Sells banks consumption-priced access to many foundation models, agent hosting and governance tooling inside the bank's own Azure subscription, under Microsoft's data protection terms; banks also reach it through software vendors such as Fiserv that build on it.
What is Microsoft Foundry, and which banks use it?
Microsoft Foundry is Microsoft's Azure platform for building AI apps and agents on models from OpenAI, Anthropic, Meta and others; Microsoft called it Azure AI Studio, then Azure AI Foundry from November 2024, and Microsoft Foundry by November 2025. Among the banks we track, Northern Trust's client-onboarding agent uses it, BNY's Eliza platform incorporates it, and Ally built call summaries on Azure OpenAI Service. Pricing is consumption-based: Microsoft publishes per-token prices by model and deployment type and says actual prices vary by agreement.
Category
Founded
2024
Headquarters
Redmond, Washington, US
Ownership
Product of Microsoft (NASDAQ: MSFT), which reported $92.3 billion of revenue and 48% Azure revenue growth for the quarter ended September 30, 2026; Microsoft holds roughly 27% of OpenAI Group, whose models it sells in Foundry, and in April 2026 gave up the exclusive right to sell them.[11][12][13]
Deployment
API and platform service (cloud) on Azure, with Global, Data Zone (EU or US) and Regional deployment types
Pricing
Consumption-based: each Foundry service has its own billing model; models are billed per million input and output tokens by model and deployment type (Global, Data Zone, Regional), with batch and hourly provisioned-throughput options
Banks among the 120
3
- The name has changed twice: Azure AI Studio became generally available in May 2024, Microsoft introduced Azure AI Foundry in November 2024, and its November 18, 2025 Ignite announcements use the name Microsoft Foundry; its documentation maps 'Azure AI Studio / Azure AI Foundry' to 'Microsoft Foundry' and 'Azure AI Services' to 'Foundry Tools'.
- Azure OpenAI Service, generally available since January 2023, is now sold as Azure OpenAI in Foundry Models; Microsoft's documentation tells Azure OpenAI customers they can upgrade the resource to a Foundry resource and keep the endpoint and API keys.
- Microsoft's pricing page lists OpenAI's GPT-6 Sol (short context) at $2 per million input tokens and $10 per million output tokens on a Global deployment, and $2.20 and $11 on a Data Zone deployment (read November 10, 2026); it calls the figures estimates that vary with the customer's agreement.
- Microsoft says Foundry is used by developers at more than 80,000 enterprises and digital natives, including 80% of Fortune 500 companies; its product page counts over 11,000 models in the catalogue, while its documentation says over 1,900.
- Named bank use among the banks we track: Northern Trust's client-onboarding agent (announced November 9, 2026), BNY's Eliza platform and an accounting research capability, and Ally's call summaries, built on Azure OpenAI Service in 2023.
- Outside the US banks counted here, Microsoft says UBS's Red assistants, built on Azure OpenAI Service and Azure AI Search, reached 30,000 employees by December 2024 in its Switzerland, Hong Kong and Singapore booking centres; that Commerzbank's Ava assistant, built with Foundry Agent Service, reportedly resolves about 75% of customer conversations autonomously; and that Commonwealth Bank's orchestration agent, built with Copilot Studio and Microsoft Foundry, sits behind more than two million conversations a month.
- Microsoft no longer has the exclusive right to sell OpenAI's models: Reuters reported on April 27, 2026 that the two renegotiated their pact, with Microsoft remaining OpenAI's primary cloud partner and holding a licence to its intellectual property through 2032. Anthropic's Claude models arrived in Foundry in November 2025 and became generally available there in June 2026.
