The best Heroku alternative depends on how much operations work you want to own. Render is the closest like-for-like platform, Fly.io suits apps that need to run near users in several regions, Railway is the quickest for small teams, Hetzner is cheapest if you run servers yourself, and AWS fits once compliance, scale or enterprise customers matter.
Prices and features below were checked on each provider's official pages in November 2026. They change often, so treat them as a starting point and confirm before you commit.
What happened to Heroku in 2026
On 6 February 2026, Heroku announced it was moving to a sustaining engineering model focused on stability, security, reliability and support rather than new features. It did not announce a shutdown.
The specifics, from Heroku's own post:
- Heroku describes itself as "an actively supported, production-ready platform". Existing apps, pipelines and add-ons keep running.
- Customers paying by card in the dashboard, new and existing, see no change to pricing, billing or service.
- Enterprise Account contracts are no longer offered to new customers. Existing Enterprise contracts are honoured and can renew.
- Salesforce is moving its product and engineering investment towards enterprise AI.
So this is not an emergency. Your app will not be switched off next month. It is a planning signal: the platform you build on will not gain features, new enterprise procurement is closed, and Salesforce has told you where its attention is. If your roadmap needs something Heroku does not do today (a specific EU region, private networking with your other systems, a compliance programme your customers will audit), it is unlikely to arrive.
One related change on AWS: App Runner, the AWS service that looked most like Heroku, is closed to new customers. AWS now points new users to ECS Express Mode instead, which matters if you were planning on App Runner as your landing spot.
The alternatives compared
All five remove the "Heroku is frozen" risk; they differ in how much of the platform you end up running yourself. This table compares them on the points that usually decide it for a European team.
| Pricing model | Managed Postgres | EU regions | Scaling | Lock-in | Ops effort | Fits | |
|---|---|---|---|---|---|---|---|
| Heroku (staying) | Fixed price per dyno | Yes, from Essential to Premium | Common Runtime "eu"; Dublin and Frankfurt in Private Spaces | Add dynos; autoscaling on higher tiers | Medium: buildpacks, add-ons | Lowest | Stable apps that need nothing new |
| Render | Workspace plan plus fixed price per instance | Yes; HA on plans with 1 CPU or more | Frankfurt | Manual or autoscaling (Pro plan) | Low to medium: Docker or native runtimes, a blueprint file | Low | The closest Heroku replacement |
| Fly.io | Usage-based per Machine, priced per region | Yes (Managed Postgres, HA included), in fewer regions | Amsterdam, Frankfurt, Paris, Stockholm (and London) | Machines in many regions, auto stop and start | Low: Docker images, but fly.toml and its networking | Medium | Latency-sensitive or multi-region apps |
| Railway | Plan fee plus per-second usage | Partly: Railway-maintained container, single node by default, HA and backups as upgrades | Amsterdam | Replicas, vertical limits per plan | Low to medium | Low | Small teams, prototypes, internal tools |
| Hetzner | Fixed monthly per server | No; you run Postgres yourself | Germany, Finland | You build it | Lowest | Highest | Teams with real ops skills and tight budgets |
| AWS (ECS Express Mode, ECS, EKS) | Pay per resource; Express Mode itself is free | Yes: RDS and Aurora, Multi-AZ, point-in-time recovery | Frankfurt, Ireland, Paris, Milan, Stockholm, Spain and more | Anything you need | Medium, though containers stay portable | Medium to high | Compliance, scale, enterprise customers |
Sources: Heroku regions, Render regions and high availability, Fly.io regions and Managed Postgres, Railway regions and Postgres, Hetzner Cloud, ECS Express Mode.
What a small app costs on each
Entry prices are close; the gap opens up with managed databases, team features and bandwidth. These are list prices for the smallest sensible production setup, before tax and traffic.
| Platform | App instance | Postgres | Platform fee |
|---|---|---|---|
| Heroku | Basic $7, Standard-1X $25 per dyno per month | Essential-0 $5, Standard-0 $50; high availability starts with Premium-0 at $200 | None |
| Render | Starter $7 (0.5 CPU, 512 MB), Standard $25 (1 CPU, 2 GB) | From $6 plus $0.30 per GB of storage | Pro workspace $25 a month flat, unlimited members |
| Fly.io | shared-cpu-1x with 1 GB about $6.70 a month in US East (varies by region) | Managed Postgres from $38 plus $0.28 per GB | Optional support from $29 a month |
| Railway | Usage: about $20 per vCPU and $10 per GB of RAM per month | Billed as usage of the database container | Hobby $5 or Pro $20 per workspace, each including that amount of usage |
| Hetzner | Shared vCPU servers from €6.49 a month excluding VAT | Runs on your own server | None |
| AWS | Fargate per vCPU-hour and GB-hour, plus a load balancer | RDS per instance-hour plus storage | None; Express Mode has no extra charge |
Sources: Heroku pricing, Render workspace plans (changed April 2026) and compute plans, Fly.io pricing, Railway pricing, Hetzner Cloud. Hetzner raised cloud prices twice in 2026, on 1 April for all servers and on 15 June for new orders and rescales, so older blog posts quoting €4 servers are out of date. Some of its cheapest plan families were also listed as unavailable when we checked.
