Skip to content
InSearch
How it works
Enterprise search software ChatGPT Enterprise pricing Coveo vs Lucidworks Dust vs GoSearch Dashworks alternatives Glean pricing Glean alternatives Microsoft Copilot alternatives Dropbox Dash pricing Dropbox Dash vs Glean Dust vs Gemini Enterprise Guru pricing Coveo pricing Coveo alternatives Bloomfire alternatives Guru alternatives Moveworks alternatives Claude Ask Your Org Claude Enterprise pricing Notion AI pricing Elastic Enterprise Search pricing Perplexity Enterprise pricing Copilot Studio pricing Amazon Kendra pricing Unleash pricing ChatGPT Business pricing Copilot Business pricing Kore.ai pricing Claude Team pricing Dust pricing Grok Business pricing Hebbia pricing Hebbia alternatives Slack AI pricing Algolia pricing Pinecone pricing Azure AI Search pricing Search management software Law firm knowledge management software Internal knowledge base software Amazon OpenSearch Service pricing Gemini Enterprise vs ChatGPT Enterprise ChatGPT Enterprise vs Copilot Amazon Quick Suite pricing Amazon Q pricing Gemini Enterprise pricing Copilot pricing Slack enterprise search Enterprise search use cases Enterprise search tools AI work assistant Knowledge base search AI enterprise search Internal company search engine Internal knowledge search Workplace search Search all company apps AI knowledge assistant Unified search Search Slack & Drive AI assistant for work Knowledge management RAG for enterprise Cited AI answers Confluence search Notion search AI for support agents Sales enablement search Engineering search Enterprise search for remote teams HR knowledge assistant Permission-aware search Secure AI search Onboarding knowledge SharePoint search Intranet search See all →
Security Pricing Blog Contact Sign in
Procurement manager comparing two printed vendor proposals at a desk

Compare · Azure AI Search pricing

Azure AI Search pricing by tier, search unit and semantic ranker, and what it costs for company search

Most pages on this topic list the tiers for developers who already chose Azure. This one is for the person who has been asked what Azure AI Search would cost behind employee search, so people can ask one question and get an answer from Drive, Slack, SharePoint and Confluence.

One search · cited answers · permission-aware

Answer Console
Your files only
Reading your files…

That request did not get through. Tap Ask InSearch again.

No account needed. Your files are read once to answer, then deleted from our servers, and never used to train any model.

Short answer

Azure AI Search is billed per search unit per hour, and a unit is one replica times one partition. In the East US price feed Basic is $0.101 an hour (about $74 a month), S1 is $0.336 (about $245), S2 is $1.344 (about $981) and S3 is $2.688 (about $1,962), and semantic ranker adds $1 per 1,000 queries. A 200 person company on Basic with three replicas pays about $221 a month before semantic ranking and before anyone builds the connectors.

The honest answer is that the service line is small. We work it through with Microsoft's own published rates and limits, and say plainly what the invoice leaves out. We sell InSearch, which competes with building on a search service, so where Azure is the cheaper line item we say so.

Drive Slack Notion Confluence Jira Salesforce

Cited to source docs

Permission-aware · never trains on your data

Why it works

What your team gets with Azure AI Search pricing

You pay for capacity, not for searches

Each search unit has an hourly rate whether it serves one query or a million. A replica times a partition is a unit, so adding a copy for availability multiplies the bill.

Availability needs more than one unit

Microsoft documents that its SLA requires two replicas for read-only workloads and three or more for read and write. A single cheap unit has no SLA.

The cheap bill is the service only

Azure AI Search indexes what you send it. Someone builds and runs the connectors for Slack, Drive and Jira, keeps permissions in sync and writes the answer step.

What it handles

One search, a cited answer, scoped to you

InSearch searches across every connected app, checks your permissions, and writes one clear answer with citations back to the exact source documents.