What does Microsoft Foundry sell to banks?
| Product | What it does | Status |
|---|---|---|
| Foundry Models[1][9] | The model catalogue and inference service: models from OpenAI, Anthropic, Meta, Google, xAI, Hugging Face and others, with benchmarking and a model router. | Generally available |
| Azure OpenAI in Foundry Models (formerly Azure OpenAI Service)[5][3] | OpenAI's models hosted by Microsoft on Azure, billed per million tokens or as provisioned throughput, on Global, Data Zone or Regional deployments. | Generally available |
| Foundry Agent Service[8][1] | Hosting and orchestration for AI agents; generally available since May 2025 under the name Azure AI Foundry Agent Service, after a public preview from November 2024. | Generally available |
| Claude in Microsoft Foundry[10][9] | Anthropic's Claude models hosted on Azure and bought through an existing Azure account; first available in November 2025, generally available from June 29, 2026. | Generally available |
| Foundry IQ[9][1] | Retrieval and grounding for agents, described by Microsoft as the evolution of Azure AI Search; announced in public preview on November 18, 2025. | Beta |
| Foundry Control Plane[9][1] | One portal for identity, policy, observability and security signals across agents, integrated with Microsoft Entra, Purview and Defender; announced in public preview on November 18, 2025. | Beta |
How much does Microsoft Foundry cost?
How it is charged: Consumption-based: each Foundry service has its own billing model; models are billed per million input and output tokens by model and deployment type (Global, Data Zone, Regional), with batch and hourly provisioned-throughput options.
| Product | Plan | List price | Unit | Checked |
|---|---|---|---|---|
| Azure OpenAI in Foundry Models | GPT-6 Sol (short context), Global deployment[3] | $2 input / $10 output | per 1M tokens (cached input $0.20); Data Zone deployment $2.20 / $11 | Nov 10, 2026 |
| Azure OpenAI in Foundry Models | GPT-6 Astra (short context), Global deployment[3] | $10 input / $50 output | per 1M tokens (cached input $1); Data Zone deployment $11 / $55 | Nov 10, 2026 |
| Azure OpenAI in Foundry Models | Provisioned throughput, GPT-6 Sol (short context), Global[3] | $1 per PTU per hour | minimum 15 PTUs; $260 per PTU with a monthly reservation, $2,652 with a yearly reservation | Nov 10, 2026 |
Microsoft publishes US-dollar list prices per model on its Azure OpenAI pricing page; the figures above were read on November 10, 2026. The Batch API is priced at a 50% discount to Global Standard pricing, and provisioned deployments are charged by the hour per provisioned throughput unit (PTU) whether or not they are used. Microsoft states that the prices are estimates only, not quotes, and that actual pricing varies with the type of agreement, the purchase date and the exchange rate. Foundry's other services (Agent Service, Foundry IQ, Foundry Tools, Control Plane) are priced separately; those prices and negotiated enterprise discounts are not listed here.
Which banks use Microsoft Foundry?
| Bank | What | Status |
|---|---|---|
| Northern Trust | Client onboarding agent for Asset Servicing's Front Office Solutions platform, built with Microsoft engineers on Azure and using Microsoft Foundry; the bank and Microsoft report data preparation time down 83%, from two and a half weeks to three days, with analysts reviewing the output (announced November 9, 2026).[15] | In production |
| BNY (Bank of New York Mellon) | BNY's in-house Eliza platform incorporates Microsoft Foundry alongside Azure, Microsoft 365 Copilot and GitHub Copilot; BNY is also building an AI research capability in its Data & Analytics Accounting platform with Azure OpenAI in Foundry Models, per two Microsoft customer stories.[16][17] | In production |
| Ally Financial | Call summaries for customer-service associates, built on Azure and Azure OpenAI Service, with the first use case targeted within eight weeks; more than 700 associates see summaries of tens of thousands of calls a week, per Microsoft's customer story.[18] | In production |
Among the 120 largest US banks on this site, from sources that name the bank: the bank, Microsoft Foundry or the press. Status is the furthest stage a source confirms.
What is Microsoft Foundry?
A unified Azure platform-as-a-service for enterprise AI operations, model builders and application development, in Microsoft's words: Foundry Models (including Azure OpenAI), Foundry Agent Service, Foundry IQ, Foundry Tools, Foundry Control Plane and Foundry Local under one Azure resource, with role-based access control, networking and policy managed in one place.[2][1]
What has Microsoft Foundry done, and when?
