Azure Machine Learning

Microsoft's managed platform for training, tracking and deploying models on Azure

As of 30 August 2026, Azure Machine Learning is free to use. Azure Machine Learning provides workspaces, scalable training compute, MLflow-compatible tracking, a model registry and managed endpoints. Softwr lists it under Machine Learning. Azure Machine Learning is made by Microsoft, launched in 1975, available on Web.

Visit siteVisit Azure Machine Learning

Overview

What Azure Machine Learning does

Azure Machine Learning is a managed service organised around a workspace, an Azure resource that holds datastores, compute targets, environments, jobs, registered models and endpoints. Training jobs are submitted to compute clusters that scale down to zero nodes between runs, runs are tracked through an MLflow-compatible tracking interface, models are registered with versions, and deployment goes either to managed online endpoints for real-time scoring or to batch endpoints for scheduled scoring. It also includes automated machine learning, a drag-and-drop designer, prompt flow for language model applications, and pipeline orchestration. The interfaces are a Python SDK, a command line extension and Studio, the browser interface. What distinguishes it is not modelling capability, it is that it is the machine learning service for an organisation that has already decided on Azure. Identity is Entra ID, networking is your virtual network with private endpoints, data sits in Azure storage, and the audit trail lands where the rest of the estate's does. That integration is why it gets chosen, because the realistic alternative is explaining a second cloud provider's security model to a review board. It follows that it is rarely evaluated on merit against SageMaker or Vertex AI; it is selected because the enterprise agreement already exists and the compliance work is already done. It is used by enterprise data teams in Microsoft shops, with a platform group maintaining the workspace and a data science group working inside it, and it does assume someone with Azure operational knowledge. The trade-off is that the workspace costs nothing and the compute costs everything, with managed online endpoints billing for their underlying virtual machines continuously and GPU capacity gated by regional quota you must request in advance. The tooling has also been through a full generational change, from the v1 SDK to a v2 SDK and command line with a different object model, so working v1 code is a migration project rather than an upgrade.

What people use it for

Enterprises standardised on Azure where using a different cloud for machine learning would mean a fresh security and compliance review
Training that needs to burst onto a GPU cluster occasionally without buying hardware, with the cluster scaling back to zero afterwards
Regulated workloads that must stay inside a virtual network with private endpoints and auditable role-based access
Teams already using MLflow who want the tracking interface they know backed by a managed service and enterprise identity

Where it falls short

Concrete and checkable, so you can decide whether any of them matter to you. This is the half of a review a vendor will not write about Azure Machine Learning.

Managed online endpoints are billed per underlying virtual machine for as long as the deployment exists, with no scale to zero, so a model answering a handful of requests a day costs the same as one answering thousands.
GPU capacity is governed by per-region, per-family quota that must be requested and approved, so a training plan can be blocked by an administrative ticket rather than by budget, and the newest accelerators are often unavailable in the region your data is required to stay in.
The v2 Python SDK and command line use a different object model from v1 and code, pipelines and examples written for v1 do not port mechanically, which has left teams maintaining two ways of doing the same thing and searching documentation that mixes both.
The workspace binds storage, key vault, container registry and compute together, so recreating or moving one is not a light operation, and configuring it properly with private endpoints and a managed virtual network is a multi-day job for somebody who already knows Azure networking.
Experiment history, registered models, environments, endpoints and pipeline definitions live inside the workspace, and although the tracking interface is MLflow-compatible, moving the accumulated lineage and orchestration elsewhere is a rebuild, so the cost of leaving grows every month the team uses it.

Cross-shopped

What people choose instead of Azure Machine Learning

Each pairing was judged by two reviewers asking whether a buyer would genuinely weigh the two against each other. The ones that failed were deleted rather than published.

Azure Machine Learningvs
AWS SageMaker
AWS SageMaker: AWS's comprehensive ML service with similar end-to-end lifecycle support and model deployment
Azure Machine Learningvs
Google Vertex AI
Google Vertex AI: Google Cloud's unified ML platform with AutoML and extensive pre-built models
Azure Machine Learningvs
DataRobot
DataRobot: Enterprise-focused AutoML platform with automated model training and deployment

Pricing

What Azure Machine Learning costs

Taken from the vendor's own pricing page. Prices move, so check before you buy.

Free Tier

Free

Limited compute
Basic features

Pay-as-you-go

$0.05 /hour

Full platform
All compute options
Enterprise features

Capabilities

Features

Workspace
An Azure resource tying together storage, key vault, container registry, compute and all ML assets
Compute clusters
Managed training clusters that scale from zero, with low-priority nodes as a cheaper interruptible option
MLflow-compatible tracking
Run, parameter, metric and artefact logging through the MLflow client interface
Model registry
Versioned model registration with lineage back to the job that produced it
Managed online endpoints
Real-time scoring endpoints with blue-green deployment and traffic splitting between versions
Batch endpoints
Scheduled or triggered scoring over large data sets without a persistent serving cost
Automated machine learning
Model search across algorithms and preprocessing for tabular, text and vision problems
Pipelines
Multi-step training and processing workflows with component reuse and caching
Prompt flow
Authoring, evaluation and deployment of language model application flows
Enterprise networking and identity
Entra ID, role-based access control, private endpoints and managed virtual network isolation

Answered, with sources

Questions people ask

Each answer names the page it came from, so you can check it rather than take our word for it.

Is there a charge for the workspace itself?

No charge for the workspace resource. You pay for the compute it runs, the storage it uses, the container registry, key vault and any endpoints left running, which is where essentially the whole bill comes from.

Does it work with MLflow?

Yes. The tracking interface is MLflow-compatible, so existing logging code generally works unchanged, and that compatibility is the least locked-in part of the platform.

What is the difference between SDK v1 and v2?

A different object model and a different way of expressing jobs, components and endpoints. v2 is the current one. v1 code does not translate mechanically and a lot of material found online still assumes v1, which is a common source of wasted time.

Do endpoints scale to zero?

Managed online endpoints do not; they hold their virtual machines. Batch endpoints only consume compute while a job runs, so intermittent workloads are much cheaper served as batch where the use case allows it.

Do I need an ML engineer to run it?

For the data science work, not necessarily. For the workspace itself, yes, somebody has to understand Azure identity, networking, quota and cost management, and on teams without that person the platform becomes the bottleneck rather than the model.

Behind it

Who makes Azure Machine Learning

Company: Microsoft
Based in: Redmond, Washington

Timeline