Heroku Is Dead: Railway vs Render vs Fly.io Cost Comparison 2026

Salesforce made it official on February 6, 2026. Heroku is transitioning to a "sustaining engineering model." That is corporate language for maintenance mode: no new features, no new enterprise contracts for new customers, and a team that has been reassigned to Agentforce. Your existing apps will keep running, and credit card customers can still sign up, but the platform has stopped evolving. For anyone planning infrastructure over the next two to three years, that is a problem.

What followed the announcement was predictable. Forums lit up. Migration threads multiplied. Developers who had been quietly deferring the decision now had to make it. The challenge is that when you evaluate five platforms at once, every comparison article reads slightly differently depending on which features the author cares about. DX, language support, managed databases, auto-scaling. There is a lot to cover. This article is not going to try to cover all of it. It focuses on one thing: what these platforms actually cost for a real, representative workload.

The Pricing Model Problem

Before you can compare numbers, you need to understand that these platforms do not even use the same pricing model. Render uses fixed monthly pricing: you pick an instance size and you pay that flat rate regardless of whether traffic is light or heavy. Railway uses usage-based pricing: you pay per GB-second of RAM and per vCPU-second of compute. Fly.io sits somewhere in between, charging per second of machine runtime based on your chosen CPU and RAM preset.

In practice, this makes comparisons tricky. Railway's usage-based model can be cheaper than Render for a side project that does not receive much traffic. But once that project runs 24/7 under consistent load, the math shifts.

Here is a concrete example. A Node.js API running continuously with 512 MB of RAM consumes roughly 0.5 GB of memory every second it is alive. Over a 30-day month, that is approximately 2.6 million GB-seconds. At Railway's published rate of $0.00000386 per GB-second, that RAM alone costs around $10 before you count any vCPU usage. The Hobby plan includes $5 in credits, so you would be looking at a bill somewhere in the $10 to $15 range once CPU is factored in. That is for a single service that never sleeps.

Render would charge you a flat $7 per month for a Starter instance with 512 MB RAM and 0.5 vCPU. Knowing which model is cheaper requires knowing your actual usage pattern, not just your instance size.

Side-by-Side Cost Table

Scenario: Solo developer, one always-on web service, 512 MB RAM, no database, approximately 50 GB outbound bandwidth per month.

A note on Vercel and Netlify: these two show up in a lot of "Heroku alternatives" lists and they should not. They are excellent platforms, but they are not designed to run persistent backend services. Vercel is the right call if you are deploying a Next.js application with serverless API routes. It is not the right call if you are running a traditional Express server or a Python API that needs to stay alive between requests.

For custom scenarios including databases, background workers, and higher traffic, use the InfraTally PaaS Cost Calculator to run your own numbers.

My actual recommendation, and the reasoning

Start with Railway if you are a solo developer or small team and you want something that just works. Connect your repo, push, done. The usage-based pricing is in your favor when traffic is low, which it always is at the start. The $5 Hobby plan includes $5 in resource credits, so a light side project often lands right at the base price. The complexity shows up later, when your app is running 24/7 under consistent load and you're trying to predict a variable monthly bill. That's the moment to revisit.

If you need to know what your bill will be before the month ends, use Render. It's the most Heroku-like option available right now: familiar workflow, solid documentation, fixed pricing. The Starter tier at $7/month for 512 MB RAM is competitive. Most production apps end up at the Standard tier at $25/month for 2 GB RAM, and that's a number you can budget around. There are no surprises in Render's billing model, which is exactly what you want when you're migrating off a platform that just told you it's going into maintenance mode.

Fly.io is the right answer for a specific situation: your application is latency-sensitive and your users are spread across geographies. Fly runs containers at the edge across 30+ global regions, which means your API can be close to users in Southeast Asia or Europe without separate CDN configuration. The per-second billing with automatic suspend when idle keeps costs reasonable. What it costs you is learning curve. Fly relies heavily on its CLI and the flyctl toolchain, and concepts like Machines, volumes, and regions require more upfront investment than Railway or Render. Worth it for the right workload. Overkill for most.

DigitalOcean App Platform gets underrated in these comparisons because DigitalOcean isn't exciting. That's actually the argument for it. They've been building cloud infrastructure longer than Railway has existed as a company. The App Platform starts at $5/month for a shared 512 MiB container, supports a git-push workflow that anyone who used Heroku will recognize, and gives you a clear upgrade path to Droplets, managed databases, and Kubernetes when you need more control. For a team migrating off Heroku who wants the fewest surprises and the most familiar workflow, this is worth a serious look.

Vercel and Netlify are not Heroku alternatives for traditional backends. They're excellent for what they do (Vercel for Next.js and serverless API routes, Netlify for static sites with edge functions) and wrong for a persistent Express or Python API that needs to stay alive between requests. They keep appearing in "Heroku alternatives" lists written by people who have never tried to run a real backend on them.

The Hidden Cost Nobody Mentions

Every platform in this comparison has an egress line item, and they are not all equal.

DigitalOcean App Platform charges $0.02 per GiB over the included allowance. Fly.io charges $0.02 per GB for outbound traffic in North America and Europe. Railway charges $0.05 per GB of egress. These are all manageable for typical backend APIs.

Vercel is the outlier that warrants a specific warning.

This is not a reason to avoid Vercel for its intended use case. It is a very good reason not to try to use Vercel as a general-purpose backend platform.

My Recommendation After 30 Years

I have watched a lot of platforms come and go. The ones that damage teams most are not the ones that shut down suddenly; they are the ones that stop evolving quietly while teams keep building on them. Heroku's February 2026 announcement is honest about the trajectory, and I respect that. It also means you should be planning your exit now, not when something breaks.

Every platform listed here has a free trial or low-cost entry point. My actual recommendation is to pick the one that fits your use case above, deploy your app, and run it for 30 days before committing. Real-world cost behavior is always more informative than any comparison table, including this one.

Pricing figures in this article were verified against published platform documentation in April 2026. Cloud pricing changes frequently. Verify current rates before making infrastructure decisions. This article contains affiliate links. See our Affiliate Disclosure.