- Nov 9, 2026
Northern Trust onboarding agent announced[15]
Northern Trust and Microsoft announced a client-onboarding agent that uses Microsoft Foundry, reporting data preparation time cut from two and a half weeks to three days and a 96.5% F1 first-pass quality score.
- Oct 28, 2026
Azure revenue up 48%[11]
Microsoft reported revenue of $92.3 billion for the quarter ended September 30, 2026, with Microsoft Cloud revenue of $63.9 billion and Azure revenue up 48%.
- Jun 29, 2026
Claude generally available in Foundry[10]
Microsoft said Claude in Microsoft Foundry was generally available, hosted on Azure and bought through a customer's existing Azure account.
- Apr 27, 2026
Microsoft and OpenAI end exclusivity[13]
Reuters reported that Microsoft and OpenAI renegotiated the pact that let Microsoft exclusively sell OpenAI's models; Microsoft stays OpenAI's primary cloud partner with an intellectual-property licence through 2032.
- Apr 20, 2026
Trade-finance proof of concept with ANZ, HSBC and Lloyds[22]
Microsoft described a trade-finance prototype developed with ANZ, HSBC and Lloyds and built on Microsoft Foundry; it is a proof of concept, not a production system.
- Feb 26, 2026
Microsoft cites Commerzbank's Ava[20]
A Microsoft banking post said Commerzbank's Ava assistant, built with Foundry Agent Service, reportedly resolves about 75% of customer conversations autonomously.
- Jan 8, 2026
Fiserv expands its use of Foundry[23]
Fiserv, a technology vendor to banks, said it would expand its use of Microsoft Foundry and had processed more than 100 billion tokens in it.
- Nov 18, 2025
Announced as Microsoft Foundry; Claude models arrive[9]
Microsoft's Ignite announcements presented the platform as Microsoft Foundry, added Anthropic's Claude Sonnet 4.5, Opus 4.1 and Haiku 4.5, made its model router generally available and put Foundry IQ and Foundry Control Plane into public preview.
- Oct 29, 2025
Azure outage lasts more than eight hours[14]
Microsoft said an outage of its Azure cloud platform, tied to its Azure Front Door network, lasted more than eight hours and affected Microsoft 365 and customers including Alaska Airlines, Reuters reported.
- May 19, 2025
Agent Service generally available[8]
Azure AI Foundry Agent Service became generally available; Microsoft said more than 10,000 customers had used it since its public preview in November 2024.
- Nov 19, 2024
Azure AI Foundry introduced[7]
Microsoft introduced Azure AI Foundry, said Azure AI Studio would become the management console and portal for it, and reported that Azure OpenAI Service consumption had more than doubled in six months.
- May 21, 2024
Azure AI Studio generally available[6]
Microsoft said Azure AI Studio, its pro-code platform for building generative-AI applications, was generally available.
- Dec 21, 2023
Ally's call summaries on Azure OpenAI Service[18]
Microsoft's customer story describes Ally's pilot on Azure OpenAI Service: more than 700 associates with summaries of tens of thousands of customer-service calls a week.
- Jan 17, 2023
Azure OpenAI Service generally available[5]
Microsoft announced general availability of Azure OpenAI Service, giving businesses access to OpenAI models including GPT-3.5, Codex and DALL-E 2 on Azure.
