What do Azure AI services cost in Australia?
Azure AI services pricing in Australia is metered per service: Azure OpenAI per million tokens, AI Search per search unit per hour, Document Intelligence per 1,000 pages, and logs per GB. As checked in October 2026, GPT-5.4 mini on Global Standard listed at about A$1.07 per million input tokens in Australia East, but search and logging usually cost more than the model.
Most Azure AI budgets start with the wrong number. Someone opens the Azure OpenAI price list, sees a mini model at about a dollar per million input tokens, multiplies by a hopeful usage guess, and gets a figure small enough to approve on a credit card. In the hypothetical knowledge assistant later in this post, model tokens come to about a fifth of the monthly bill. The search index, the second search replica you need for an uptime commitment, the log workspace capturing every prompt, and the non-production environment nobody switched off make up the rest.
How the prices in this post were checked. Every Azure unit price below comes from Microsoft's Azure Retail Prices API (the data behind azure.microsoft.com/pricing and the Azure pricing calculator), queried for the Australia East region in AUD on 8 October 2026. Microsoft's AUD list prices on that date worked out at about A$1.4246 per US$1, and where Microsoft quotes only USD we convert at that rate. Figures exclude GST and any enterprise agreement discount. Prices change often, so confirm current rates in the Azure pricing calculator before you commit a budget.
If you are still choosing a cloud platform, start with our AWS vs Azure vs Google Cloud for AI in Australia comparison. This post assumes Azure is already the platform.
Key takeaways
- In a typical retrieval assistant, Azure AI Search, not model tokens, is the largest line. An uptime SLA needs two replicas, which doubles it.
- Data residency changes the price list. Keeping prompt processing inside Australia on a current-generation model needs Regional Provisioned throughput, with a minimum of 25 PTUs.
- Provisioned throughput only matches pay-as-you-go cost at near-full utilisation around the clock.
- Microsoft 365 Copilot listed at A$44.90 per user per month paid yearly (ex GST, October 2026). Audit permissions and pilot before licensing at scale.
- A four-week measured pilot gives a budget finance can defend.
How is an Azure AI bill built?
An Azure AI workload is several metered services wired together, each billing on a different unit. Knowing the unit tells you which number to estimate.
- IngestBlob storage per GB, Document Intelligence per 1,000 pages
- IndexEmbeddings per token, AI Search per hour per search unit
- GenerateModel tokens per million, or PTUs per hour
- ObserveLog Analytics per GB ingested and retained
- RunPeople time for governance, evaluation and support
| Service | What you pay for | Australia East list price (AUD, ex GST) |
|---|---|---|
| Azure OpenAI GPT-5.4 mini, Global Standard | Per million tokens | A$1.0685 input, A$0.1068 cached input, A$6.4107 output |
| Azure OpenAI GPT-4.1 mini, regional Standard | Per million tokens | A$0.63 input, A$0.16 cached input, A$2.51 output (converted from US$0.44, US$0.11, US$1.76) |
| text-embedding-3-small, regional | Per million tokens | about A$0.03 (converted from US$0.022) |
| Azure AI Search Basic | Per search unit per hour | A$0.1895 (about A$138 a month) |
| Azure AI Search Standard S1 | Per search unit per hour | A$0.6325 (about A$462 a month) |
| Semantic ranker, standard plan | Per 1,000 queries | A$1.4246 |
| Document Intelligence Read | Per 1,000 pages | A$2.1369 (A$0.8548 beyond 1 million pages) |
| Document Intelligence prebuilt models | Per 1,000 pages | A$14.246 |
| Document Intelligence custom extraction | Per 1,000 pages | A$42.7381 |
| Speech to text, real-time / batch | Per audio hour | A$1.4246 / A$0.2564 |
| Neural text to speech | Per million characters | A$21.369 |
| Log Analytics ingestion | Per GB after 5 GB free each month | A$4.7582 |
| Blob storage, Hot LRS | Per GB per month | A$0.0285 |
| Internet data transfer out | Per GB after the first 100 GB a month | A$0.171 (Microsoft network routing) |
Source: Microsoft Azure Retail Prices API, Australia East, AUD, as checked October 2026. Monthly figures assume 730 hours.
