Justin McKelvey

Justin McKelvey

Ran snowboard shops. Taught myself to code. Now I build AI that does real work.

• Vibe Code Rescue • 8 min read •

Neon Pricing (2026): Free Plan Limits, $0.106 per CU-Hour, and When Scale-to-Zero Stops Saving You Money

Quick Answer

Neon has a free plan (100 compute hours and 1 GB per project) and two usage-based paid plans: Launch at $0.106 per CU-hour and Scale at $0.222, both with storage at $0.35 per GB-month. A small database that sleeps when nobody's using it runs about $15 a month on Launch. The same database left on all month is about $21. A busy one is about $80 on Launch and $163 on Scale.

Prices from neon.com/pricing, read November 20, 2026 · Author: Justin McKelvey, AI consultant, former CTO at an AI startup, 50+ products shipped

TL;DR: Neon is cheap right up until your database stops sleeping

Neon's whole pricing model is a bet that your database is idle most of the time. It scales to zero after 5 minutes without queries and wakes up in a few hundred milliseconds when one arrives, and you only pay for the hours it's awake. For a side project, an internal tool or an app with business-hours traffic, that bet pays off.

The bill changes when something keeps the database awake: a health check every minute, a background job that polls, a busy app. Then you're paying for every hour of the month, and the question becomes how big the compute is and which plan you're on.

Neon plans and prices (as of November 2026)

Free Launch Scale
Price $0, no card Usage-based (Neon's "typical spend" $15/mo) Usage-based (typical spend $701/mo)
Compute 100 CU-hours per project per month $0.106 per CU-hour $0.222 per CU-hour
Largest compute 2 CU (8 GB RAM) 16 CU (64 GB RAM) 56 CU (224 GB RAM)
Storage 1 GB per project (20 GB total) $0.35 per GB-month $0.35 per GB-month
Restore history Up to 6 hours or 1 GB of changes Up to 7 days, $0.20 per GB-month of changes Up to 30 days, same rate
Branches per project 10 10 included, then $0.002 per branch-hour 25 included, then $0.002 per branch-hour
Scale to zero After 5 minutes, always on After 5 minutes, can be turned off Configurable
Network transfer 5 GB per project 500 GB per project, then $0.10/GB 500 GB per project, then $0.10/GB
Extras Community support Billing support, 3-day logs SLAs, HIPAA, private networking

A CU (compute unit) is about 4 GB of memory plus CPU, and the smallest size is a quarter of one. Your compute bill is the average size times the hours it runs times the plan's rate. Every branch has its own compute, and they all count.

What a Neon database actually costs: four real bills

Database Compute Storage and history Monthly bill
A. Side project or internal tool, sleeps when idle (Neon's "intermittent" example) 140 CU-hours x $0.106 = $14.84 1 GB = $0.35 $15.19 on Launch, or $0 on Free if it stays under 100 CU-hours
B. Small app that never sleeps, smallest size (0.25 CU, scale to zero off) 182.5 CU-hours x $0.106 = $19.35 5 GB = $1.75, plus about $0.20 of history $21.30 on Launch
C. Busy production app, scaling between 0.5 and 4 CU (Neon's "medium" example) 720 CU-hours x $0.106 = $76.32 10 GB = $3.50 $79.82 on Launch, $163.34 on Scale
D. A preview branch a developer uses 2 hours a workday 11 CU-hours x $0.106 = $1.17 Shares the parent's storage until data changes about $1.17 extra (inside the 10 included branches)

The math under rows A and C uses Neon's own workload examples; the dollar figures are my arithmetic on today's rates. The jump that matters is A to B. The smallest database costs $15 when it sleeps and $21 when it doesn't, and if it could sleep half the day (12 hours at 0.25 CU, 90 CU-hours) it would be about $9.54 plus storage. Find what keeps your database awake before you pick a plan.

The Free plan's day-17 problem

Neon's Free plan is generous on paper: 100 projects, 100 CU-hours each, no credit card, no expiry. But scale to zero can't be turned off on Free, and the allowance is per month. If something keeps a Free database busy around the clock at the smallest size, 100 CU-hours lasts 400 hours, about 16 and a half days. Neon then suspends the compute until the next billing month. Your data stays; your app stops reaching it.