What do banks use Microsoft Foundry for?
| Use | What it does | Status | Use case |
|---|---|---|---|
| Client-onboarding data agent[15] | Northern Trust's agent converts new clients' historical transaction data into the bank's standard format, with analysts reviewing outputs and resolving exceptions. | In production | Generative & agentic AI |
| In-house AI platform[16] | BNY's Eliza platform, which Microsoft says holds more than 300 solutions, incorporates Microsoft Foundry alongside Azure, Microsoft 365 Copilot and GitHub Copilot. | In production | Generative & agentic AI |
| Contact-centre call summaries[18] | Ally's customer-service associates receive AI-written summaries of calls, which they can edit, built on Azure OpenAI Service. | In production | Customer-facing chatbots |
| Choice of where prompts are processed[4][3] | Regional deployments process prompts within the chosen geography; Data Zone deployments stay within the EU or the US; Global deployments may process them in any geography where the model is deployed. | In production | Data & privacy |
| Bank vendors building on Foundry[23] | Fiserv says it uses Microsoft Foundry in its products, client servicing and developer gateway, so a bank can depend on Foundry through a vendor without contracting for it. | In production | Third-party & vendor AI |
What has Microsoft Foundry disclosed in numbers?
| Metric | Value | As of |
|---|---|---|
| Enterprises using Foundry (company-reported) | More than 80,000 enterprises and digital natives, including 80% of Fortune 500 companies[1] | Nov 10, 2026 |
| Models in the catalogue (company-reported; Microsoft's pages differ) | Over 11,000 on the product page; over 1,900 in the documentation[1][2] | Nov 10, 2026 |
| Azure revenue growth | 48% year over year in the quarter ended September 30, 2026[11] | Sep 30, 2026 |
| Northern Trust onboarding data preparation (bank- and vendor-reported) | Down 83%, from two and a half weeks to three days; 96.5% F1 first-pass quality score[15] | Nov 9, 2026 |
| BNY Eliza (as reported by Microsoft) | More than 300 solutions; client settlement inquiries resolved 80% faster[16] | Nov 1, 2026 |
| Ally call summaries (as reported by Microsoft) | More than 700 associates; tens of thousands of calls a week[18] | Dec 21, 2023 |
| Fiserv tokens processed in Foundry (company-reported) | More than 100 billion[23] | Jan 8, 2026 |
| UBS Red assistants on Azure OpenAI Service (as reported by Microsoft; Switzerland, Hong Kong and Singapore booking centres, not the US bank) | Rolled out to 30,000 employees[19] | Dec 31, 2024 |
| Commonwealth Bank customer-service conversations on a Copilot Studio, Dynamics 365 and Foundry build (as reported by Microsoft; not a US bank) | More than two million a month[21] | Jul 9, 2026 |
“When we started to build out Eliza, we selected the Microsoft Azure cloud foundation, then added Microsoft Foundry as part of the solution.”[16]
Sarthak Pattanaik · Chief Data and AI Officer, BNY · Nov 1, 2026
“The solution reduces the time required to prepare and validate clients’ data while maintaining the oversight and controls our clients expect, creating capacity for both sides to focus on higher-value work.”[15]
Pete Cherecwich · Chief Operating Officer, Northern Trust · Nov 9, 2026
“The massive investment Microsoft has made in OpenAI helped move things forward, but the key thing is that Microsoft really understands our need to implement robust, reliable, and secure capabilities,”[18]
Scott Richardson · Chief Information Officer of Enterprise Data and Analytics, Ally · Dec 21, 2023
Who runs Microsoft Foundry?
Which rules apply to a bank using Microsoft Foundry?
| Authority | Why it matters here | Documents |
|---|---|---|
| OCC | A bank building on Foundry depends on a chain of Microsoft's Azure, a model provider such as OpenAI or Anthropic and often a software vendor such as Fiserv; the 2023 interagency third-party guidance expects banks to understand subcontractor and concentration risk, including cloud outages like the eight-hour Azure incident of October 2025. | OCC Bulletin 2023-17 FDIC FIL-29-2023 |
| Federal Reserve | SR 26-2 covers traditional and machine-learning models but leaves generative and agentic AI outside its scope, so agents built on Foundry, such as an onboarding agent whose output analysts review, need governance through the bank's broader risk framework. | SR 26-2 |
| NIST | NIST's Generative AI Profile lists value-chain and third-party model risks that apply when one platform serves models from several providers; Microsoft's documented data commitments are written for 'models sold by Azure', so a bank has to check which catalogue models carry that designation. | NIST AI 600-1 (Generative AI Profile) |
Linked authorities have a page in the Regulation tracker; the others are named for completeness.
