Azure SQL Database Pricing, Decoded for a Broke Founder (2026)

The Day I Opened the Azure Pricing Calculator and Closed My Laptop

It was late 2024. I was helping a friend who runs a small B2B tool — the kind of guy who came up through a Windows shop, wrote everything in .NET, and genuinely likes Microsoft. Nothing wrong with that. He asked me a simple question: “How much will it cost to put my database on Azure?”

I opened the Azure SQL Database pricing page, expecting a number. Instead I got a decision tree. DTU or vCore? Provisioned or serverless? General Purpose or Business Critical? Do you have a SQL Server license with Software Assurance, or not? Locally-redundant or zone-redundant backups?

I’m an engineer. I like knobs. But this was the moment I remembered why I run Postgres on Neon and sleep fine at night. Let me decode the whole thing for you, because if you’re going to pay Microsoft, you should at least know what you’re paying for.

First, the one fact everyone glosses over: Azure SQL Database is managed SQL Server in the cloud. It’s not a generic database. It’s Microsoft’s own engine, and the license for that engine is baked into the price whether you notice it or not. Keep that in your back pocket — it explains everything.

---

Model One: DTU — The Bundle That Hides the Math

The older pricing model is DTU-based. A DTU (Database Transaction Unit) is Microsoft’s made-up blend of CPU, memory, and I/O rolled into a single dial. You don’t buy 2 vCPUs and 8 GB of RAM; you buy “50 DTUs” and trust the blender.

It comes in three tiers:

• Basic — 5 DTUs, 2 GB storage included, about $4.90/month. Fine for a toy. Useless for anything real.
• Standard — from S0 at 10 DTUs (~$15/month) up through S12. This is where most small production databases actually live.
• Premium — from P1 (125 DTUs, ~$456/month) all the way to P15 at 4,000 DTUs, which runs somewhere around $11,000/month.

The appeal of DTU is that it’s one number — compute and a slug of storage bundled together, billed hourly (roughly 730 hours a month). The problem with DTU is also that it’s one number. You have no idea whether you’re CPU-bound or I/O-bound, so when your app gets slow, you can’t reason about it. You just bump to the next tier and pay more. It’s a black box you feed money into.

Microsoft themselves now nudge new projects toward vCore. DTU still exists mostly for people who don’t want to migrate. My take: if you’re starting fresh on Azure in 2026, skip it.

---

Model Two: vCore — Real Hardware, Real Invoice

The vCore model is the grown-up option, and it’s what the docs now recommend. Here you buy actual, nameable resources: a number of vCores, an amount of memory, and storage priced separately. That transparency is genuinely better — you can see what you’re renting.

You can also see how much it costs.

There are two service tiers that matter:

• General Purpose — the default. In East US as of 2026, a 4-vCore General Purpose database runs about $1.008/hour for compute — call it ~$736/month — before storage. Storage is roughly $0.115/GB per month on top.
• Business Critical — faster local SSD, built-in high availability replicas. A 4-vCore Business Critical database is around $2.71/hour, or ~$1,978/month, with storage at about $0.25/GB per month.

Read those numbers again. Seven hundred and thirty-six dollars a month is the floor for a modest General Purpose database, and that’s compute alone. My whole SaaS stack — the one I broke down in the broke solopreneur’s survival guide — never came close to that even at its most bloated.

The License You’re Renting Without Realizing It

Here’s the part that stings. That vCore price includes a SQL Server license. You are renting Microsoft’s database engine by the hour, forever, on top of the hardware. It never becomes yours.

If your company already owns SQL Server licenses with Software Assurance — classic for a Windows-first enterprise — Azure lets you apply them through Azure Hybrid Benefit and knock roughly 30–55% off the compute price. Which is Microsoft’s polite way of admitting how much of that bill was the license in the first place. For a solo founder who has never bought a SQL Server license in their life? You pay full freight. You’re subsidizing an enterprise licensing model you have no part in.

You can shave more with 1-year or 3-year reserved capacity commitments (only available on vCore, not DTU). But committing three years to a database platform before you have paying customers is exactly the kind of over-engineering that killed my earlier projects.

---

Serverless: The One Genuinely Good Idea Here

The best thing Azure SQL Database offers is serverless — a sub-flavor of the vCore model, currently on General Purpose. Instead of paying for a fixed machine 24/7, you’re billed per second based on the vCores you actually use (around $0.0001450 per vCore-second), between a floor and a ceiling you set.

