MongoDB Atlas Pricing, Explained (2026)

The Cluster That Cost Me Nothing, Then Everything

In 2023, while I was neck-deep in ClickHouse and Parquet files for Clickly, I took a detour. Everyone on Twitter was building AI side-projects on MongoDB, throwing unstructured JSON at it like confetti. I thought, fine, I’m a Postgres guy, but let me see what the fuss is about. I spun up a free cluster on MongoDB Atlas.

And here’s the thing: it was genuinely lovely. Free. Fast enough. No credit card. I built a little scratchpad app, threw documents at it, and never thought about a schema once. For about a week, I was a convert.

Then I tried to imagine what happened when I got real. I opened the pricing page. And the number that greeted me on the other side of that free tier was not $5. It was not $12. It was the kind of jump that made me close the tab and quietly go back to Neon. That gap — between “free” and “real” — is the whole story of Atlas pricing, and almost nobody explains it before you’re already committed. So let me.

---

The Free Ride: M0 and Its 512 MB Ceiling

Atlas gives you a genuinely free tier called M0, and credit where it’s due — it’s free forever, not a 12-month trial that mugs you on day 366 (looking at you, AWS).

Here’s what $0 buys you:

• 512 MB of storage. Half a gigabyte. Enough for a prototype, a demo, a class project.
• Shared compute. You’re on a multi-tenant box with strangers, so performance is a coin flip.
• No automated backups. If your data goes, it’s gone. Do not put anything you’d cry over on M0.

For learning MongoDB or standing up a weekend demo, M0 is fantastic. I have zero complaints about it. The complaint starts the moment your prototype gets a heartbeat and you outgrow 512 MB — because the next rung up is where things get interesting.

---

The Flex Tier: $8 to $30, and Where the Old Tiers Went

If you started with MongoDB a couple of years ago, you might remember the M2 and M5 “shared” clusters, or the “serverless” instances that promised to scale to zero. Forget both. As of February 2025, MongoDB folded all of that into one thing: the Flex tier. The old shared clusters and serverless instances hit end-of-life on January 22, 2026, and any accounts still on them got auto-migrated to Flex.

Flex is actually the sanest pricing MongoDB has ever shipped, and I’ll give them full marks for it:

• $8/month base fee, which includes 5 GB of storage and up to 100 operations per second.
• Usage-based on top of that, but — and this is the good part — capped at $30/month. No matter how spiky your traffic gets, Flex will never bill you more than thirty bucks. No surprise 3 AM invoice.
• Unlimited data transfer, and you get the grown-up features too: Atlas Search, Vector Search, triggers, change streams.

That $30 ceiling is the thing serverless never gave you. Serverless pricing on any platform is a slot machine — I’ve watched a single runaway function eat a bill alive. Flex removes the anxiety. If MongoDB were a religion for me, Flex is the tier I’d actually recommend. It’s cheap, it’s predictable, and 5 GB goes a long way for a young SaaS.

The problem isn’t Flex. The problem is what happens after Flex.

---

The Cliff: Dedicated Clusters Start Around $57

The moment you need a production-grade database — real backups, dedicated compute, no noisy neighbors, the ability to actually scale — you leave Flex and land on a dedicated cluster, starting at M10.

An M10 on AWS US East runs about $0.08 per hour. Sounds tiny. It’s not, because databases run 24/7, and that hourly rate compounds into roughly $57 a month for the compute alone. Your storage starts at 10 GB (expandable to 120 GB). And “for the compute alone” is doing a lot of work in that sentence, because backups and data transfer aren’t included — more on that in a second.

So the ladder looks like this:

Tier	What you get	Cost (2026)
M0	512 MB, shared, no backups	$0 forever
Flex	5 GB, capped, usage-based	$8–$30/mo
M10 (dedicated)	10 GB, 2 GB RAM, real backups available	~$57/mo + add-ons
M700 (top end)	Enterprise beast	~$33/hr (~$24k/mo)

Look at the middle of that table. You go from a hard ceiling of $30/month on Flex straight to a ~$57/month floor on M10 — and that floor only goes up once you turn on the things a real production database needs. There is no gentle $35 or $45 step in between. It’s a cliff. One month you’re paying $30 with change to spare, the next you’ve nearly doubled it just to get off the shared box.

