Index
Book a demo

Azure Machine Learning Pricing, Plans, and Availability

#2 in MLOps Platforms

by Microsoft · microsoft.com ↗

Microsoft’s machine learning platform for model lifecycle management, deployment, and monitoring.

#2MLOps PlatformsEnterprise
Visit website ↗

Pricing at a glance

◇ Pricing model

Usage-based cloud pricing with no publicly disclosed standalone Azure Machine Learning service fee; legacy Machine Learning Studio classic pricing includes a Free tier and quote-like/undisclosed paid tiers.

↻ Billing notes

Microsoft publicly offers Azure Machine Learning pricing through pay-as-you-go, Azure savings plan for compute, and reservations. The official page says pay-as-you-go is billed by the second, with no long-term commitments or upfront payments. It also says Azure savings plan for compute requires committing to spend a fixed hourly amount for 1 or 3 years, while reservations are available for one-year or three-year terms and can reduce cost versus pay-as-you-go. For the classic Machine Learning Studio offering, Microsoft says prices are estimates only and that actual pricing may vary by agreement, purchase date, and currency exchange rate. The classic page also includes a request-a-pricing-quote path and shows some public prices as not disclosed.

Plans and pricing

9 tiers

Pay as you go

Per second for compute capacity
Not publicly disclosed
  • No long-term commitments
  • No upfront payments

Compute capacity is billed by the second; usage increases or decreases on demand. Microsoft does not publish a standalone Azure Machine Learning subscription fee on the page.

Azure savings plan for compute

1 year or 3 years
Not publicly disclosed
  • Lower prices after committing to a fixed hourly spend

Applies to select compute services globally and is suited for dynamic workloads.

Reservations

1 year or 3 years
Not publicly disclosed
  • Significant cost reduction compared with pay-as-you-go

Described as Azure Reserved Virtual Machine Instances for stable, predictable workloads.

Machine Learning Studio (classic) Free

Not publicly disclosed
Free
  • Azure subscription not required
  • Max number of modules per experiment: 100
  • Max experiment duration: 1 hour per experiment
  • Max storage space: 10 GB
  • Single node execution/performance
  • No Production Web API

This tier is part of the classic Machine Learning Studio pricing page, not the current Azure Machine Learning service pricing page.

Machine Learning Studio (classic) Standard

Per ML studio workspace per month and per studio experimentation hour
Not publicly disclosed
  • Azure subscription required
  • Unlimited modules per experiment
  • Up to 7 days per experiment with a maximum of 24 hours per module
  • Unlimited storage - BYO
  • Read data from On-Premises SQL Preview: Yes
  • Multiple nodes execution/performance

Microsoft shows the Standard tier price fields as blank dashes on the public page, so the amount is not publicly disclosed.

Production Web API Dev/Test

Per month
Not publicly disclosed
  • 1,000 included transactions per month
  • 2 included compute hours per month
  • 2 total web services

Testing tier is limited to two concurrent RRS calls.

Production Web API Standard S1

Per month
Not publicly disclosed
  • 100,000 included transactions per month
  • 25 included compute hours per month
  • 10 total web services

Overage rates are listed, but the public page does not disclose the actual prices.

Production Web API Standard S2

Per month
Not publicly disclosed
  • 2,000,000 included transactions per month
  • 500 included compute hours per month
  • 100 total web services

Overage rates are listed, but the public page does not disclose the actual prices.

Production Web API Standard S3

Per month
Not publicly disclosed
  • 50,000,000 included transactions per month
  • 12,500 included compute hours per month
  • 500 total web services

Overage rates are listed, but the public page does not disclose the actual prices.

Add-ons and extras

Azure Blob Storage

Not publicly disclosed

Microsoft says Azure Machine Learning incurs separate charges for other Azure services consumed, including Azure Blob Storage.

Azure Key Vault

Not publicly disclosed

Microsoft says Azure Machine Learning incurs separate charges for other Azure services consumed, including Azure Key Vault.

Azure Container Registry

Not publicly disclosed

Microsoft says Azure Machine Learning incurs separate charges for other Azure services consumed, including Azure Container Registry.

Azure Application Insights

Not publicly disclosed

Microsoft says Azure Machine Learning incurs separate charges for other Azure services consumed, including Azure Application Insights.

Hidden and indirect costs

Worth budgeting for
⚠  Azure Machine Learning itself has no additional service charge, but the official pricing page says you will incur separate charges for compute and for other Azure services used alongside it. That means the real bill can include storage, secrets management, registry, monitoring, and the virtual machines or compute instances that actually run training and deployment workloads. Microsoft also notes that prices are estimates only and may vary by agreement, purchase date, and currency exchange rate, so the number a buyer sees can differ from the published table.
⚠  The classic Machine Learning Studio pricing page shows several price fields as dashes rather than published amounts. In practice, that means some tiers and overage meters are quote-like or calculator-based on the public page, so buyers should expect to confirm live rates in the Azure pricing calculator or with sales before budgeting. The page also states that hourly charges only apply to active use, which can make usage patterns a meaningful cost driver.

What buyers actually pay

A team wants to run a few development experiments and compare cost with a production-ready Azure Machine Learning setup.

Expected costNot publicly disclosed; compute and supporting Azure services are billed separately, and the amount depends on the instance type, region, and runtime.
Compute runtimeVM sizeStorageMonitoringRegistry and key management services

A buyer only needs the legacy Machine Learning Studio classic Free tier for basic experimentation.

Expected costFree
Experiment countStorage cap of 10 GBOne-hour experiment duration limit

A team publishes classic web services and needs a larger production tier.

Expected costNot publicly disclosed; the public page lists Dev/Test, Standard S1, S2, and S3 but hides the dollar amounts.
Monthly transactionsCompute hoursNumber of web servicesOverage usage

Pricing FAQ

Microsoft’s official pricing page says there is no additional charge to use Azure Machine Learning. Instead, buyers pay for compute capacity and for other Azure services they consume around the workload. That makes the platform more usage-based than seat-based or license-based.

The public pricing page lists pay-as-you-go, Azure savings plan for compute, and reservations. Pay-as-you-go is billed by the second with no long-term commitment or upfront payment. Savings plans and reservations are designed for buyers who can commit to longer usage horizons.

No. The classic pricing page shows a Free tier with published limits, but many Standard and web service price fields appear as dashes instead of dollar amounts. Microsoft also directs buyers to request a pricing quote and use the Azure pricing calculator, which suggests that some legacy pricing is not fully public.

↓Next: Alternatives