And the killer feature: auto-pause. If your database sits idle past a timeout you configure, it pauses and compute billing drops to zero. When a request comes in, it resumes. For a bootstrapper whose app gets three visitors on a Tuesday, this is genuinely the right shape — you don’t pay for compute at 4 AM when nobody’s awake.

Sound familiar? It should. It’s the same scale-to-zero idea Neon has offered on Postgres for years. Azure caught up on the concept, which tells you it’s a good one. But notice what auto-pause doesn’t pause.

---

The Costs Nobody Puts in the Headline: Storage and Backups

Compute is the number on the pricing page. Storage and backups are the numbers on the actual invoice, and they don’t pause when your database does.

Storage is billed per GB per month, separately from compute in the vCore model — around $0.115/GB on General Purpose. Modest, until your data grows.

Backups are the sneaky one. Point-in-time restore backup is included up to roughly 100% of your database size; go beyond that retention window, or add long-term retention for compliance, and you’re paying extra for backup storage. Choose zone- or geo-redundant backups and it climbs again. I’ve watched people budget carefully for compute and then get blindsided by a backup-storage line item they never modeled. When your database auto-pauses to save compute, the storage and backup meters keep ticking. That’s the fine print.

So the “cheap serverless database” is compute-cheap and quietly storage-and-backup-expensive. Every number needs a footnote. Which brings me to why I don’t do this to myself.

---

The Broken-Engineer Verdict: This Is Microsoft-Shop Gear

Azure SQL Database is a genuinely excellent product — for a specific person. If you already live in the Microsoft ecosystem, write .NET, run Active Directory, and your team’s muscle memory is T-SQL and SQL Server Management Studio, then Azure SQL Database is the path of least resistance and you should take it. Enterprises run mission-critical workloads on it for good reason. The Hybrid Benefit exists precisely because those shops already bought the licenses.

You are (probably) not that shop. You’re a founder trying to stretch every rupee until it screams. And for you, the whole thing is wrong on three counts:

1. You’re renting a license you’ll never own. Half the point of open-source Postgres is that there’s no engine tax. Azure SQL Database bakes the SQL Server license into every hour.
2. There’s no real $0 floor. Even serverless, which auto-pauses compute, keeps billing storage and backups. Basic tier’s ~$5/month is the closest to free, and it’s too weak for production.
3. It’s lock-in with a Microsoft accent. SQL Server’s T-SQL dialect isn’t standard Postgres or even standard MySQL. Migrating off later means rewriting queries. I refuse to get married on the first date.

For what it’s worth, this is the same reason I tell people to skip the big clouds entirely when they’re starting out. I laid out the full argument in Azure vs AWS for a broke founder — short version: you want neither yet.

What I Actually Run Instead

Postgres on Neon. That’s it. End of story.

Neon’s free tier costs $0/month and includes 100 projects, 0.5 GB of storage per project, and 100 compute-hours per project each month — with scale-to-zero always on, pausing after five minutes idle. No license. No Hybrid Benefit gymnastics. No T-SQL lock-in. When you outgrow free, storage on paid plans is about $0.35/GB-month (they cut it dramatically after the Databricks acquisition), and compute is pay-as-you-go.

I’ve run seven databases on that free tier at once — here’s exactly how. Try doing that on Azure SQL Database without a five-figure annual number appearing somewhere.

If you’re weighing where to run Postgres in the first place, I compared the serverless options in Neon vs Supabase vs Vercel Postgres. And if you’re still stuck on which engine to pick, PostgreSQL vs MySQL settles it: Postgres for ~99% of SaaS, and exactly when MySQL earns the exception. Notice that “managed SQL Server” isn’t on either list. That’s not an accident.

---

The Bottom Line

Azure SQL Database pricing isn’t expensive so much as it’s built for someone else’s budget. DTU hides the math, vCore shows you the math and it’s a big number, and serverless is a great idea wrapped around a license you’ll rent forever. Every model quietly assumes you’re a company that already pays Microsoft — because you mostly are, once you buy in.

You’re not that company yet. You’re one person trying to keep a product alive on $0 MRR. So don’t rent SQL Server by the second. Run Postgres, let it scale to zero on Neon’s free tier, and spend the $736/month you just saved on literally anything that gets you a customer.

This is the Broken Engineer Guide — I over-engineer everything, fail at business, and hand you the receipts so you don’t repeat my mistakes. Pick the boring free database and go build something.