Microsoft Foundry Pricing and Reservation

Question

I am currently trying to understand the pricing structure of Microsoft Foundry and how reservations apply to it. I have a few questions:

The Azure pricing documentation provides pricing details for all Foundry models: https://azure.microsoft.com/en-gb/pricing/details/ai-foundry-models/aoai/

However, in the Azure Pricing Calculator (https://azure.microsoft.com/en-in/pricing/calculator/), I can only find Azure OpenAI, and not the other models. Are the other models listed under a different product name?

Based on my current understanding, only OpenAI, Microsoft, and DeepSeek models can currently be covered by reservations. Are there any other models that are eligible for reservations?

1. Does Azure Advisor provide reservation recommendations for Microsoft Foundry? If so, can these reservations be retrieved using the Reservation Recommendations API? https://learn.microsoft.com/en-us/rest/api/consumption/reservation-recommendations/list?view=rest-consumption-2024-08-01&tabs=HTTP At the moment, I do not see Azure Foundry or Azure OpenAI listed as a supported resourceType in the documentation.

Answer 1

Hello Ananya,

Thank you for your patience while we clarify how pricing, reservations, and recommendations work for Azure AI Foundry today. I’ll address each of your questions below.

1. Why you only see “Azure OpenAI” in the Pricing Calculator

Currently, the Azure Pricing Calculator only provides a first-class entry for Azure OpenAI Service.

Many models available through Azure AI Foundry (including Microsoft-branded models or partner models such as DeepSeek) are not yet surfaced as separate tiles in the calculator UI. This means they may appear on the Foundry pricing documentation page, but not as individual selectable products in the calculator.

In practice:

• Pricing for these models is published on the Azure AI Foundry pricing page (token-based PAYG or PTU-based pricing).
• To estimate costs today, you can reference the per-token or per-PTU rates from the pricing page and:
  • Use the calculator’s Custom cost entries, or
  • Estimate usage manually until the models appear in the calculator.
• It is also common for new or preview models to appear in documentation before they are added to the calculator UI, so manual estimation is the typical workaround in the meantime.

2. Which Foundry models support reservations

Reservations apply only to Provisioned Throughput Unit (PTU) deployments, not to token-based pay-as-you-go usage.

Currently, reservation eligibility includes deployments such as:

Azure OpenAI Service models deployed using PTU (provisioned throughput)

Microsoft Foundry provisioned throughput capacity used for supported Foundry models or tools

Some partner model deployments (for example DeepSeek) when they expose a PTU-based SKU

Workloads that are strictly PAYG/token-based or fixed-rate service SKUs (for example transaction-based AI services) are not eligible for reservations.

At this time, OpenAI models, Microsoft-hosted models, and DeepSeek engines are the primary model families that can be covered by PTU reservations. There are currently no additional model lines beyond these that support reservation coverage.

3. Azure Advisor and Reservation Recommendations API

Regarding reservation recommendations:

Azure Advisor does not currently generate reservation recommendations for Azure OpenAI or Azure AI Foundry PTU usage.

Similarly, the Reservation Recommendations REST API does not list resource types such as:

Microsoft.Foundry

  `Microsoft.CognitiveServices/accounts/openai`
  

Because these resource types are not currently supported, reservation recommendations for these services cannot yet be retrieved via the Advisor API.

Support for Advisor-driven PTU reservation guidance is planned but not yet available.

In the meantime, the recommended approach is to:

Track usage through Azure Cost Analysis

Export usage data and estimate long-term PTU needs

Purchase reservations manually through the Azure portal under All Services → Reservations → Microsoft Foundry Provisioned Throughput

The pricing calculator currently exposes Foundry usage primarily through Azure OpenAI, while other Foundry models may not yet appear as separate calculator entries.

Reservations only apply to PTU-based deployments, not token-based PAYG workloads.

Azure Advisor and the Reservation Recommendations API do not yet provide reservation guidance for Foundry/OpenAI resources, so reservations must currently be evaluated and purchased manually.

Please refer this

Buy Foundry PTU reservation: https://learn.microsoft.com/azure/cost-management-billing/reservations/microsoft-foundry

I Hope this helps. Do let me know if you have any further queries.

---

If this answers your query, please do click Accept Answer and Yes for was this answer helpful.

Thank you!

Answer 2

Hello Ananya ,
Azure AI Foundry does not appear as a separate item in the Azure Pricing Calculator because it functions as a control and orchestration layer rather than a billable service; costs are charged through the underlying Azure AI services it uses, such as Azure OpenAI Service, Azure AI Search, Azure Vision, Azure Speech, and Azure Document Intelligence. Model reservations in Foundry are available through Provisioned Throughput Units (PTUs) and currently support only OpenAI, Microsoft Phi, and DeepSeek models, while third-party models remain pay-as-you-go.
Additionally, reservation recommendations for these services are not provided by Azure Advisor or the Reservation Recommendations API, meaning PTU reservations must be purchased and managed manually through the Azure Portal or Foundry interface and monitored using Azure Cost Management tools.