How does Australian data residency change Azure AI pricing?
This is the line item most budgets get wrong, and it sits in a dropdown labelled "deployment type".
Microsoft's deployment types documentation (updated August 2026) states that data stored at rest stays in your chosen Azure geography for every deployment type. Where the prompt is processed is separate:
- Global deployments may process prompts in any Azure region.
- Data Zone deployments process within a Microsoft-defined zone. For this region that is Asia Pacific, which covers several countries beyond Australia.
- Standard (regional) and Regional Provisioned deployments process prompts within the Azure geography you choose.
The catch is availability. On Microsoft's region tables (checked October 2026), pay-as-you-go regional Standard in Australia East offers GPT-4.1 mini, GPT-4o and the embedding models. The GPT-5 family is available in Australia East on Global Standard, Data Zone Standard and Regional Provisioned. Australia Southeast did not appear in those Azure OpenAI tables, which matters if you planned a second Australian region for disaster recovery.
| Global Standard (any region) | Onshore options | |
|---|---|---|
| Billing | Per token, A$1.0685 in / A$6.4107 out per 1M | Data Zone APAC: A$1.2821 in / A$7.6929 out per 1M (processing in APAC, not Australia only) |
| Processing inside Australia | Not guaranteed | Regional Provisioned only, minimum 25 PTUs |
| Cheapest onshore entry point | No minimum spend | 25 PTUs x A$464.42 per PTU per month reservation = A$11,610.51 a month |
| Same deployment on hourly PTU billing | Not applicable | 25 PTUs x A$3.2481 x 730 hours = A$59,277.82 a month |
| Onshore pay-per-token alternative | Not applicable | Older GPT-4.1 mini on regional Standard, A$0.63 in / A$2.51 out per 1M |
Read that table before a residency requirement gets written into a project charter. If your policy says "prompts must be processed in Australia" and you want a current-generation model, the floor is a provisioned reservation of roughly A$11,600 a month before a single question is answered. If the policy is "data at rest in Australia, processing in Asia Pacific", Data Zone Standard costs about 20% more per token than Global with no minimum. If GPT-4.1 mini is good enough, regional Standard keeps processing onshore on a per-token bill.
Which of those you need is a legal and risk question before it is a technical one. Our guide on where AI data actually goes covers the Privacy Act and contract angles, and our private AI infrastructure service explains when running a model on your own infrastructure becomes the cheaper way to meet a strict residency rule.
What does your data policy require for prompt processing?
- No location restriction on processing, data at rest in Australia is enoughGlobal Standard: lowest token price, newest models first
- Processing must stay in Asia PacificData Zone Standard: about 20% more per token, no minimum
- Processing must stay in Australia, an older mini model is acceptableRegional Standard with GPT-4.1 mini, per-token billing
- Processing must stay in Australia on a current modelRegional Provisioned: 25 PTU minimum, budget A$11.6K+ a month
- Large overnight jobs with no latency needBatch: half the Global Standard rate, 24-hour target turnaround
Is pay-as-you-go or provisioned throughput cheaper?
Provisioned throughput units (PTUs) reserve model capacity, billed per PTU per hour whether or not you send a request. Microsoft's provisioned throughput sizing guide (updated September 2026) lists, for GPT-5.4 mini, a Global and Data Zone minimum of 15 PTUs, a regional minimum of 25 PTUs, and 7,900 input tokens per minute per PTU, with one output token counting as six input tokens.
That lets you check break-even yourself. Fifteen Global PTUs on a monthly reservation cost 15 x A$370.40 = A$5,555.95 a month. At 100% utilisation for 30 days they process about 5.1 billion normalised tokens, which would cost roughly A$5,470 on Global Standard. A GPT-5.4 mini reservation only matches pay-as-you-go when it runs flat out every hour of the month. On hourly billing the same 15 PTUs cost A$15,599.37.
PTUs earn their keep when you need predictable latency for a customer-facing or high-volume system, or when onshore processing on a current model is mandatory. For an internal assistant used in business hours, pay-as-you-go is nearly always cheaper. Microsoft also notes a reservation does not guarantee capacity, so create the deployment first and buy the reservation second.