  • Lists the Free, Basic, S1, S2, S3, L1 and L2 hourly rates and the storage each partition holds, tied to Microsoft sources
  • Converts the hourly rates into monthly figures for one, two and three replicas
  • Prices semantic ranker, image extraction and the new serverless preview, which most summaries skip
  • Works out a monthly bill for 200 employees of internal search and says which tier fits
  • Names the build work the invoice does not include, so the total cost is honest
Answer Cited

Written answer

Full-time employees get 20 days of PTO per year plus 10 company holidays1, requested in Workday with manager approval2.

Only sources you can access

CF People Handbook · Time Off [1]
DR PTO-request-process.pdf [2]
Drive · Slack · Confluence Never trains on your data

Why InSearch

One search across every app, with a cited answer

Not eight separate search boxes, not a wall of blue links. InSearch unifies your apps and returns one written answer with citations, scoped to exactly what you can see.

One search, every app

Drive, Slack, Notion, Confluence, Gmail, Jira and Salesforce searched in a single query, so you stop opening a different search box for each tool.

A cited answer

A written answer in plain language with inline citations back to the exact source docs, so you can trust it and verify it instead of reading a page of links.

Permission-aware

Google Drive, Slack and Gmail access carries over item by item and is enforced at query time. For Notion, Confluence, Jira, Salesforce and SharePoint the admin chooses at connection whether everyone or only the admin sees their content. It never trains on your data.

Compared

Azure AI Search tiers and published hourly rates side by side

Read from the Azure retail prices feed (East US, pay as you go) and learn.microsoft.com on November 16, 2026. US dollars. Monthly figures multiply the hourly rate by 730 hours for one search unit.

Tier Per unit per hour About per month Storage per partition Built for
Free $0 $0 50 MB, shared hardware Trying the API. No SLA, and an idle service can be deleted
Basic $0.101 $74 15 GB, up to 3 partitions and 3 replicas Small production indexes such as one company knowledge base
Standard S1 $0.336 $245 160 GB Mid sized indexes and private endpoint indexing
Standard S2 $1.344 $981 512 GB Large indexes and heavier query load
Standard S3 $2.688 $1,962 1,024 GB Very large indexes, plus a high density variant for many small indexes
Storage Optimized L1 and L2 $3.839 and $7.677 $2,802 and $5,604 2,048 GB and 4,096 GB Large, slowly changing corpora with lower query rates
InSearch Team / Business / Scale Per user $20 / $32 / $49 per user billed annually Not metered All connectors, permission-aware search and cited answers, no build

Storage figures are for services created after May 2024 in higher capacity regions such as East US, and older or other region services hold less. Semantic ranker ($1 per 1,000 queries), image extraction ($0.65 to $1.00 per 1,000 images), agentic retrieval and embedding model calls are billed on top. The serverless preview, billed since September 13, 2026, uses a $0.24 per hour compute unit and $0.20 per GB a month for storage, and has no SLA. Prices vary by region, so confirm yours in the Azure pricing calculator before you commit.

How much does Azure AI Search cost?

It costs from $0 on the Free tier to about $74 a month for one Basic search unit, and about $245 a month for one S1 unit. Multiply by the number of replicas and partitions you run, because every copy is billed as its own unit. Microsoft prices the service per search unit per hour, and a search unit is one replica multiplied by one partition. Two replicas and one partition is two units.

Three things then move the number. The tier sets the hourly rate and how much data a partition holds. The replica count sets how many queries you can serve at once and whether you meet the SLA. And optional features such as semantic ranker, image extraction from documents and agentic retrieval are billed by usage on top of the unit charge. Nothing here depends on how many people search, which is the opposite of a per-seat product.

A search unit is the billing block: one replica of one partition. A partition holds the index data and a replica is a full copy of the engine that answers queries. Search units equal replicas times partitions, so three replicas on two partitions is six units and six times the hourly rate. Basic allows up to three partitions and three replicas, nine units in total, on services created after April 3, 2024.

Partitions are how you add storage and replicas are how you add query capacity and resilience. Microsoft documents that the SLA requires two replicas for read-only workloads and three or more for read and write, so a production service rarely runs on one unit. That is why the sticker price of $74 for Basic is not the number to budget.