The second limit people miss is transfer. A Free project gets 5 GB of outbound data a month, shared across the database, file storage and functions, and hitting it also suspends the compute. An app that returns big query results or serves files out of Neon's object storage can hit that before it hits the compute limit.

So Free is right for prototypes, demos and anything that's quiet most of the day. The moment real users depend on it, Launch at roughly $15 to $21 a month is the cheap insurance.

Launch or Scale?

Scale costs $0.222 per CU-hour against Launch's $0.106, about 2.1 times as much for the same compute. It doesn't make your database faster at the same size; it buys you bigger maximum sizes (up to 56 CU), 25 branches, 30 days of restore history, an uptime SLA, HIPAA and private networking. If you don't need one of those, stay on Launch. If you're handling health data or a customer contract requires an SLA, Scale is the plan, and row C above shows what that does to the bill.

Neon vs Supabase, DigitalOcean and Railway for one small database

Where the database runs Small always-on database What else the price includes
Neon Launch (0.25 CU, about 1 GB RAM) about $21.30 Branching, restore history, scale to zero if you allow it
Supabase Pro $25 Auth for 100,000 users, file storage, an always-on 1 GB database
DigitalOcean managed PostgreSQL (1 GiB) $15.15 A managed database that's always on; no scale to zero, no branching

Neon wins when the database can sleep, and when you want a fresh copy of production data for every pull request. Supabase wins when you also need login and file storage, because its $25 covers all three; the breakdown is on Supabase pricing, and the bigger backend decision is on Supabase vs Firebase. A managed database on DigitalOcean is the cheapest always-on option, priced next to Droplets on DigitalOcean pricing. And if your app already lives on Railway, its built-in Postgres runs in the same bill; the whole app-plus-database math is on Railway pricing.

For what it's worth, this site doesn't use Postgres at all. It runs on SQLite on a Railway volume, which costs nothing extra and is plenty for one business's CRM and blog. Plenty of small apps are paying for a database server they don't need yet.

What to watch out for

  • Things that keep the database awake. Uptime monitors, connection-pool health checks, cron jobs and ORMs that ping on a timer all stop scale to zero. Check Neon's compute graph for a flat line before you assume you're on the cheap side of the table.
  • Autoscaling raises the ceiling and the bill. A database allowed to scale to 4 CU bills 4 CU while it's there. Set a sensible maximum size; Launch allows up to 16 CU.
  • Every branch is a compute. Preview branches are cheap when they sleep, but a branch someone left a script running against bills like production.
  • History isn't free on paid plans. Restore history costs $0.20 per GB-month of data that changes inside the window. A write-heavy app with a 7-day window pays for it.
  • The Free plan suspends, it doesn't slow down. Past 100 CU-hours or 5 GB of transfer, the compute is off until next month. Don't put a paying customer on it.

If an app you built with an AI tool is running up a database bill you can't explain, or keeps going down for reasons nobody can find, that's what Vibe Code Rescue fixes.

Frequently asked questions about Neon pricing

How much does Neon cost per month?

Neon's Free plan is $0. The paid Launch plan has no fixed monthly fee: you pay $0.106 per CU-hour of compute and $0.35 per GB-month of storage, and Neon lists a typical Launch bill as $15 a month. The Scale plan charges $0.222 per CU-hour and lists a typical bill of $701. Prices from neon.com/pricing, read November 20, 2026.

What is a CU-hour on Neon?

A CU (compute unit) is about 4 GB of memory plus CPU, and a CU-hour is one CU running for one hour. Your monthly CU-hours are the average compute size times the hours it runs. A database at the smallest size (0.25 CU) running all month uses about 182.5 CU-hours, which is $19.35 on Launch.

What are the Neon Free plan limits?