That’s the number that made me close the tab in 2023. It’s not that $57 is outrageous for a serious company — it isn’t. It’s that as a broke solopreneur with $0 MRR, doubling my database bill overnight to serve zero paying customers was exactly the kind of pain I built this whole blog to help people avoid. (I added up every line of that “keeping the lights on” bill in what it really costs to run a SaaS.)

---

The Add-Ons Nobody Warns You About

Here’s where the sticker price and the real bill part ways. On a dedicated cluster, the cluster rate is the beginning of your invoice, not the end. Three line items quietly stack on top:

Backups. M0 and Flex don’t charge separately, but on dedicated clusters, cloud backups are billed per GB of snapshot storage. Skip them and you’re running production without a safety net. Enable them — as you should — and your bill grows with your data.

Data transfer. This is the classic cloud trap. Getting data out of your cluster (to your app, across regions, to the internet) costs money per GB on dedicated tiers. It’s the same egress-fee game that makes S3 and AWS bills mysterious, and it scales with your traffic in ways that are genuinely hard to predict up front.

Search nodes. Atlas Search and Vector Search are included as features, but if you want them to perform under real load, MongoDB will nudge you toward dedicated Search Nodes — separate infrastructure, separately billed. Suddenly your “one database” is two invoices.

None of this is a scam. It’s how managed infrastructure works. But it means the honest answer to “what does a production MongoDB Atlas cluster cost?” is “start at ~$57 and expect noticeably more.” Anyone quoting you the base rate is quoting you the trailer, not the movie.

---

The Broken Engineer’s Verdict: You Probably Want Postgres

You knew this was coming. If you’ve read anything else on this site, you know I’m a PostgreSQL-on-Neon zealot, and I’m not going to pretend otherwise here. But let me make the case honestly, because Atlas is a good product — it’s just rarely the right product for a bootstrapper.

The core reason founders reach for MongoDB is “my data is unstructured, I want to throw JSON at it and not think about schemas.” Here’s the punchline: Postgres does that too. The JSONB column type lets you store, index, and query documents inside a relational database. You get the flexibility of documents and real joins, real transactions, real constraints for the day your data actually needs structure — which, trust me, is sooner than you think. (I laid out the full head-to-head in MongoDB vs PostgreSQL, and the short version is: Postgres for ~95% of SaaS.)

Now line up the money. On Neon, I run seven Postgres databases on the free tier for $0, and its pay-as-you-go pricing separates storage from compute so a quiet database costs almost nothing — no $57 cliff, no doubling your bill to escape a shared box. When you do start paying, you scale up smoothly by the dollar instead of leaping off a ledge. (I compared the cheap serverless Postgres options — Neon, Supabase, and the ghost of Vercel Postgres — in this breakdown.)

So who should pay for Atlas? If you have a genuinely document-shaped workload at scale, a team that already knows Mongo cold, and revenue that makes $60–$100/month a rounding error — then Flex to prototype and a dedicated cluster to grow is a perfectly respectable path. There’s a reason MongoDB is a multi-billion-dollar company. It works.

But if you’re where I was — broke, ambitious, staring at a pricing page and doing the mental math on a product with no users — the answer is almost always the boring one. Prototype on M0 if you’re curious. Then go build on Postgres, where “unstructured” and “cheap” aren’t a contradiction. (And if you want to see just how deep the database rabbit hole goes, I documented my entire accidental database degree — ClickHouse, Timescale, Turso, all of it.)

---

The Bottom Line

MongoDB Atlas pricing isn’t dishonest — it’s just shaped like a cliff. M0 is a genuinely generous free tier. Flex, at a capped $30/month, is the most founder-friendly thing MongoDB has ever done. And then dedicated clusters start near $57 and climb from there once backups, data transfer, and search nodes join the party.

The free tier will happily let you fall in love. Just know that the honeymoon ends at a number, and the number is steeper than you’re bracing for. For most of us, the smarter move is to keep the document flexibility, drop the document database, and let Postgres and its JSONB column do the same job for a fraction of the anxiety.

Prototype where it’s free. Grow where it’s cheap. And never, ever double your database bill for a product that hasn’t earned its first rupee.

This is the Broken Engineer Guide — I over-engineer everything, fail at business, and share the scars so you spend your money on customers, not clusters.