What should a bank note before buying Microsoft Foundry?
Analysis by BankingNewsAI from the sources cited on this page. Observations, not advice or a recommendation.
01
Cost follows usage and deployment type
Models are billed per token, or per hour for provisioned throughput, and the same model costs about 10% more on a Data Zone deployment than on a Global one in Microsoft's list prices.
02
List prices are estimates
Microsoft says the published figures are not quotes and that actual pricing depends on the type of agreement, the purchase date and the exchange rate.
03
Three names in two years
Contracts, documentation and job postings may say Azure AI Studio, Azure AI Foundry, Azure OpenAI Service or Microsoft Foundry for parts of the same platform.
04
Where prompts are processed is a deployment choice
Microsoft's documentation says Global deployments may process prompts in any geography where the model is deployed, while Data Zone and Regional deployments narrow that.
05
Most named results are vendor-reported
The figures for Northern Trust, BNY, Ally, UBS and Commerzbank come from Microsoft's releases and customer stories, some of them jointly with the bank; none is independently audited.
06
Microsoft's own pages disagree on the catalogue size
The product page says over 11,000 models and the documentation says over 1,900; the number of models a bank can use depends on which carry Microsoft's terms and are deployed in its regions.
What are the alternatives to Microsoft Foundry?
Amazon Bedrock (AWS)
AWS service giving one API to foundation models from Amazon, Anthropic, OpenAI, Meta and others, plus AgentCore for running agents.
ChatGPT Enterprise (OpenAI)
OpenAI's workplace ChatGPT with admin controls and no training on business data by default, plus API access to GPT models.
Claude (Anthropic)
Anthropic's Claude models and workplace plans (Team, Enterprise), sold direct and through AWS and Microsoft.
Gemini Enterprise (Google Cloud)
Google Cloud's workplace AI app and agent platform on Gemini models, connecting to Google Workspace, Microsoft 365 and business data.
Microsoft 365 Copilot
AI assistant inside Word, Excel, Outlook and Teams, grounded in a firm's Microsoft 365 data, plus Copilot Studio for building agents.
GitHub Copilot
AI coding assistant and coding agent in the IDE, CLI and GitHub.com, with Business and Enterprise plans billed per seat plus AI credits.
How much does Microsoft Foundry cost a bank?
There is no single price: each Foundry service is billed on consumption. For models, Microsoft's pricing page on November 10, 2026 listed OpenAI's GPT-6 Sol (short context) at $2 per million input tokens and $10 per million output tokens on a Global deployment, and $2.20 and $11 on a Data Zone deployment. Provisioned throughput for the same model was $1 per unit per hour with a 15-unit minimum. Microsoft says these are estimates and that actual prices vary by agreement.
Which banks use Microsoft Foundry or Azure OpenAI?
Among the US banks we track, three are named in sources we could read: Northern Trust, whose client-onboarding agent uses Microsoft Foundry (announced November 9, 2026); BNY, whose Eliza platform incorporates it; and Ally, which built call summaries on Azure OpenAI Service in 2023. Outside that list, Microsoft names UBS for assistants built on Azure OpenAI Service in Switzerland, Hong Kong and Singapore, Commerzbank for an assistant on Foundry Agent Service and Commonwealth Bank for a customer-service orchestration agent.
Is Microsoft Foundry the same as Azure AI Foundry?
Yes. Microsoft's documentation lists 'Azure AI Studio / Azure AI Foundry' as the previous brand and 'Microsoft Foundry' as the current one. Azure AI Studio became generally available in May 2024, Microsoft introduced the Azure AI Foundry name in November 2024 and presented the platform as Microsoft Foundry in November 2025. Azure AI Services is now called Foundry Tools, and Azure OpenAI Service is sold as Azure OpenAI in Foundry Models.
