EU-resident replacement for Azure OpenAI Whisper before its 15.12.2026 retirement?

For an EU-resident production replacement today, migrate the workload to regional Azure Speech to text in Sweden Central—using real-time transcription for live calls, or fast/batch transcription for recordings—rather than relying on an unannounced EU deployment of gpt-transcribe.

Current status of gpt-transcribe

• Microsoft’s current availability matrix does not list gpt-transcribe under Data Zone Standard in Europe. Its listed availability is Global Standard, which can process inference in any Azure region and therefore does not meet an EU-only processing requirement.^1^
• There is no published commitment or date for Data Zone EU or regional availability. Microsoft documents the general launch order as Global → Data Zone → geography-based/regional, but explicitly says geography-based availability has no guaranteed date and depends on capacity.^2^
• Customers cannot infer a future EU release from Global availability. Continue checking the model availability matrix, but treat any future deployment as eligible only when it explicitly appears under Data Zone Standard—Europe or a suitable Standard/Regional table.

A support request can clarify roadmap or subscription visibility, but it cannot establish a compliance commitment that is absent from the published availability matrix.

Microsoft’s replacement position

The retirement schedule currently lists:

• whisper version 001: December 15, 2026
• Replacement: not yet named
• gpt-4o-transcribe version 2025-03-20: October 15, 2026, not February 26, 2027
• gpt-4o-mini-transcribe version 2025-12-15: June 15, 2027^3^

The blank replacement entry is expected at this stage: Microsoft says it normally selects and publishes an official replacement approximately 90–120 days before retirement. It also advises customers not to wait for that designation before evaluating alternatives.^4^

Although the transcription quickstart identifies gpt-transcribe for offline file transcription, that is a functional recommendation—not an EU-residency guarantee.^5^

EU-resident production alternative

Azure Speech stores and processes speech data in the region where the Speech resource is created, except for specifically documented exceptions such as Voice Live. Sweden Central supports regional:

• Real-time transcription
• Fast transcription
• Batch transcription^6^

For your workload:

1. Create an Azure Speech resource in Sweden Central.
2. For live phone audio, use the Speech SDK or real-time Speech-to-text API.
3. For completed call recordings:
   • Use fast transcription when a synchronous result with predictable latency is required.
   • Use batch transcription for larger volumes and asynchronous processing.
4. Start with the default Azure Transcribe model. Microsoft describes it as the general-purpose option with diarization, phrase lists, language identification, and timestamps. Fast transcription can also use Azure LLM Speech or MAI-Transcribe-2, but availability and production status must be checked independently.^7^
5. Benchmark it against representative calls for word error rate, speaker separation, multilingual/code-switching behavior, latency, and cost before cutover.

Azure Speech is specifically recommended for phone-call recordings and analytics, as well as real-time call-center transcription.^8^

For recordings, keep both source audio and transcript output in EU/EEA storage. Batch transcription lets the customer choose and control those storage locations and retention; real-time audio is processed in server memory without storage at rest.^9^

Important residency distinction

• Azure Speech in Sweden Central: regional processing, suitable when Sweden/EU location is required.
• Data Zone EU: processing can occur anywhere within the EU Data Boundary, which can include EFTA locations such as Switzerland and Norway.
• Global Standard: processing may occur in any Azure region and should not be used where inference must remain in the EU/EEA.^2^

Because your calls include Swiss residents, confirm with your privacy/legal team whether the requirement is strictly EU, EU Data Boundary/EEA, or specifically Sweden. The regional Azure Speech option provides the narrowest documented processing scope among these choices.

References

1. Region availability for Foundry Models sold by Azure
2. Understanding deployment types in Microsoft Foundry Models - Microsoft ...
3. Model retirement schedule - Microsoft Foundry | Microsoft Learn
4. Microsoft Foundry Models lifecycle and support policy
5. Speech to text with transcription models - Microsoft Foundry
6. Supported regions for Azure Speech
7. Use the fast transcription API with Azure Speech in Foundry Tools (programming-language-csharp)
8. What are OpenAI transcription models?
9. Data and Privacy for Speech to text