AWS vs Azure vs Google Cloud in Australia: regions, IRAP, cost and fit

The short version, by situation

For most Australian organisations, the right cloud is the one that matches what you already run, because all three meet the common Australian requirements. Here is where each tends to win.

Your situation	Usually the best fit	Why
Microsoft 365, Entra ID and Windows servers already in place	Azure	Identity, licensing and support carry over; one vendor relationship
Custom software product or SaaS built by a development team	AWS	Broadest managed service catalogue and the largest pool of engineers who know it
Data warehouse, analytics or Google Workspace at the centre	Google Cloud	BigQuery and the data tooling are its strongest area
Commonwealth or state agency work at PROTECTED	Azure or AWS, then Google Cloud	All three have IRAP scope; Azure adds Canberra regions
AI features that must process data in Australia	AWS	Widest onshore model list at the time of writing
Small team with one app and modest traffic	Whichever your developer knows best	Skills matter more than the price differences

The rest of this page explains the detail behind each row, so you can test it against your own case.

Where are each provider’s Australian regions?

AWS, Azure and Google Cloud each run a Sydney and a Melbourne region; only Azure also has regions in Canberra. The table reflects each provider’s region list at the time of writing.

Provider	Sydney	Melbourne	Canberra
AWS	Asia Pacific (Sydney), ap-southeast-2, 3 availability zones	Asia Pacific (Melbourne), ap-southeast-4, 3 availability zones, opt in	None
Azure	Australia East (New South Wales), 3 availability zones	Australia Southeast (Victoria), no availability zones listed	Australia Central and Australia Central 2, restricted access
Google Cloud	australia-southeast1, 3 zones	australia-southeast2, 3 zones	None

Availability zones matter more than region counts. A region with three zones lets you run a database and application across separate data centres, so one facility failing doesn’t take you offline. On Azure, Australia Southeast has no availability zones listed, which makes it a disaster recovery partner for Australia East rather than a first choice for a highly available primary system.

Azure’s Canberra regions sit in CDC data centres and Microsoft describes them as designed for government and critical infrastructure. Access is restricted, so most businesses won’t use them.

Latency is similar across providers between Sydney and Melbourne. If your users are in Perth, all three serve them from the east coast, and the difference between providers is small next to the distance.

What does IRAP mean for choosing a cloud?

All three providers have IRAP assessments covering listed services at the PROTECTED level, so IRAP rarely separates them; the services in scope do. IRAP is the Australian Signals Directorate’s program for independent assessment against the Information Security Manual.

• AWS states its IRAP assessment covers in scope services in the Sydney and Melbourne regions, publishes the services list, and provides the PROTECTED report through AWS Artifact.
• Microsoft states the IRAP assessment of Azure covers the four Australian regions for data up to and including PROTECTED, and offers an Azure Policy initiative mapped to ISM PROTECTED controls.
• Google states an independent assessor found Google Cloud and Google Workspace strongly aligned with PROTECTED level controls, and publishes a products in scope list.

The catch is in the word “listed”. A newer managed service, or a feature in preview, may not be in scope yet. Before designing a government system, check every service you plan to use against the scope list, and remember that the agency, not the cloud provider, authorises the system. Our IRAP explained guide covers who decides what.

Which AI models can you run in Australian regions?

At the time of writing, AWS offers the most AI models processed onshore, through Amazon Bedrock’s Sydney and Melbourne regions. Bedrock’s regional availability list includes several Claude models and many open weight models in Australia. Azure OpenAI and Google’s Gemini platform each keep processing onshore for a narrower set of models.

This changes month to month, and the newest models usually launch on global routing first. We keep the detail in a dedicated guide, Bedrock vs Azure OpenAI vs Vertex AI in Australia, rather than repeat it here. If onshore AI processing is a hard requirement, check that guide and the vendor’s page before choosing a cloud for the rest of the system.

Which price differences actually matter?

List prices for compute are close enough that they rarely decide the choice; egress, support plans, commitment discounts and currency do.

Cost item	AWS	Azure	Google Cloud
Internet egress from Australia	US$0.114 per GB for the first 10 TB from Sydney	First 100 GB a month free, then US$0.12 per GB (Microsoft network)	Check the network pricing page for your tier
Transfer between Australian regions	US$0.08 per GB Sydney to Melbourne	US$0.08 per GB between Oceania regions	Check the network pricing page
Billing currency	USD, billed by AWS Australia with GST	Depends on your agreement or partner	AUD available for Australian accounts
Commitment discounts	Savings Plans and Reserved Instances	Reservations and savings plans	Committed use discounts

Prices are at the time of writing and change; confirm on the linked pages. Support is the other line people forget: production systems usually need a paid support plan with a response time target, and its price is often a percentage of monthly spend. The practical lessons:

1. Egress adds up for media and data heavy apps. An app serving 5 TB a month of files from Sydney pays around US$570 in AWS internet egress alone, before a CDN.
2. Replication across regions costs money every month. A Sydney primary with a Melbourne disaster recovery copy pays transfer on every change.
3. USD billing moves with the exchange rate. Budget a buffer, or prefer AUD billing where it is offered.
4. Commit only after the workload settles. Our guide to reducing cloud costs explains how to size commitments without locking in waste, and covers serverless versus containers.

How much does Microsoft 365 or Google Workspace tip the decision?

A lot, because identity is the hardest thing to run twice. If your staff already sign in with Entra ID, Azure uses the same accounts, groups and conditional access policies, and your Microsoft agreement may already include Azure credits or support.

That doesn’t force everything onto Azure. AWS and Google Cloud both accept Entra ID as an identity provider through single sign on, so a Microsoft 365 business can run a custom app on AWS without a second set of staff passwords. See our guide to SSO with Entra ID for business apps.

Google Workspace works the same way in the other direction. A business that runs on Workspace and keeps its reporting in BigQuery will find Google Cloud’s identity and data tools familiar, while still being free to host an individual app elsewhere.

Skills count as much as licences. If your internal IT team knows Azure and your developer knows AWS, decide who will support the system in year three, not who builds it in month one.

Does an Australian region make your data sovereign?

No. Choosing Sydney or Melbourne gives you data residency, meaning the data is stored in Australia; it doesn’t change who owns the provider. AWS, Microsoft and Google are all US headquartered, so foreign laws can apply to them whichever region you choose.

For most businesses, residency plus encryption, access logging and contract terms is enough. For some government, defence and critical infrastructure work, buyers ask about sovereignty directly. Our guide to data residency vs data sovereignty explains the difference and what Australian buyers typically require.

When is multi cloud worth the overhead?

Multi cloud is worth it when each cloud does a different job, and rarely when the same system is spread across two.

Sensible patterns:

• Microsoft 365 and Entra ID for staff, with the product on AWS or Google Cloud.
• BigQuery for analytics while the operational app runs elsewhere.
• One cloud’s AI models called from an app hosted on another, where onshore model availability requires it.

Patterns to avoid at SMB and mid market scale: running the same database across two providers for resilience, or building everything with lowest common denominator tools to keep the option of moving. Both double the security, monitoring and skills you need, and a second region with the same provider usually gives better resilience for less effort.

How AWLabs approaches cloud choice

We build on AWS, Azure and Google Cloud and hold no partner status with any of them, so the recommendation follows your systems and team. We use Australian regions where data needs to stay onshore, as agreed in the Schedule of Work, and list any service that sits outside that boundary. Senior engineers lead every build, and the infrastructure is written as code so it can be reviewed and handed over. See cloud infrastructure and our cloud migration cost guide.