Which Azure AI costs do people miss?
Prompt length, every time. Retrieval assistants send system instructions, retrieved chunks and conversation history with every question. A 6,000-token prompt for a one-line question is normal, so retrieval settings are a cost lever as well as a quality one.
Caching only works on identical openings. Per Microsoft's prompt caching documentation, a prompt must be at least 1,024 tokens and its first 1,024 tokens must match exactly for a cache hit. In-memory caches typically clear after 5 to 10 minutes of inactivity. Put stable instructions first and anything that varies last. On GPT-5.6 and later models, cache writes can also be charged.
Reasoning tokens bill as output. GPT-5 family models generate internal reasoning tokens billed at the output rate. A higher reasoning effort setting can multiply output spend on the same question.
AI Search runs by the hour. A search service bills every hour it exists. Microsoft's service limits page says the query uptime SLA needs two or more replicas. The same page notes Basic does not support private connections to a Foundry resource for indexers, so private networking for AI enrichment pushes you to S1 or higher.
Logging the conversation. Capturing full prompts and responses for audit is sensible, and at A$4.76 per GB after the free 5 GB it adds up on a busy assistant. Decide retention on purpose.
Non-production environments. A dev and test copy carries its own search service and logs. A Basic search unit left running all month is A$138.
Data transfer. The first 100 GB a month of egress is free and AI responses are small. Egress matters when you move large document sets between regions or clouds.
People time. Someone owns evaluation, prompt changes, access reviews, model upgrades and the monthly cost review. That never appears on the Azure invoice, and on a modest workload it can exceed everything that does. Our AI implementation cost breakdown covers build and run labour.
Worked example 1: what does an internal knowledge assistant cost?
Consider a hypothetical mid-sized organisation that puts its policies, procedures and contract templates behind an internal assistant. These assumptions are illustrative only:
- 40,000 questions a month (for example, 100 regular users asking 20 questions on each of 20 working days)
- 6,000 input tokens per question, a quarter served from cache
- 500 output tokens per answer
- 20,000 documents, about 50 million tokens to embed once
- GPT-5.4 mini on Global Standard, AI Search S1 with two replicas for the SLA, semantic ranker on every query
- 20 GB a month of logs, 50 GB of source documents, a Basic search unit for dev and test
- Model input, 180M uncached tokens x A$1.0685/MA$192.33
- Model cached input, 60M tokens x A$0.1068/MA$6.41
- Model output, 20M tokens x A$6.4107/MA$128.21
- Embeddings, 52M tokens (one-off index plus queries)A$1.63
- AI Search S1, 2 replicas x 730 hrs x A$0.6325A$923.45
- Semantic ranker, 40,000 queries x A$1.4246/1,000A$56.98
- Log Analytics, 15 GB billable x A$4.7582A$71.37
- Blob storage, 50 GB x A$0.0285A$1.43
- Dev and test search, Basic x 730 hrs x A$0.1895A$138.34
- Estimated monthly Azure total (excludes app hosting and people)A$1,520.15
Model tokens are about 22% of this total. Search, at roughly 65%, is the line to optimise first:
- Moving to AI Search Basic with two replicas brings search to A$276.67, if the index fits Basic's 15 GB storage and 5 GB vector quota per partition and you do not need private connections for enrichment.
- Switching to GPT-4.1 mini on regional Standard keeps prompt processing in Australia and drops the token lines to about A$172.38, in exchange for an older, less capable model.
App hosting (App Service, Container Apps or a Copilot Studio front end) is left out because it varies so much by design.
Worked example 2: what does invoice and form processing cost?