For AWS the honest answer is "it depends on the shape". A single small service on Fargate behind a load balancer often costs more than Render at the bottom end, because the load balancer and NAT are fixed costs. AWS gets cheaper relative to PaaS as you grow, especially with Savings Plans. Our NAT Gateway cost calculator shows one of the line items people forget.
How to choose
Decide by what you need in the next two years, not by the cheapest starter plan. In our experience these rules of thumb hold up:
- You want Heroku, just not frozen: Render. Same mental model (services, workers, cron, managed Postgres, preview environments), a Frankfurt region, and the least retraining.
- Your users are spread across continents, or latency matters: Fly.io. Running the same app in several regions is its core feature. Check that Managed Postgres is offered where you need it; today in Europe that is Amsterdam, Frankfurt and London.
- You are two or three developers on an early product: Railway or Render. Railway is faster to start; turn on Postgres HA and backups before you have real customer data.
- Cost is the main constraint and someone on the team likes running servers: Hetzner with a deployment tool on top. You take on patching, backups, Postgres and on-call. Our Hetzner vs AWS comparison covers the trade-offs.
- Customers send security questionnaires, you need a specific EU region, or you are heading for SOC 2, ISO 27001 or NIS2: AWS.
- None of the above, and the app is stable: staying on Heroku is a legitimate choice. Revisit it every six months.
What makes any move off Heroku harder than it looks
The app is the easy part; the database and the add-ons are where moves slip. Plan for these whichever target you choose.
- Postgres. Classic Heroku Postgres (Essential, Standard, Premium) does not give you the superuser or replication roles that tools such as AWS DMS need, so a near-zero-downtime move by replication is not available. You plan a dump-and-restore window, sized by testing it on a copy. Heroku Postgres Advanced does offer managed logical replication to an external database, which changes the plan if you are on it. Our guide to zero-downtime Postgres migration to RDS covers both paths.
- Build. Moving from buildpacks to a Dockerfile (or building with Cloud Native Buildpacks) is where hidden assumptions show up: system packages, asset builds, the release phase.
- Add-ons. List every one: Redis, scheduler, logging, email, search, monitoring. Each needs a replacement and a data move.
- Config and secrets. Config vars become environment variables or a secrets store. Rotate anything that was shared widely while you are at it.
- DNS and TLS. Lower TTLs a few days before the cutover, and check that certificates are issued on the new platform before you switch.
When a full AWS move makes sense
Move to AWS when the platform has become the constraint on selling, not just a line on the bill. The usual triggers:
- Customers or regulators ask hard questions. Where exactly is the data, who can access it, show me the audit logs and the restore test. On AWS you control all of it; on a PaaS you inherit someone else's answers.
- You need a specific region or private networking. A Paris or Frankfurt region, VPC peering with a partner, private connectivity to a customer.
- You have many services and environments. Past a handful of services, per-instance PaaS pricing and per-environment duplication add up, and Kubernetes or ECS starts to pay for itself.
- You need the wider platform. Managed queues, vector search in Postgres, data pipelines, AI services, all in the same account and network.
For a small number of services, start with ECS on Fargate (Express Mode makes the first deploy close to Heroku-simple); see ECS vs EKS for startups for when Kubernetes is worth it. Our step-by-step Heroku to AWS migration guide covers the concept mapping, pipeline, database and cutover.
For a real example, we moved a cybersecurity certification SaaS from Heroku to EKS in Paris: every service containerised and migrated one environment at a time, deploys reaching staging about ten minutes after merge, daily backups across twelve environments, and non-production compute cut by a third. That client sold to security-conscious buyers, which is exactly when the extra work pays off.
Where to start
Write down your services, add-ons, database size and the regions and compliance answers your customers expect. That list usually makes the choice obvious. Our free cloud readiness scorecard is a quick way to see what a move would involve, and the cloud migration checklist covers the rest.
If the answer is AWS and you would rather not spend a quarter learning it the hard way, our cloud migration service plans and runs the move with your team, database cutover included, and leaves you with infrastructure as code you own.
Frequently asked questions
Is Heroku shutting down in 2026?
No. In February 2026 Heroku moved to a sustaining engineering model focused on stability, security, reliability and support, with no new features. Existing apps keep running, card-paying customers saw no pricing changes, and existing Enterprise contracts can renew, but new customers can no longer sign Enterprise contracts. No shutdown date has been announced.
What is the closest alternative to Heroku?
Render is the closest like-for-like replacement: web services, workers, cron jobs, managed Postgres and preview environments, with a Frankfurt region and a flat $25 a month Pro workspace plan as of November 2026. Fly.io and Railway are also close but differ in pricing model, Postgres maturity and how much networking you manage.
Can I migrate Heroku Postgres with zero downtime?
Usually not on classic plans. Heroku Postgres Essential, Standard and Premium do not grant the superuser or replication roles that replication tools need, so most teams plan a tested dump-and-restore window. Heroku Postgres Advanced supports managed logical replication to an external database, which allows a much shorter cutover.
Is AWS App Runner a good Heroku replacement?
Not for new customers: AWS closed App Runner to new customers in 2026 and recommends Amazon ECS Express Mode instead. Express Mode deploys a container image to Fargate with a load balancer, autoscaling and HTTPS from a single call, at no extra charge beyond the underlying resources.