Deploying Your Side Projects Without Going Broke in 2026


I’ve got six side projects deployed right now. Two are barely used and just running for portfolio purposes. Two get a few hundred users a month. One has about 4,000 monthly active users. The last is a project I run with a friend that has small but recurring revenue.

Total monthly hosting cost across all six: $47 USD.

This wasn’t always the case. A few years back I was paying probably four times that for less functionality and considerably more stress about overage charges. The hosting landscape has changed quite a bit for small-scale stuff, and not all the changes are advertised loudly. Here’s what I’m actually using and how I’d pick differently if I were starting over.

The Honest Tier Breakdown

Side project hosting decisions are about matching the project’s actual traffic and complexity to the right product. Most of the bad decisions come from over-engineering for hypothetical scale.

The categories I think about:

Static or nearly-static sites: portfolio, blog, marketing pages, simple JavaScript apps that talk to external APIs. These should cost zero or nearly zero.

Small fullstack apps with their own backend: things like a side project SaaS, a tool with user accounts, an app with a database. These can cost $5-20 depending on choices.

Apps with real user load (low thousands of monthly active users and up): the cost question gets more interesting and the choices matter more.

I’ll work through each category.

Static and Nearly-Static

For my portfolio and blog, I use Cloudflare Pages. It’s free, fast, and the build minutes allowance is more than enough for anything personal. The custom domain setup is straightforward and includes free SSL.

Vercel and Netlify are the obvious alternatives. Both have generous free tiers but the bandwidth limits are tighter than Cloudflare’s. I’ve had projects exceed Vercel’s free bandwidth allowance unexpectedly when a post got some attention; I’ve never had that happen on Cloudflare.

The catch with Cloudflare Pages is that the build environment is less flexible than Vercel’s. If you’re doing anything unusual with your build process, Vercel’s environment will probably accommodate it more easily. For straightforward static site generators (Astro, Hugo, plain HTML), Cloudflare is fine.

GitHub Pages is the option people forget about. For purely static content with no build complexity, it’s free and works well. I use it for one project just because.

Small Fullstack Apps

This is where the most has changed in the last two years. The options I’d seriously consider:

Fly.io is what I use for most fullstack side projects. The free tier got reduced last year and now requires a credit card, but the paid tier starts at $5 per app per month for the smallest VM size, which is enough for projects with low-thousands of users. Fly handles deployment well, supports most languages and frameworks, and the developer experience is genuinely pleasant once you get past the initial CLI setup.

Railway is the slightly more polished alternative. Pricing starts at $5 per month for the developer tier, which includes $5 of usage credits. For projects that need a managed database, Railway’s Postgres integration is simpler than rolling your own on Fly. The trade-off is that costs can creep up if you don’t pay attention.

Render has a free tier that includes a small Postgres database. The catch is that the free web services sleep after 15 minutes of inactivity, which means cold starts of 30-60 seconds on next request. For projects that are genuinely low-traffic, this is acceptable; for anything you want to use yourself regularly, it’s annoying.

Hetzner Cloud is the option I recommend to friends who are comfortable with a bit more sysadmin work. €4.51 per month gets you a CX22 instance with 4GB of RAM and 40GB of storage, which is dramatically more compute than the PaaS options offer at the same price point. The cost is that you’re managing the OS, the database, backups, and so on. For someone learning sysadmin, this is a feature; for someone who just wants to deploy stuff, it’s a chore.

DigitalOcean has fallen behind on pricing relative to Hetzner but is still competitive and has a polished interface. Their App Platform is a reasonable PaaS option starting at $5/month per service.

Apps With Real Traffic

When projects start getting actual usage, the cost equation shifts in ways that aren’t always intuitive.

For my project with 4,000 MAU, total hosting cost is $19 per month. That covers a small Fly app, a managed Postgres instance, and Cloudflare R2 for storage. R2 deserves its own mention—it’s S3-compatible object storage with no egress fees, which is a meaningful cost savings if your app serves much media.

If I were starting that same project from scratch today, I might consider Coolify on a Hetzner box. It’s an open-source PaaS that gives you a Heroku-like experience on hardware you control. For a $5-7 per month Hetzner instance, you can run multiple projects.

The general principle is that as traffic grows, the cost of bandwidth and database operations matters more than the cost of compute. Picking hosts with reasonable bandwidth pricing (most of the new generation) over hosts with old-school bandwidth pricing (AWS, GCP at default settings) makes a much larger difference than picking the cheapest compute.

What I’d Actually Avoid

Some categories I’d avoid for side projects, with the caveat that they make sense for production work at companies:

The big three clouds (AWS, GCP, Azure) for side projects. The free tiers are time-limited, the pricing is complex, and the chance of an unexpected $200 bill from misconfigured services is high. I’ve been there. The educational value of learning AWS for a side project is real but consider whether that’s actually what you want to optimise for.

App-store-aligned platforms like Heroku have improved but still aren’t priced for hobby use after their free tier ended. The convenience hasn’t kept up with the alternatives.

Anything with autoscaling enabled by default. Side projects rarely need autoscaling. The cost downside (an unexpected traffic spike billed at autoscale rates) outweighs the upside (your project handling viral traffic gracefully). If you actually go viral, deal with it then.

My Practical Setup Right Now

For anyone wanting a starting point, here’s what I’d actually recommend a self-taught dev set up in 2026:

Cloudflare account for DNS, R2 storage, and Pages hosting. The account is free and almost everything you’ll need fits in the free tier.

Fly.io account for any fullstack apps that need a server. The $5/month minimum is reasonable and the platform is small enough to fit in your head.

A Hetzner or Linode account if you want a long-running personal server for experiments. €5-10 per month.

GitHub for code, GitHub Actions for CI (the free tier is plenty).

Total monthly cost for the setup: about $5-15 depending on how many active projects you have running. You can deploy nearly anything in this stack with reasonable performance, you’ll learn skills that transfer to professional work, and you won’t get surprised by bills.

The biggest hosting lesson I’ve learned: optimise for not getting surprised, not for hypothetical scale. Most side projects don’t need to be ready for millions of users. They need to be cheap enough that you don’t kill them when life gets busy. That’s the metric that’s kept my six projects alive across years where my actual time available to maintain them has varied wildly. More info on the platforms I’ve touched is on team400.ai if you’re curious what some of these patterns look like at slightly larger scale.