Now a hypothetical finance team processing 10,000 supplier invoices a month, averaging two pages each, so 20,000 pages.
| Approach | Monthly Azure cost and trade-off | |
|---|---|---|
| Prebuilt invoice model | Document Intelligence, 20 x A$14.246 per 1,000 pages | A$284.92. Least build effort, fields arrive structured |
| Prebuilt on a commitment tier | 20,000-page pre-built commitment tier | A$270.67 a month flat. Only worth it if volume is steady |
| Read OCR plus a language model | Read at A$2.1369 per 1,000 pages, then GPT-5.4 mini on Batch | A$42.74 + A$34.19 = A$76.93. Cheapest, but you build and test the extraction prompts |
| Custom extraction model | Document Intelligence custom, 20 x A$42.7381 per 1,000 pages | A$854.76. For unusual layouts the prebuilt model misses |
The language model line assumes 4,000 input and 400 output tokens per invoice on Global Batch (A$0.5342 and A$3.2054 per million). Batch suits this job because invoices can wait overnight, and Microsoft prices Batch at half the Global Standard rate with a 24-hour target turnaround.
Add about A$24 for 10 GB of logs and around A$1 a month of storage, and the bill sits between roughly A$100 and A$310 a month depending on the path. The cheaper path costs more in engineering and ongoing testing. For the architecture behind it, see our multi-agent document processing guide.
Speech follows the same logic. Transcribing 500 hours of recorded meetings on batch costs A$128.20; the same audio in real time costs A$712.30. If nobody needs the transcript during the meeting, batch is the default.
How much do Microsoft 365 Copilot and Copilot Studio cost in Australia?
Copilot licences come from a different budget line (per user, not per token) but end up in the same conversation.
- Microsoft 365 Copilot was listed on Microsoft's Australian site at A$44.90 per user per month paid yearly, or A$47.15 per user per month paid monthly on an annual commitment, excluding GST (October 2026). It needs a qualifying Microsoft 365 licence underneath. A Copilot Business plan is also listed for organisations up to 300 users; check Microsoft's site because it has carried promotional pricing.
- Copilot Studio bills in Copilot Credits. Microsoft's Australian pricing page listed a 25,000-credit capacity pack at A$299.30 per pack per month, and the Azure pay-as-you-go meter is A$0.0142 per credit (US$0.01). Credit use depends on what the agent does (scripted answers, generative answers, actions, grounding on your organisation's data), as set out in Microsoft's Copilot Studio licensing guide.
At list price, 100 Copilot users cost A$4,490 a month, or A$53,880 a year. That is why order matters: run the Copilot readiness and permissions audit first, license a pilot group, measure, then expand. Our Copilot readiness service covers that sequence.
How do you get an Azure AI budget you can defend?
A spreadsheet estimate rests on guessed prompt sizes and guessed usage. A short pilot replaces both guesses with measurements.
- Week 1Set the boundariesAgree the residency rule, the model shortlist and the deployment type. This decides the price list you are using.
- Week 2Build a thin pilotGlobal or regional Standard, Basic search, logging on. Tag every resource with a cost centre.
- Week 3Measure real usagePull tokens per question, cache hit rate and search queries from Azure Monitor and Cost Management.
- Week 4Model the production billScale measured unit costs to forecast volume, add SLA replicas, non-production and logs, set budget alerts.
After go-live, a few habits keep the bill honest: Azure budget alerts at 50%, 80% and 100% of forecast, a monthly cost review by tag, a record of which model version each deployment runs so an upgrade does not quietly change the price, and a quarterly recheck of the price list. Where nobody internal has time, that is the ongoing work our managed AI services cover.
How Solve8 can help
Our consultants model Azure AI costs against your residency rule and real usage, then build the pilot that turns the estimate into a measured number. Our AI strategy work starts with exactly this cost and residency model. Contact us for a second opinion on an estimate.
Sources: Unit prices from the Microsoft Azure Retail Prices API (prices.azure.com, Australia East, AUD), checked 8 October 2026. Deployment types and data residency from Microsoft Learn, "Understanding deployment types in Microsoft Foundry Models" (August 2026). Model availability from Microsoft Learn, "Region availability for Foundry Models sold by Azure" (September 2026). PTU minimums and throughput from Microsoft Learn, "Determine PTU sizing for a workload" (September 2026). Prompt caching rules from Microsoft Learn, "Prompt caching with Azure OpenAI" (August 2026). AI Search limits from Microsoft Learn, "Service limits for tiers and SKUs" (September 2026). Microsoft 365 Copilot and Copilot Studio AUD prices from microsoft.com/en-au, checked October 2026. All worked examples are hypothetical.