Reference Document-

1. https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/microsoft/
   2.https://learn.microsoft.com/en-us/azure/cost-management-billing/reservations/reserved-instance-purchase-recommendations
   3.https://docs.azure.cn/en-us/cost-management-billing/dataset-schema/reservation-recommendations-mca

Answer 3

Microsoft Foundry pricing and reservations are tied to Azure AI services and Provisioned Throughput Units (PTUs) for models available in Microsoft Foundry Models.

1. Pricing visibility in the Azure Pricing Calculator
   Foundry resources are billed through the underlying Azure AI services they use (for example, Azure OpenAI, Azure Speech in Foundry, Azure Vision in Foundry, etc.). Costs are calculated per service and feature, not as a single “Foundry” line item:

• When a Foundry resource is created, charges accrue for the Azure services it uses, such as Azure OpenAI, Azure Speech in Foundry, Content Safety, Azure Vision in Foundry, Azure Document Intelligence, and Azure Language in Foundry.
• Two billing models exist: pay‑as‑you‑go (serverless API) and commitment tiers.

Because of this, services may appear in the pricing calculator under their individual product names (for example, Azure OpenAI) rather than a generic “Foundry” product. The context does not list alternative calculator product names for each Foundry model beyond the underlying services mentioned.

2. Which models can be covered by reservations
   Reservations apply to Microsoft Foundry Provisioned Throughput for “models available in Microsoft Foundry Models.” Key points:

• A Microsoft Foundry Provisioned Throughput reservation covers PTU usage for models available in Microsoft Foundry Models when deployed as Global, Data Zone, or Regional Provisioned deployments.
• The reservation discount applies automatically to provisioned throughput pay‑as‑you‑go meters for those models, within the reservation’s scope and region.
• Reservations for Global, Data Zone, and Regional deployments are not interchangeable; a separate reservation is required for each deployment type.

The context explicitly calls out Azure OpenAI Service and DeepSeek as examples of “models available in Microsoft Foundry Models,” but does not enumerate all other eligible models. It states that the reservation mechanism “works on models available in Microsoft Foundry Models (e.g. Azure OpenAI Service, DeepSeek, etc.)” and refers to the Microsoft Foundry Models pricing page for the full list. No additional specific model families beyond these examples are listed in the provided material.

3. How Microsoft Foundry Provisioned Throughput reservations work

• Reservations are purchased for a specific region, quantity of PTUs, and deployment type (Global, Data Zone, or Regional).
• The reservation discount is applied hourly by comparing reserved PTUs to deployed PTUs for eligible models in scope:
  • Deployed PTUs up to the reserved amount are covered by the reservation.
  • Any deployed PTUs above the reserved amount are billed at the hourly pay‑as‑you‑go rate.
  • Partial‑hour deployments are pro‑rated by minutes within the hour.
• Reservations apply only to provisioned deployments and do not cover standard (serverless) deployments or fine‑tuning.
• Reservations do not guarantee capacity; best practice is to create deployments first, then buy reservations sized to those deployments.
• If deployments are deleted, the reservation is not automatically changed or canceled; it must be canceled or exchanged manually via Azure Reservations.

4. Buying and scoping reservations

• Reservations are purchased in the Azure portal under Microsoft Foundry Provisioned Throughput.
• Supported subscription types include Enterprise Agreement, Microsoft Customer Agreement, and pay‑as‑you‑go.
• Scopes can be:
  • Single resource group
  • Single subscription
  • Shared scope across eligible subscriptions in the same billing context
  • Management group (subscriptions in the management group and billing scope)
• Reservation size should match the total PTUs deployed for each deployment type (Global, Data Zone, Regional). Reservations are purchased in one‑PTU increments.

5. Azure Advisor and Reservation Recommendations API
   The provided context describes how reservations work for Azure AI Foundry Provisioned Throughput and how to purchase and scope them, but it does not state that Azure Advisor provides reservation recommendations specifically for Microsoft Foundry or Azure OpenAI, nor does it document support for these services in the Reservation Recommendations API. The context also does not list any resourceType values or mappings for Foundry/OpenAI in that API.

Based on the available material, only the general reservation behavior for Azure AI Foundry Provisioned Throughput is documented; there is no explicit confirmation here that Azure Advisor or the Reservation Recommendations API currently surface recommendations for Microsoft Foundry or Azure OpenAI PTU reservations.

---

References:

• Save costs with Microsoft Foundry Provisioned Throughput Reservations
• Plan and manage costs for Microsoft Foundry
• Understanding costs associated with provisioned throughput units (PTU)