Heroku is finally dead. The original Platform as a Service company that made it much easier (for a cost) to deploy web applications made an announcement yesterday, An Update on Heroku, confirming what has long been suspected: Salesforce (the owners of Heroku since 2010) is officially pulling the plug.

Heroku is Dead Image

Frustratingly, the announcement doesn’t just say that. Instead, it notes moving to a new “sustaining engineering model” while simultaneously no longer accepting any new enterprise customers. That’s corporate-speak for: we are keeping the lights on and will keep accepting your money, but the product will be run by an absolute skeleton crew of engineers and support until it finally dies a slow death.

It is time to migrate elsewhere.

For those of us who have been using Heroku and Django for a while, this feels like déjà vu all over again. Despite the heroic work of a dwindling number of staffers, Heroku hasn’t had any meaningful product innovations in over a decade, and the platform problems have noticeably accelerated in recent years: security incidents in 2022, ending free plans in 2024, and in June last year suffering a widespread outage lasting over 15 hours followed by another lasting over 8 hours. If you look at the Heroku Status archive ,it is littered with similar unplanned downtime, all of which speaks to a general lack of investment and focus.

So what next? Fortunately, there are now a number of next-generation hosting solutions, most notably Render, Fly.io, Railway, Digital Ocean App Platform, Platform.sh, and PythonAnywhere. There is also a growing list of exciting self-hosted options, such as Coolify, Dokku, and Appliku, that promise a PaaS-level experience on a VPS you control.

And I have to mention the heroic efforts from Eric Matthes, author of Python Crash Course, around django-simple-deploy, a third-party package providing a common interface for Django deployments across several providers, currently Fly.io, Upsun, and Heroku, though it has also supported others in the past.

VPS + PaaS

If you are a solo developer, a consultancy, or a small-to-medium-sized web business, I think Platforms as a Service can still make sense. You are paying someone else to manage your infrastructure, ensure as close to zero downtime as possible, manage your database, and so on. There are often add-ons to add popular tools such as queues, email support, and more.

It’s about finding the right equilibrium for you and your team. Newer self-hosted options listed above offer the option of self-hosting on a VPS, meaning you purchase and own the underlying virtual server and use software to make it as close to a PaaS as possible. One wrinkle is that while you can set up automatic database backups, you don’t really have a managed database that’s guaranteed to be up-to-date, regularly backed up, and more. Popular options for this around Postgres include CrunchyData, Supabase, Neon, and the usual options from AWS.

It all becomes a spectrum balancing what you are willing to pay and how much performance you really need.

What’s Next?

Speaking for myself, I will finally give a proper try to the leading PaaS and VPS alternatives. I know readers will want a firm recommendation, so when I have one, I will share it and likely publish new tutorials. But my super-quick take is that if you are a solo or small developer, then a PaaS probably makes sense, but once you start managing multiple sites with actual traffic, the motivation to switch to a VPS option grows and grows, and it has never been easier with newer tools.