These are our estimates, not Microsoft quotes. Assume 200 employees, a 10 GB index of documents, messages and tickets, 20 searches a working day and 22 days a month, so 88,000 queries. 10 GB fits inside a 15 GB Basic partition, so one partition and three replicas, which meets the read and write SLA rule.

Three Basic units at $0.101 an hour for 730 hours is about $221 a month. If every query also runs the semantic ranker at $1 per 1,000 queries, that is about $88 more, and the free monthly allowance Microsoft gives lowers it slightly. The service bill is roughly $300 a month, or $3,700 a year. The same design on three S1 units is about $736 a month before semantic ranking.

The one-off and variable lines come on top: embedding calls to turn text into vectors, which you pay a model provider for, and image extraction if your documents are scanned. Against that, InSearch Business for 200 users at $32 per user billed annually is $76,800 a year. The two figures buy different things, which is the next section.

What is not in the Azure AI Search bill?

Azure AI Search indexes the documents you send it and answers queries. For employee search that leaves a long list on your side. A connector and sync job for every source, Slack, Google Drive, Confluence, Gmail, Jira, Salesforce, with deletions and renames handled. Microsoft supplies indexers for Azure sources and SharePoint in Microsoft 365, and for anything else you write the pipeline. Chunking, embedding and re-embedding when documents change. Per-user access control so a restricted document never reaches the wrong person, usually by storing rights on each record and filtering every query. And the answer step: a model that writes a reply with citations, which you pay for per token.

None of this is a criticism. It is what a search service is. It means the real cost of an Azure project is engineering time and maintenance, and only you can price that against your own team. If you assume a $150,000 fully loaded engineer, three months to build and a day a week to run it, the first year is roughly $68,000 in labor, on top of the $3,700.

It is the right choice when retrieval is a feature of a product you ship, when your data already lives in Azure storage, or when you have a platform team that wants control over ranking and a fixed hourly bill. The tiers are well documented, scale is large, and the entry price is low.

It is the wrong shape when the job is "let our 150 employees ask one question and get a cited answer from Slack, Drive and Confluence by Friday". That job is the whole InSearch product: connect the apps, permissions are enforced at query time, answers cite sources, and the price is per user. If you are weighing engines anyway, compare Algolia, Pinecone and Elastic before you commit, and read the Azure AI Search alternative for internal search breakdown. Microsoft sells a package for its own apps too, priced on Microsoft Copilot pricing, and Microsoft documents the tiers in its Azure AI Search tier guide.

Good questions

Questions about Azure AI Search pricing

Azure AI Search has a Free tier with 50 MB of storage on shared hardware. It has no SLA, cannot scale, and an unused service can be deleted. It suits trying the API. Any production index moves to Basic or higher, which starts at $0.101 per search unit per hour.
Basic is $0.101 per search unit per hour in the East US price feed, about $74 a month for one unit. It holds 15 GB per partition on newer services and allows up to three partitions and three replicas. Three replicas for the SLA is about $221 a month.
A search unit is one replica of one partition and is the unit Microsoft bills by the hour. Units equal replicas times partitions, so two replicas on three partitions is six units. Partitions add storage and replicas add query capacity and availability.
Semantic ranker is priced at $1 per 1,000 queries in the East US feed on the standard plan, after a monthly free allowance. On the free plan, requests fail with a billing error once the allowance is used. It needs Basic tier or higher for the standard plan.
It depends on whether you want a component or a product. Elasticsearch, Algolia, Pinecone and Azure AI Search are parts you build on. InSearch is a finished product that searches Drive, Slack, Notion, Confluence, Gmail, Jira and Salesforce with permission-aware, cited answers at a per-user price.

Explore more

More ways teams find answers with InSearch

Give your team one place to find every answer.

One search across all your company apps, a clear cited answer, scoped to exactly what each person can see, and never trained on your data.

See pricing

One search across every app · cited answers · permission-aware · never trains on your data