Azure vs AWS: who is winning AI workloads in 2025

TL;DR:

AWS leads overall cloud share, per Canalys and Synergy.
Azure is growing faster and wins marquee AI ties like OpenAI.
AWS offers the broadest model menu on Bedrock and custom chips.
Azure leans on OpenAI, adds Mistral and more in its model catalog.
Your winner depends on workload, data gravity, and tooling

Who is winning depends on what you measure. By revenue share, AWS leads. By AI headline workloads and Microsoft 365 integration, Azure has an edge. Growth favors Azure.

Market share. Canalys pegs AWS at roughly one-third of the market in early 2025, with Azure in the low-20s and growing faster. Synergy Research shows a similar pattern.
Growth. Canalys notes Azure and Google Cloud grew above 30 percent year over year in Q2 2025, while AWS grew about 17 percent.
Customer momentum. ETR survey data shows stronger spending velocity for Azure than AWS in July 2025. Treat surveys as sentiment, not revenue.

Model catalogs and ecosystems

AWS Bedrock

Bedrock supplies a large model menu through one API, including Anthropic Claude, Meta Llama, Mistral, AI21, Cohere, Amazon Nova and Titan, and more. This favors teams that want several providers with one contract and governance model.

Azure AI Foundry and Azure OpenAI

Azure features OpenAI models through Azure OpenAI and a broad catalog that includes models from Mistral, Meta, Cohere, and others. Microsoft also offers its own small models such as Phi. The catalog approach helps when you want OpenAI plus alternates in one portal.

What’s changing now. On September 11, 2025, OpenAI and Microsoft signed a new MOU to shape the next phase of their partnership. Final terms are pending, so watch for any hosting or exclusivity shifts that could change where the largest training runs land.

Chips and performance

AWS

AWS fields first-party silicon for AI. Trainium2 is generally available and aims for better price-performance than prior GPU instances. AWS also advanced agentic tooling in Bedrock and Q. This mix can lower unit costs at scale.

Azure

Azure leans on NVIDIA’s latest parts and shows benchmark gains with large H200 clusters in MLPerf, plus expanding Grace Blackwell support. For many teams, the question is not peak speed alone, but when capacity is available in your region.

Signature workloads

OpenAI on Azure. Microsoft says OpenAI’s API runs on Azure and is also available through Azure OpenAI Service. That has pulled high-end training and inference onto Azure. The new MOU could adjust these lines, so track the definitive agreement.
Anthropic on AWS. AWS positions Claude as a flagship Bedrock model and highlights growing agent features. Many enterprises standardize on Claude for long-context tasks and safety features.

Where each wins today

Azure advantages

Tight tie-ins to Microsoft 365, Teams, SharePoint, and Power Platform.
Easy path to ship Copilot features to Office users and secure them with Entra ID and Purview.
Access to OpenAI models in the same cloud that hosts your Microsoft data.

AWS advantages

Breadth of models in Bedrock, plus managed agents and guardrails.
Custom silicon that can cut training and inference costs at scale.
Deep ecosystem in data lakes, eventing, and serverless that many AI apps need.

Decision guide: pick by workload, not headlines

Use this quick table to decide.

Goal	Likely fit	Why
Build with many model options, swap fast	AWS	Bedrock’s catalog and guardrails across providers.
Standardize on OpenAI with Microsoft data	Azure	Azure OpenAI plus Microsoft 365 and security stack.
Lowest training cost at large scale	AWS	Trainium2 focus on price-performance, growing availability.
Ship Copilot into Office quickly	Azure	Native hooks to Word, Excel, Outlook, Teams.
Mix Anthropic with others under one roof	AWS	Claude family is first-class on Bedrock.

What to watch next

Definitive Microsoft–OpenAI terms. Hosting rights and model access could move workloads.
GPU and Trainium supply. Capacity dictates where training lands this quarter.
Market share updates. Track Canalys and Synergy into Q3 and Q4 2025.

Why it matters

If you measure by total cloud revenue, AWS is still on top. If you measure by Microsoft 365 integration and OpenAI alignment, Azure has a real edge. Most companies will be multicloud for AI to balance price, speed, and model choice.

Sources: