If you are going to sit one cloud certification in Australia in 2026, sit an AWS one. In our count of 315 AI job ads from July to September 2026, AWS was named in 16% of postings, Azure in 8% and Google Cloud in 6% — and that ratio matches what Seek and LinkedIn show for Australian roles, where AWS has led for a decade. But read the next number before you book the exam: a certification of any kind was mentioned in 7% of those ads. Certifications open a door in a few specific places. They are not the thing most employers are looking for.
Which cloud do Australian employers actually use?
AWS, by a wide margin, with Azure second and Google Cloud a distant third — and the gap is wider in Australia than globally. AWS opened its Sydney region in 2012 and its Melbourne region in 2023; the big banks, Atlassian, Canva and most of the startup ecosystem run on it. Azure's share is concentrated where Microsoft already is: government, health, education and the enterprises that bought Office 365 and Teams and let the cloud follow. Google Cloud has real footholds in analytics-heavy shops (BigQuery) and in some of the retailers and media companies, but it is the platform you learn because a specific employer runs it, not because the market does.
So the first question is not "which certification" but "which employers". If you are targeting government or a Microsoft-shop enterprise, Azure. Everywhere else, AWS.
Which certification, then: Practitioner, Associate or Specialty?
For an AI role, the vendor's AI-specific entry certification, followed — if at all — by an associate-level engineering one. The three candidates:
| Vendor | Entry AI certification | What it signals |
|---|---|---|
| AWS | AWS Certified AI Practitioner | You know the AWS AI service catalogue (Bedrock, SageMaker, the managed services), the basics of prompting and RAG, and responsible-AI vocabulary. Foundational, not hands-on. |
| Microsoft | Azure AI Fundamentals (AI-900) | The same layer for Azure OpenAI, Cognitive Services and Azure ML. Recognised inside government and enterprise Microsoft shops. |
| Google Cloud | Generative AI Leader | Business-level generative-AI literacy on Vertex and Gemini. Lightest of the three; aimed at non-engineers. |
All three are multiple-choice exams that test vocabulary and service knowledge, and all three can be passed in three to six weeks of study alongside a job. None of them proves you can build anything — which is exactly why they appear in 7% of ads and shipped appears in 39%.
The associate-level engineering certifications (AWS Solutions Architect Associate, Azure Administrator, Google Professional Cloud Engineer) are worth more to an employer, take two to four months, and are about cloud, not AI. Do one only if your target role is infrastructure-heavy — MLOps, platform, or the LLMOps roles that are starting to appear.
When is a certification actually worth the money in Australia?
In four situations, and in all four it is the door the certification opens rather than the knowledge:
- Government and defence panels. Recruitment scoring often gives points for a named certification; two candidates with equal portfolios are separated by it.
- Consultancies and managed-service providers. AWS, Microsoft and Google partner tiers require a headcount of certified staff. A consultancy will pay for your exam and sometimes hire for it, because your certification counts towards their partner status.
- Career changers with no tech employment history. A certification is a cheap, fast, legible signal that you are serious, and it gets a CV past the first screen. It does not get you through the interview.
- Visa and skills-assessment processes, where a recognised qualification has legal weight that a GitHub profile does not.
Outside those four, the exam fee (US$100 for AI Practitioner, US$99 for AI-900, US$99 for the Google exam at the time of writing) buys you less than the same money spent on cloud credits to deploy something.
What do the ads ask for instead of certifications?
Evidence of shipping. The skills most named alongside AWS in our sample were monitoring and observability (17%), Docker (10%), CI/CD (10%), Kubernetes (8%) and Terraform (7%) — the tools you use to put something in production and keep it there. A candidate who can show a deployed model-backed service with a dashboard, a deploy pipeline and a cost line has demonstrated every one of those. A candidate with AI Practitioner has demonstrated that they can pass AI Practitioner.
The strongest position is both: the certification to get through the screen, the deployed project to get through the interview. If you can only do one, do the project.
How should I prepare so the certification is not wasted?
Study by building on the platform rather than by reading about it. Every AWS AI Practitioner domain — Bedrock, RAG on AWS, guardrails, the responsible-AI material — can be learned by deploying one small application that uses them, and the application is what you show afterwards. Square 1's exam prep is built on this idea: a free lane with the full question bank, and a Pass Track that grades you on practice sittings plus the build, so the certificate and the portfolio arrive together. Expect three to six weeks for AI Practitioner or AI-900 at five hours a week.
For the broader question of whether any certification moves an employer, are AI certifications worth anything to employers is the companion piece, written for the hiring manager's side of the table.
The short answer, by situation
- Targeting startups or SaaS engineering teams: skip the certification; deploy three things; learn AWS by using it.
- Targeting government, banks or a Microsoft-shop enterprise: Azure AI-900 first, then a portfolio.
- Targeting a consultancy or MSP: AWS AI Practitioner now, Solutions Architect Associate within a year — they will likely pay.
- Career changer with no tech history: AWS AI Practitioner for the screen, plus one deployed project for the interview. Do them in the same six weeks.