Each project gets 100 CU-hours of compute, 1 GB of storage and 5 GB of network transfer a month, with up to 100 projects, 10 branches per project and 20 GB of storage across all projects. If a project uses up its CU-hours or transfer, Neon suspends its compute until the next month. Your data isn't deleted.

Can I keep a Neon Free database running 24/7?

Not for a whole month. Free plan databases scale to zero after 5 minutes idle and you can't turn that off. If something keeps the database busy around the clock at the smallest size (0.25 CU), 100 CU-hours lasts about 400 hours, so the compute is suspended around day 17.

Is Neon cheaper than Supabase?

For the database alone, usually yes: a small always-on Neon database works out to about $21 a month on Launch, and one that sleeps at night closer to $10 to $15, against $25 for Supabase Pro. But Supabase's $25 also includes auth, file storage and an always-on database. Compare the whole backend you need, not just Postgres.

When should I pay for Neon's Scale plan?

When you need what it adds beyond compute: an uptime SLA, HIPAA, private networking, 25 branches per project and up to 30 days of restore history. It charges $0.222 per CU-hour, about 2.1 times Launch, so a busy database costs roughly twice as much on Scale. Buy it for compliance, not speed.

Sources

  • Neon pricing, read November 20, 2026 (page last updated November 4, 2026): plans, CU-hour and storage rates, branch, history, transfer and snapshot prices, the Free plan FAQ and the workload examples.
  • Neon plans documentation, read November 20, 2026: Free plan transfer allowance, history window defaults, and how branch computes count toward usage.

Next step Get the free repo audit →

Free Resource Justin McKelvey

What your AI stack actually costs

Prices changed 3x this year. The always-current cost sheet: sticker price vs what heavy use actually costs for Cursor, Claude, Replit, Lovable, Bolt & more.

Frequently Asked Questions

How much does Neon cost per month?
Neon's Free plan is $0. The paid Launch plan has no fixed monthly fee: you pay $0.106 per CU-hour of compute and $0.35 per GB-month of storage, and Neon lists a typical Launch bill as $15 a month. The Scale plan charges $0.222 per CU-hour and lists a typical bill of $701. Prices from neon.com/pricing, read November 20, 2026.
What is a CU-hour on Neon?
A CU (compute unit) is about 4 GB of memory plus CPU, and a CU-hour is one CU running for one hour. Your monthly CU-hours are the average compute size times the hours it runs. A database at the smallest size (0.25 CU) running all month uses about 182.5 CU-hours, which is $19.35 on Launch.
What are the Neon Free plan limits?
Each project gets 100 CU-hours of compute, 1 GB of storage and 5 GB of network transfer a month, with up to 100 projects, 10 branches per project and 20 GB of storage across all projects. If a project uses up its CU-hours or transfer, Neon suspends its compute until the next month. Your data isn't deleted.
Can I keep a Neon Free database running 24/7?
Not for a whole month. Free plan databases scale to zero after 5 minutes idle and you can't turn that off. If something keeps the database busy around the clock at the smallest size (0.25 CU), 100 CU-hours lasts about 400 hours, so the compute is suspended around day 17.
Is Neon cheaper than Supabase?
For the database alone, usually yes: a small always-on Neon database works out to about $21 a month on Launch, and one that sleeps at night closer to $10 to $15, against $25 for Supabase Pro. But Supabase's $25 also includes auth, file storage and an always-on database. Compare the whole backend you need, not just Postgres.
When should I pay for Neon's Scale plan?
When you need what it adds beyond compute: an uptime SLA, HIPAA, private networking, 25 branches per project and up to 30 days of restore history. It charges $0.222 per CU-hour, about 2.1 times Launch, so a busy database costs roughly twice as much on Scale. Buy it for compliance, not speed.
Justin McKelvey, Austin, TX

Written by

Justin McKelvey

I spent 12 years running snowboard and mountain-bike shops in Colorado, then taught myself to code. Since then: 50+ products shipped, Head of Product at a national sports platform, CTO at an AI startup. Now I put AI to work in real businesses in Austin, TX and remote, and write down what actually works.

Work with me

Want more help? Get a free 20-min repo audit · Grab the real AI stack cost sheet