AWS vs Azure vs GCP for Mid-Market: How to Actually Choose
Read Time 13 mins | Written by: Vinayak Bhagat
Most comparisons of AWS, Microsoft Azure and Google Cloud are feature tables. They list services, regions and price points, and they go out of date as fast as the providers ship. For a mid-market company they also answer the wrong question. All three can run the systems a mid-market business needs. The question is which one your business can run well.
That depends far less on the providers than on you: the skills your people already have, the systems and licences you already own, where your data will live, and the commercial terms you can actually get. Those forces pull every company toward one provider before anyone opens a feature table.
So measure the pull. The right primary cloud for a mid-market company is the one it already leans toward on skills, estate, data and commercial terms, confirmed by running one real workload. We call those the Four Gravity Tests.
How should a mid-market company choose between AWS, Azure and Google Cloud?
Choose on fit, not features. Score each provider on four tests: which one your people can already run, how well it fits the identity, licences and systems you already own, where your data and analytics will live, and the commercial terms you can get. Then prove the leading choice on one real workload before committing. For most mid-market companies one primary cloud, chosen this way, beats spreading across all three.
Why Feature Tables Mislead Mid-Market Buyers
A feature table rewards breadth: more services, more regions, more options. A mid-market company uses a small, ordinary part of any provider: compute, storage, databases, networking, identity, backup, some analytics and, increasingly, a few AI services. On that part, the three providers are close enough that the differences rarely decide anything.
What does decide things is everything a feature table leaves out. A provider your team cannot operate will cost more in people than it saves in price. A provider that fights your existing identity and licensing will add work to every project. And a list price tells you little about the bill, which depends on how you buy and how you run. Those are the four tests.
The Four Gravity Tests
Score each provider on each test from your own evidence. The scores matter less than the reasons you write next to them, because the reasons are what you will have to defend.
| Test | The question | Evidence to collect |
|---|---|---|
| 1. Skills | Who can run it? | What your team has already built and supported, and on which provider |
| 2. Estate | What does it have to fit? | Your identity system, existing licences, and the systems that must connect to it |
| 3. Data | Where will your data live and be used? | Where your main data sits today, and the analytics and AI work planned on it |
| 4. Commercial | What terms can you actually get? | Written offers, commitment options, and how you buy software today |
Test 1: Skills — who can run it
The cloud you choose is the cloud your team will operate every day: patching, monitoring, troubleshooting at night, and explaining the bill. Look at what your engineers have actually built and supported, not what they would like to learn. Count the people who could run a production incident on each provider today.
A provider with no in-house skills is not ruled out, but it carries a cost you must plan for: training, hiring or outside help for the first year at least. Put that cost in the comparison, next to the price.
Test 2: Estate — what it has to fit
Every mid-market company already has an estate: an identity system that controls who can sign in to what, software licences, and core systems such as finance, CRM and file storage. A new cloud has to fit that estate, and the fit is not equal across providers.
If your identity and productivity tools already run on Microsoft, for example, how each provider works with that identity is a real factor to weigh. If your existing licences can or cannot be carried to a given cloud, that changes the cost. Ask each provider, in writing, how your specific identity system and licences would work with them, and test the answer in the one-workload proof below.
Test 3: Data — where it will live
Data is the heaviest thing you own in the cloud. It is expensive to move out once it is in, and the analytics and AI work you plan will follow it. Start from where your most important data sits today and what you intend to do with it over the next few years: reporting, a data warehouse, machine learning, or AI assistants on your own documents.
Then check the specific services you would use for that work on each provider, by running a small piece of it rather than reading about it. If some of your data must stay in a particular country, check that each provider has a region there that offers the services you need. And write down what it would take to move the data out again, because that is the cost of changing your mind later.
Test 4: Commercial — the terms you can actually get
List prices are the least useful number in the decision. What you will pay depends on commitment discounts, migration credits, support plans, and whether you buy through an agreement you already hold or a partner. Ask each provider for a written offer for the workloads you actually plan to run, over the period you expect to run them.
