The Free Tier Graveyard: every dev free tier that died, 2022-2026

Originally published on FreeTierWatch, where we hand-verify and track the free tiers of developer services.

Every entry below broke thousands of apps and invalidated thousands of tutorials. Each one is verified against the provider's own announcement — linked, dated, no folklore.

If you build on free tiers, this list is why this site exists.

The timeline of deaths

2022: Heroku — the extinction event

The free tier that taught a generation to deploy. For over a decade, git push heroku main was the first deployment command most developers ever ran.

Salesforce killed free dynos, free Postgres, and free Redis on November 28, 2022, citing "fraud and abuse." Free databases on personal accounts were deleted, not just paused. Inactive accounts were deleted too, starting a month earlier.

What it broke: countless student projects, portfolio apps, Discord bots, and the Nightscout diabetes-monitoring community (big enough that Heroku's own FAQ has a dedicated entry for it).

The lesson that everyone learned the hard way: "free forever" is a pricing decision, and pricing decisions change with ownership. Heroku's free tier died 12 years in — longevity is not a guarantee.

2023: Railway — the quiet conversion

Railway replaced its recurring free tier with a one-time $5 trial credit. Once spent, you pay the $5/month Hobby minimum.

This one still catches people in 2026 because old tutorials never got updated. Our Railway data page exists largely to answer "wait, I thought Railway was free?"

The lesson: a trial credit dressed as a free tier is the most common soft-kill. Check whether the "free" is recurring or one-time before you build.

2024: PlanetScale — profitability, announced plainly

PlanetScale's CEO published "PlanetScale forever" in March 2024: the company was prioritizing profitability, laying off sales/marketing, and retiring the Hobby plan — with one month's notice.

To their credit, it's the most honest free-tier death notice ever written: "We do not need to give away endless amounts of free resources in order to keep growing."

What it broke: thousands of side projects running on free MySQL, with 33 days to migrate.

The lesson: VC-funded generosity is a customer-acquisition expense. When the market turns, it's the first line item cut. The most generous tier on the market is often the most fragile.

2024: Fly.io — the silent sunset

No dramatic blog post here. Fly.io's famous "3 free shared VMs" simply stopped applying to new organizations around October 2024; new signups are pay-as-you-go with a card required.

This is the hardest kind of death to catch: nothing was announced loudly, existing users kept their allowances, and thousands of tutorials continue to describe a free tier that new users can't get. Our Fly.io page tracks the current reality.

The lesson: grandfathering hides the change. The free tier a tutorial author used in 2023 may not exist for you today, even though their app still runs.

2025: Glitch — total shutdown

Glitch didn't shrink its free tier — it stopped hosting apps altogether on July 8, 2025. Projects, profiles, the in-browser editor as a host: gone. Subdomain redirects run through end of 2026 as a courtesy.

What it broke: a decade of creative web experiments, classroom projects, and the lowest-friction "remix and run" culture the web had.

The lesson: sometimes the risk isn't the tier — it's the company. Glitch was beloved and still died. Affection is not a business model.

2025: Netlify — the stealth nerf

Netlify didn't kill anything. It just moved all new accounts to a credit-based plan (300 credits/month) where bandwidth costs 20 credits/GB — an effective ceiling around 15 GB/month, versus the legacy tier's 100 GB. Existing accounts were grandfathered, so most writeups never noticed.

Full numbers on our Netlify page.

The lesson: watch for repricing disguised as "simplification." Credits, units, and tokens usually mean the old number got smaller.

The pattern (and how to predict the next death)

Across all seven deaths, the same warning signs repeat:

1. Ownership change — Heroku died under Salesforce. Koyeb just joined Mistral AI; Turso is joining Supabase. Acquisitions reset pricing philosophy. We flag these on our data pages the day we see them.
2. "Fraud and abuse" language in blog posts — the standard prelude to free-tier removal.
3. Profitability pivots — layoffs + "sustainable business" posts preceded the PlanetScale kill.
4. A new paid tier appears just below the free tier's capability — the free tier is now cannibalizing revenue; its days are numbered.
5. Grandfathering begins — "existing users keep X" means X is already dead; it just hasn't reached you yet.

If you're building on a free tier today

• Export paths beat free capacity. Prefer services where leaving is one pg_dump away. The Heroku deletion hurt people who had no backup habit.
• Treat card-required free tiers as short-term loans. When the tier dies, the card is already on file.
• Keep your data layer portable — standard Postgres/MySQL/SQLite, no proprietary APIs you can't re-implement in a weekend.
• The boring alternative is immune. A ~$5/month VPS running Docker has no free tier to lose. When you outgrow free, that's the upgrade that can't be rug-pulled.

Every service above has a tracked data page with dated verification. When the next one dies — and the pattern says one will — subscribers hear it first.