Where each tends to fit
Generalisations, not rules:
- AWS — widest range of services and the largest partner and hiring pool
- Azure — natural fit for Microsoft-heavy organisations and hybrid with Windows Server
- Google Cloud — strong in data analytics, BigQuery and Kubernetes
What should decide it
In practice:
- Licences you already pay for (Microsoft especially)
- Your team's existing skills
- The managed services you will really use
- Regions close to your users
- Support and partner availability
Regions matter
Choose regions near your users. For US businesses that is usually one of the Virginia, Ohio, Oregon or California AWS regions, or the Azure and Google Cloud regions near them — each US state page on our site lists the nearest ones.
Keep the exit open
Infrastructure as code, containers and standard databases keep a future move possible. Lock-in is a trade-off, not a sin — just make it deliberately.
Compute and containers
All three providers offer virtual machines in a wide range of sizes, managed Kubernetes (EKS on AWS, AKS on Azure, GKE on Google Cloud), serverless functions (Lambda, Azure Functions, Cloud Functions) and serverless containers (App Runner and Fargate, Azure Container Apps, Cloud Run). For most general-purpose workloads, any of them will do the job well.
Differences show up at the edges: specific instance types, GPU availability in particular regions, how quickly new hardware arrives, and how each service is priced. If your workload depends on something specialised, check availability in the regions you need before deciding.
Databases and data
Each provider runs managed versions of the popular open-source databases — MySQL, PostgreSQL and others — plus its own designs: Aurora and DynamoDB on AWS, Azure SQL and Cosmos DB on Azure, Cloud SQL, Spanner, Firestore and BigQuery on Google Cloud. Microsoft SQL Server runs on all three, but licensing is often simplest on Azure.
For analytics, BigQuery is a common reason to choose Google Cloud; AWS and Azure have strong alternatives in Redshift and Synapse or Fabric. If your company already uses one provider's analytics stack, that usually outweighs small differences in features.
AI and machine learning
All three now offer managed AI platforms and access to large language models: Amazon Bedrock and SageMaker, Azure AI services including Azure OpenAI, and Google Cloud's Vertex AI. The models and features each provider offers change quickly, so compare them against your specific use case at the time you decide rather than relying on last year's comparison.
Consider where your data already lives. Moving large datasets between clouds to use a particular AI service adds cost and complexity.
Pricing and discounts
List prices for comparable compute are close between the three, and they change. What makes the real difference is how you buy:
- Commitment discounts — Savings Plans and Reserved Instances on AWS, reservations and savings plans on Azure, committed-use discounts on Google Cloud
- Spot or preemptible capacity for interruptible workloads
- Licence benefits — especially bringing Windows Server and SQL Server licences to Azure
- Sustained-use and automatic discounts on some Google Cloud services
- Data transfer charges, which vary by direction and region and are easy to overlook
Identity, security and governance
Identity is where the providers differ most in day-to-day use. Azure is built around Microsoft Entra ID, which many companies already use for Microsoft 365, so users and groups carry over naturally. AWS uses IAM with IAM Identity Center for single sign-on, and connects to most identity providers. Google Cloud uses Google identities and connects to others through Cloud Identity.
All three provide organisation-level controls, policy enforcement, audit logging and security posture tools. The best one for you is often the one your security team already knows.
Hiring and partners
Skills matter as much as features. AWS has the largest pool of engineers and partners, which can make hiring easier; Azure skills are common in companies with a Microsoft background; Google Cloud specialists are fewer but strong in data and Kubernetes. Whichever you choose, budget for training your own team, and consider a managed service or dedicated engineer to cover the gap while you build skills.
Multi-cloud: when it makes sense
Running across several clouds sounds like insurance, but it doubles much of the operational effort: two sets of identity, networking, monitoring, security tooling and skills. It makes sense when there is a concrete reason — an acquisition that brought a second cloud, a specific service only one provider offers, a customer requirement — and less sense as a general hedge.
A more practical form of insurance is portability: infrastructure as code, containers and standard databases make a future move possible without running two clouds today. Cloud Ninza manages AWS, Azure, Google Cloud and DigitalOcean, and can help you choose based on your workloads rather than any provider's preference.
A simple way to decide
If you are still undecided, a short structured process helps. Write down your three most important workloads and what each needs. Note the licences and skills you already have. Shortlist the providers that cover those needs in regions near your users. Build a small proof of concept of one real workload on each shortlisted provider and compare the experience, the performance and an honest cost estimate including support and data transfer.
Most companies find that one provider is a clear fit once they look at their own situation rather than general comparisons. If two are genuinely close, choose the one your team can run with confidence.
Support and account management
All three providers sell support plans with different response times and access to technical staff, priced in tiers or as a share of spend. For many small and mid-sized companies, the provider's basic or developer tier plus a partner or managed service provider is better value than the top support tier. Whatever you choose, make sure someone is responsible for opening cases, following them up and learning from them.
Getting started
If you are choosing a provider, send us your main workloads, your existing licences and where your users are. We compare the options for your case, help you run a small proof of concept, and set up whichever platform you choose with a secure landing zone. If you already run on one of them, we can manage it for you or review it against the provider's well-architected framework.
Whichever provider you choose, the account and the bill are yours, and our fee is always a separate line.
Remember that the decision is less permanent than it feels. Companies move workloads between clouds, add a second provider for a specific service, or bring workloads back to their own servers when the numbers change. Choosing well today matters, but designing so that you can change your mind tomorrow — with infrastructure as code, containers, standard databases and clean documentation — matters at least as much, whichever provider you end up with. A provider you can leave is a provider you can negotiate with.
Key takeaways
The short version:
- AWS for breadth, Azure for Microsoft shops, Google Cloud for data and Kubernetes
- Decide by licences, skills and services you will use
- Pick regions near your users
- Keep infrastructure as code so a move stays possible
AWS Management
AWS account management — landing zones, VPC design, EKS, RDS, SageMaker and Control Tower governance. Landing zone and Control Tower setup · VPC, subnet and transit design · EKS cluster operations and upgrades.
Get a costed configuration within one business day
VPS, dedicated servers, GPU clusters, migration and colocation are sized to the workload rather than sold off a list. Tell us what you are running and an engineer replies with a configuration and a price — provider cost and our service fee shown as two separate numbers, so you can see exactly what you are paying us for.