Read the conditions as carefully as the discount. Credits and discounts usually come with commitments, and a commitment is only a saving if you will really use what you promised to buy. Whatever you choose, the bill after the move depends more on how you run than on the rate you negotiated, which is what a FinOps practice that does not slow engineering is for.
Prove It on One Workload
The four tests usually point clearly at one provider. Before you sign anything long, prove that answer on one real workload: something that matters enough to be representative, but not so much that a problem would hurt the business. A reporting database, an internal application or a file service is often a good candidate.
Run it end to end: build it, connect it to your identity system, back it up, restore it, watch it for a few weeks and read the bill. The proof answers questions no sales conversation can: whether your team is comfortable operating it, whether the estate fit is as smooth as promised, and what a month really costs.
Each provider publishes an architecture framework that is worth reading during the proof, whichever you choose: AWS Well-Architected, the Azure Well-Architected Framework and the Google Cloud Architecture Framework. Once the proof holds, plan the move itself with our four-wave migration roadmap.
The Cloud Provider Scorecard
This article explains the four tests. The scorecard is where you run them on your own business: a page per test with the questions to answer and room to score each provider with your reasons, a plan for the one-workload proof, and a decision page for whoever signs off. It comes with no scores filled in, because the answer depends on you.
When Multi-Cloud Makes Sense, and When It Does Not
Running on several clouds sounds like protection against any one provider. For most mid-market companies it doubles the skills you need, the identity integrations you maintain and the bills you have to read, without removing much risk. One primary cloud, chosen well and run well, is usually the stronger position.
There are good reasons for a second cloud: a specific service you cannot get elsewhere, an acquisition that arrives on another provider, or a customer who requires it. Treat each as a deliberate exception with an owner, not a strategy. For transparency: Ontrac sells cloud services across providers, so we have no stake in which one you choose.
Three Mistakes in Choosing a Cloud
Choosing on the biggest discount
A large first-year credit can hide a poor fit on skills or estate, and the fit costs you every year after the credit runs out. Weigh the offer last, after the other three tests.
Letting one enthusiast decide
An engineer who loves one provider is an asset, but the decision outlives them. Make the four tests the record, so the choice survives a change of team.
Committing before the proof
A multi-year commitment signed before any workload runs locks in an answer you have not tested. Prove first, then commit to what you now know you will use.
We run the four tests and the proof with you, on your workloads.
Our cloud team works across providers. We help mid-market and private-equity-backed companies score the options on their own evidence, run the one-workload proof, and plan the move, with FinOps built in from the first month.
Book a cloud workshopFrequently Asked Questions
Which cloud is best for a mid-market company?
The one your business already leans toward. AWS, Azure and Google Cloud can all run what a mid-market company needs, so the deciding factors are your team's skills, how each fits your identity, licences and existing systems, where your data will live, and the terms you can get. Score each provider on those, then prove the leading choice on one workload.
Is it cheaper to choose the provider with the lowest prices?
Not necessarily. Your bill depends on commitment discounts, credits, support plans and, above all, how you run once you are there. A provider your team cannot operate well usually costs more in people and rework than it saves on price. Compare written offers for your own workloads, and plan how you will manage cost after the move.
Should a mid-market company use more than one cloud?
Usually not as a strategy. A second cloud doubles the skills, integrations and bills to manage. Add one only for a specific reason, such as a service you cannot get elsewhere, an acquisition already running there, or a customer requirement, and give it a named owner.
How do you test a cloud provider before committing?
Run one real but low-risk workload end to end: build it, connect it to your identity system, back it up and restore it, operate it for a few weeks and read the bill. That shows whether your team can run it, whether it fits your estate, and what it really costs, before any long commitment.
Keep Reading
→ Cloud Migration for Mid-Market: A Practical Modernization Roadmap — how to move once you have chosen.
→ FinOps in 2026: How to Cut Cloud Costs Without Slowing Engineering — keeping the bill in check after the move.
This article is general information about choosing a cloud provider and is not legal, financial or procurement advice. Provider services, terms and prices change; confirm every detail with the providers and review commitments with appropriate advisors.
