 1 GB volume) | ~$2.09/month |
New accounts receive $5 in free trial credits. After that, a credit card is required and all resources are billed.
## Fly.io Machine Pricing: Shared vs Performance
Fly.io Machines are micro-VMs that run your application. You pay per second for CPU and RAM, billed only when the Machine is running.
### Shared CPU Machines
Shared CPU Machines share physical cores with other workloads on the same host. Suitable for most web apps, APIs, and background workers.
| Machine Size | vCPU | RAM | Monthly (always-on) |
|---|---|---|---|
| shared-cpu-1x | 1 shared | 256 MB | ~$1.94 |
| shared-cpu-1x | 1 shared | 512 MB | ~$3.19 |
| shared-cpu-1x | 1 shared | 1 GB | ~$5.70 |
| shared-cpu-1x | 1 shared | 2 GB | ~$10.58 |
| shared-cpu-2x | 2 shared | 4 GB | ~$21.17 |
| shared-cpu-4x | 4 shared | 8 GB | ~$42.34 |
| shared-cpu-8x | 8 shared | 16 GB | ~$84.67 |
### Performance CPU Machines
Performance Machines have dedicated CPU cores. Suitable for workloads that need consistent, uncontested compute.
| Machine Size | vCPU | RAM | Monthly (always-on) |
|---|---|---|---|
| performance-1x | 1 dedicated | 2 GB | ~$99.07 |
| performance-2x | 2 dedicated | 4 GB | ~$198.14 |
| performance-4x | 4 dedicated | 8 GB | ~$396.29 |
| performance-8x | 8 dedicated | 16 GB | ~$792.58 |
### Scale to Zero
Fly.io supports scale-to-zero. Machines that receive no traffic can stop automatically, pausing compute billing. When a new request arrives, the Machine starts and handles it.
The tradeoff is a cold start delay of a few seconds on the first request after idle. For production APIs where response time matters, this hurts user experience. For internal tools or batch workloads, it is a valid cost control option.
## Fly.io Postgres Pricing
Fly Postgres is not a fully managed database. It is a Postgres app running inside a Fly Machine with a persistent volume attached. You manage backups, version upgrades, and failover yourself. Fly provides tooling (`fly postgres` commands) but the operational responsibility is yours.
| Component | Price |
|---|---|
| Machine (shared-cpu-1x, 256 MB) | ~$1.94/month |
| Machine (shared-cpu-1x, 1 GB) | ~$5.70/month |
| Volume (1 GB) | $0.15/month |
| Volume (10 GB) | $1.50/month |
| Volume (50 GB) | $7.50/month |
| Snapshot backups | $0.02/GB/month |
A minimal Fly Postgres setup (shared-cpu-1x, 256 MB, 1 GB volume) costs roughly $2.09 per month. A practical production database (1 GB RAM, 10 GB storage) runs around $7.20 per month. For high-availability with automatic failover, you run a primary and a replica, which doubles the Machine cost.
## Fly.io Volume and Bandwidth Pricing
Volumes are persistent block storage required for any stateful workload. They are billed whether or not the attached Machine is running.
| Storage | Monthly Price |
|---|---|
| 1 GB | $0.15 |
| 10 GB | $1.50 |
| 50 GB | $7.50 |
| 100 GB | $15.00 |
| 500 GB | $75.00 |
| Bandwidth | Price |
|---|---|
| First 100 GB outbound/month | Free |
| Outbound after 100 GB | $0.02/GB |
| Inbound | Free |
The volume billing regardless of Machine state is worth noting. If you use scale-to-zero to save on compute, your volumes keep accruing charges. A stopped app is not a free app.
## What a Real App Costs on Fly.io in 2026
Here is a realistic cost estimate for a typical production app: a Node.js API with two instances for redundancy and a Postgres database.
| Resource | Spec | Monthly Cost |
|---|---|---|
| API Machine x2 (always-on) | shared-cpu-1x, 512 MB each | ~$6.38 |
| Fly Postgres Machine | shared-cpu-1x, 1 GB RAM | ~$5.70 |
| Postgres Volume | 10 GB | $1.50 |
| Outbound bandwidth | Up to 100 GB | Free |
| **Total** | | **~$13.58/month** |
This is a lean setup with no Redis, no staging environment, and no backup replica. A more complete production setup runs $20 to $40 per month. Add a Redis instance and a Postgres replica and you are over $30 before accounting for any traffic overage.
## The Hidden Costs of Fly.io
### Every App Needs a Dockerfile and fly.toml
Fly.io does not auto-detect your stack. Every app requires a `fly.toml` configuration file and a Dockerfile. If your team is not comfortable with Docker, this is a real barrier before your first deploy. If you have multiple services, each needs its own config maintained separately.
### Fly Postgres Is Not Truly Managed
Unlike managed databases on platforms like Supabase or AWS RDS, Fly Postgres puts operational responsibility on you. Backups are not automatic unless you configure them. Version upgrades are manual. Failover requires a primary-replica setup you provision and manage yourself. Teams used to managed databases are often caught off guard by this.
