Azure Machine Learning SLA Credits & Refunds

Microsoft publishes a separate SLA document for nearly every Azure AI/ML service, and Machine Learning is no exception. The catch: the credit you're owed depends on configuration choices (zone redundancy, replication tier, pricing model) most teams forget to verify after an incident. Here's what to check for Machine Learning specifically.

What this guide covers

• The official Azure Machine Learning uptime commitment and credit tiers
• Which incidents qualify (and which exclusions silently disqualify claims)
• How to file a Machine Learning credit request inside the Azure claim window
• Why manual claim recovery typically leaves money on the table

Frequently asked questions about Azure Machine Learning SLAs

What is the typical SLA uptime guarantee for Azure Machine Learning?

Azure guarantees 99.9% uptime for Managed Online Endpoints used to serve real-time inference, and 99.9% for the workspace management plane. Training compute, pipelines, and batch endpoints are best-effort and do not carry an availability SLA. If Azure fails to meet the inference commitment during a billing cycle, you are eligible to receive a portion of your Machine Learning spend back as a service credit.

How do I claim Azure Machine Learning SLA credits after an outage?

Submit a billing support request through the Azure portal: Help + Support → New support request → Issue type: Billing → Problem type: Service credit request. Within two months of the billing period in question, provide the affected Subscription ID and Resource ID, the start and end timestamps of the impacted period, your evidence (Azure Monitor logs, Resource Health alerts, or independent monitoring), and your calculated Monthly Uptime Percentage for Machine Learning. Microsoft validates against its internal incident records before issuing the credit to your billing account.

What exclusions apply to the Azure Machine Learning SLA?

Critically, training jobs, automated ML runs, pipeline executions, and customer-managed compute clusters fall outside the SLA — only managed online endpoints are covered for availability.

Why is it difficult to get refunds for Machine Learning outages manually?

AI/ML SLAs are still maturing, and Machine Learning carries some of the most nuanced terms in the cloud catalog. Rate limits, queue depths, and model availability all get measured differently, and the SLA often excludes throttling that the provider deems "expected." Teams that successfully claim Machine Learning credits do so by capturing per-request latency and error-code data and matching it precisely against the published terms.

Related Azure SLA guides

Other Azure services creditable through the same portal-based billing request process:

• Azure Azure OpenAI SLA credits — AI/ML
• Azure Cognitive Services SLA credits — AI/ML
• Azure Virtual Machines SLA credits — Compute
• Azure Blob Storage SLA credits — Storage

Recover Azure credits without a portal grind

Azure billing support requests for Machine Learning aren't difficult to file — they're tedious. Each one takes the same kind of subscription-ID, resource-ID, timestamp, and uptime-calculation packaging, repeated for every incident across every subscription you own.

Next Signal detects Machine Learning SLA breaches across your Azure tenants, packages the credit request in the format Microsoft expects, and submits it. See how it works or start a free trial.