Who owns Microsoft Foundry?
Microsoft (NASDAQ: MSFT), which runs it as part of Azure. Microsoft reported revenue of $92.3 billion for the quarter ended September 30, 2026, with Azure revenue up 48%. It holds roughly 27% of OpenAI Group, according to OpenAI, and remains OpenAI's primary cloud partner, but since April 2026 it no longer has the exclusive right to sell OpenAI's models, Reuters reported.
Does Microsoft use a bank's prompts to train models?
Microsoft's documentation for models sold by Azure, which include the Azure OpenAI models, says prompts, completions, embeddings and training data are not available to other customers or to OpenAI and are not used to train generative-AI foundation models without the customer's permission. It also says the models are stateless. Prompts are checked in real time for harmful content, and the commitments are written for models designated as sold by Azure.
Where is a bank's data processed on Foundry?
It depends on the deployment type. Microsoft's documentation says prompts and responses are processed within the customer-specified geography for standard deployments. Its pricing page describes Data Zone deployments as geography-based, covering the EU or the US, and Regional deployments as local to one of up to 27 regions. Global deployments may process prompts in any geography where the model is deployed, at lower list prices.
How does Microsoft Foundry compare with Amazon Bedrock?
Both sell consumption-priced access to foundation models from several providers inside a cloud. Until April 2026 Microsoft had the exclusive right to sell OpenAI's models; Reuters reported on April 27, 2026 that the pact had been renegotiated and that OpenAI had by then launched on Amazon's cloud. Microsoft's November 2025 statement that Azure was the only cloud offering both OpenAI and Anthropic models predates that change. This site's Amazon Bedrock page lists Bedrock's published prices and named bank customers for a side-by-side reading.
- Microsoft Foundry (undated page, retrieved 2026-11-10)
- What is Microsoft Foundry? (undated page, retrieved 2026-11-10)
- Azure OpenAI Service - Pricing (undated page, retrieved 2026-11-10)
- Data, privacy, and security for Foundry Models sold by Azure in Microsoft Foundry (undated page, retrieved 2026-11-10)
- General availability of Azure OpenAI Service expands access to large, advanced AI models with added enterprise benefits
- From code to production: New ways Azure helps you build transformational AI experiences
- The next wave of Azure innovation: Azure AI Foundry, intelligent data, and more
- Announcing General Availability of Azure AI Foundry Agent Service
- Microsoft Foundry: Scale innovation on a modular, interoperable, and secure agent stack
- Claude in Microsoft Foundry is now generally available
- Microsoft announces first quarter results
- Our Structure (undated page, retrieved 2026-11-10)
- Microsoft, OpenAI change terms of deal so startup can court Amazon and others
- Microsoft Azure's services restored after global outage
- Northern Trust and Microsoft Frontier Company collaborate to transform client onboarding through AI
- Frontier Firm BNY resolves client inquires 80% faster with Microsoft AI powered Eliza (undated page, retrieved 2026-11-01)
- BNY uses Azure OpenAI in Foundry Models to scale trust in accounting operations (undated page, retrieved 2026-11-10)
- Ally Financial empowers customer service associates to focus on human engagement by using Azure OpenAI Service
- UBS and Microsoft unite: co-creating the future of banking with Azure AI (undated; figures as of December 2024)
- Agentic AI in Banking: A Blueprint for Modern Customer Experience
- How Commonwealth Bank and Microsoft are reimagining the future of customer service
- Reimagining trade finance with AI: a collaborative proof of concept from Microsoft, ANZ, HSBC, and Lloyds
- Fiserv Collaborates with Microsoft to Accelerate AI-Driven Innovation
When a bank signs with Microsoft Foundry, you'll read it here first.
when a bank picks an AI vendor, or a vendor ships for banks, it is in the next morning's brief · six sourced stories · 7 am PT · free
plus every regulation, bank and agent page update, the morning after · leave any morning