Plan and manage costs of an Azure AI Search service

This article explains how Azure AI Search is billed, including fixed and variable costs, and provides guidance for cost management.

Before you create a search service, use the Azure pricing calculator to estimate costs based on your planned capacity and features. Another resource is a capacity-planning worksheet that models your expected index size, indexing throughput, and indexing costs.

As your search workload evolves, follow our tips to minimize costs during both deployment and operation. You can also use built-in metrics to monitor query requests and Cost Management to create budgets, alerts, and data exports.

Understand the billing model

Azure AI Search has both fixed and pay-as-you-go billing. You pay a fixed rate for your search service as long as it exists, while premium features are billed according to your usage.

Costs for Azure AI Search are only a portion of the monthly costs in your Azure bill. Although this article focuses on planning and managing Azure AI Search costs, you're billed for all Azure services and resources used in your Azure subscription, including non-Microsoft services.

How you're charged for the base service

When you create or use search resources, you're charged for the minimum required replica and partition combination (R × P) at the prorated hourly rate of your pricing tier. As your search units increase or decrease, so do your costs. For more information and an example of the billing model, see Billing rates.

How you're charged for premium features

Premium features are charged in addition to the base cost of your search service. The following table lists premium features and their billing units. All of these features are optional, so if you don't use them, you don't incur any charges.

¹ Refers to images extracted from a file within the indexer pipeline. Text extraction is free. Image extraction is billed when you enable the indexAction parameter or when you call the Document Extraction skill.

² Charges for Azure OpenAI models and Azure AI Foundry models appear on your bill for those services.

How you're otherwise charged

Depending on your configuration and usage, the following charges might apply:

• Data traffic might incur networking costs. See the bandwidth pricing.

• Several premium features, such as knowledge stores, debug sessions, and enrichment caches, depend on Azure Storage and incur storage costs. Charges for these features appear on your Azure Storage bill.

• Customer-managed keys, which provide double encryption of sensitive content, require a billable Azure Key Vault.

• A skillset can include billable built-in skills, nonbillable built-in utility skills, and custom skills. Nonbillable utility skills include Conditional, Shaper, Text Merge, and Text Split. They don't have an API key requirement or 20-document limit.

• A custom skill is functionality you provide. Custom skills are billable only if they call other billable services. They don't have an API key requirement or 20-document limit.

Estimate and plan costs

Use the Azure pricing calculator to estimate your baseline costs for Azure AI Search. You can also find estimated costs and tier comparisons on the Select Pricing Tier page during service creation.

For initial testing, we recommend that you create a capacity-planning worksheet. The worksheet helps you understand the index-to-source ratio and the effect of enrichment or vector features on both capacity and cost.

To create a capacity-planning worksheet:

1. Index a small sample (1–5%) of your data. Include any OCR, enrichment, or embedding skills you plan to use.

2. Measure the index size, indexing throughput, and indexing costs.

3. Extrapolate the results to estimate the full-scale requirements for your data.

Minimize costs

To minimize the costs of your Azure AI Search solution, use the following strategies:

Deployment and configuration

• Create a search service in a region with more storage per partition.

• Create all related Azure resources in the same region (or as few regions as possible) to minimize or eliminate bandwidth charges.

• Choose the lightest pricing tier that meets your needs. Basic and S1 offer full access to the modern API at the lowest hourly rate per SU.

• Use Azure Web Apps for your front-end application to keep requests and responses within the data center boundary.

Scaling

• Add partitions only when the index size or ingestion throughput requires it.

• Add replicas only when your queries per second increase, when complex queries are throttling your service, or when high availability is required.

• Scale up for resource-intensive operations, such as indexing, and then readjust downwards for regular query workloads.

• Write code to automate scaling for predictable workload patterns.

• Remember that capacity and pricing aren't linear. Doubling capacity more than doubles costs on the same tier. For better performance at a similar price, consider switching to a higher tier.

Indexing and enrichment

• Use incremental indexing to process only new or changed data.

• Use enrichment caching and a knowledge store to reuse previously enriched content. Although caching incurs a storage charge, it lowers the cumulative cost of AI enrichment.

• Keep vector payloads compact. For vector search, see the vector compression best practices.

Monitor costs

At the service level, you can monitor built-in metrics for your queries per second (QPS), search latency, throttled queries, and index size. You can then create an Azure Monitor dashboard that overlays QPS, latency, and cost data to determine when to add or remove replicas.

At the subscription or resource group level, Cost Management provides tools to track, analyze, and control costs. You can use Cost Management to:

• Create budgets that define and track progress against spending limits. For more granular monitoring, customize your budgets using filters for specific Azure resources or services. Filters prevent you from accidentally creating resources that incur extra costs.

• Create alerts that automatically notify stakeholders of spending anomalies or overspending risks. Alerts are based on spending compared to budget and cost thresholds. Both budgets and alerts are created for subscriptions and resource groups, making them useful for monitoring overall costs.

• Export cost data to a storage account. This is helpful when you or others need to perform more cost analysis. For example, a finance team can analyze the data using Excel or Power BI. You can export your costs on a daily, weekly, or monthly schedule and set a custom date range. Exporting cost data is the recommended method for retrieving cost datasets.

FAQ

Can I temporarily shut down a search service to save on costs?

Search runs as a continuous service. Dedicated resources are always operational and allocated for your exclusive use for the lifetime of your service. To stop billing entirely, you must delete the service. Deleting a service is permanent and also deletes its associated data.

Can I change the billing rate (tier) of an existing search service?

Existing services can be switched between Basic and Standard (S1, S2, and S3) tiers. Currently, you can only switch from a lower tier to a higher tier, such as going from Basic to S1. For more information, see Change your pricing tier.

Related content

• Azure AI Search pricing
• Choose a pricing tier for Azure AI Search
• Optimize your cloud investment with Cost Management
• Quickstart: Start using Cost analysis