### Cold Starts with Scale-to-Zero
Scale-to-zero sounds like a cost saving, but it trades cost for latency. The first request after a Machine stops takes several seconds to respond. For any user-facing API, this creates a noticeable degradation. Avoiding cold starts means keeping Machines running, which means paying the always-on rate.
### No Free Tier Since 2024
Fly.io was a popular choice for side projects precisely because of its generous free tier. When that was removed in 2024, every app that was free started incurring charges. Developers who built habits around Fly.io's free tier have had to revisit their hosting costs.
## Fly.io vs Kuberns: How They Compare
Fly.io pricing can be reasonable for small apps, but the operational overhead never disappears. Every app needs a Dockerfile and config file. Postgres requires manual management. Scaling requires manual rules. For teams that want production-grade reliability without that configuration layer, [Kuberns](https://kuberns.com) takes a fundamentally different approach.
Kuberns is an agentic AI deployment platform built on AWS. You connect your GitHub repository and click Deploy. The agentic AI detects your stack, builds your app, provisions infrastructure, handles SSL, and manages auto-scaling automatically. No Dockerfile required. No config files.
| Feature | Fly.io | Kuberns |
|---|---|---|
| Dockerfile required | Yes | No |
| Config file required | fly.toml | None |
| Free tier | None (removed 2024) | Yes, free credits |
| Auto-scaling | Manual rules via Machines API | Agentic AI, automatic |
| Cold starts | Yes (scale-to-zero) | No cold starts |
| Postgres | Self-managed Fly Postgres | External via env vars |
| CI/CD | Via fly.toml deploy hooks | Built-in, zero config |
| Monitoring | Basic Fly dashboard | Unified dashboard |
| Starting price | ~$2/month | $7/month |
| Built on | Fly Machines | AWS |
The core difference is not price. It is how much you have to configure and maintain. Fly.io gives you control. Kuberns gives you automation.
## Why Teams Switch from Fly.io to Kuberns
The teams that move off Fly.io typically hit one of three points. The Dockerfile and `fly.toml` maintenance becomes overhead as the number of services grows. Fly Postgres operational responsibility becomes a distraction from building. Or they need real auto-scaling without writing Machines API logic.
Kuberns removes all three of those friction points.
- Agentic AI deployment that reads your repository and configures the entire pipeline without any config file
- Zero cold starts so every request is served immediately regardless of traffic patterns
- AI-driven auto-scaling that responds to real traffic without manual rules
- Built-in CI/CD from day one, every Git push triggers a redeploy automatically
- Unified monitoring with logs, metrics, and alerts in a single dashboard
- Free credits on signup to deploy your first app at no cost
If you are currently on Fly.io and dealing with config overhead or Postgres management, migrating to Kuberns is straightforward. Connect your GitHub repo, add your environment variables, and your app is live on AWS with HTTPS and auto-scaling in under five minutes.
For a wider view of how Fly.io compares against other platforms, see our full comparison of [Fly.io vs AWS vs Kuberns](https://kuberns.com/blogs/flyio-vs-aws-vs-kuberns/) and [Fly.io vs Render vs Kuberns](https://kuberns.com/blogs/flyio-vs-render-vs-kuberns-ai/).
[Start deploying on Kuberns for free](https://dashboard.kuberns.com/)
## Frequently Asked Questions
### Is Fly.io free in 2026?
Fly.io removed their permanent free tier in 2024. New accounts receive $5 in trial credits. After that, all resources are billed. There is no permanently free tier for Machines, Postgres, or bandwidth in 2026.
### How much does Fly.io cost per month?
A minimal app (one shared-cpu-1x Machine, 256 MB RAM) costs around $1.94 per month always-on. A typical production setup with two API instances and a Postgres database runs $13 to $20 per month. Performance Machines with dedicated CPU start at $99 per month.
### What happened to the Fly.io free tier?
Fly.io eliminated permanent free resource allowances in 2024. The previous free tier included three Machines, a Postgres instance, and 160 GB of bandwidth per month at no cost. This was replaced with a $5 trial credit for new accounts.
### Does Fly.io charge for stopped Machines?
Fly.io does not charge CPU and RAM for stopped Machines. However, volumes attached to stopped Machines continue to be billed at $0.15 per GB per month.
### How does Fly.io Postgres pricing work?
Fly Postgres is billed as a standard Fly Machine plus a persistent volume. A minimal setup (shared-cpu-1x, 256 MB, 1 GB volume) costs around $2.09 per month always-on. Backups cost $0.02 per GB per month if enabled. You manage backups, upgrades, and failover yourself.
### What is a good alternative to Fly.io?
[Kuberns](https://kuberns.com) is a strong alternative. It is an agentic AI deployment platform built on AWS that deploys any stack from GitHub without a Dockerfile or config files. New accounts get free credits to start, with no cold starts, auto-scaling included, and CI/CD built in.
### How does Fly.io pricing compare to Kuberns?
Fly.io charges per Machine-hour plus storage and bandwidth, and requires Dockerfiles and fly.toml for every app. Kuberns charges based on actual compute with no config files, includes agentic AI deployment, zero cold starts, and AI-driven auto-scaling from day one.
---
- [More Alternatives articles](https://kuberns.com/blogs/category/alternatives/1/)
- [All articles](https://kuberns.com/blogs/)