Microsoft Foundry passes 120,000 customers as revenue more than triples
Microsoft official logo (public domain, Wikimedia Commons) — CryptoBriefing brand treatment
Microsoft's AI agent platform is growing fast, and open-source models are doing a surprising share of the lifting
Microsoft Foundry now counts more than 120,000 customers. Over the same stretch, its revenue more than tripled year over year.
The growth details are arguably more interesting than the headline count. Multi-model agent builders on Foundry grew fivefold. Revenue from open-source models on the platform jumped tenfold year over year.
What the numbers show
Foundry is Microsoft’s unified Azure platform for building AI agents and applications.
Foundry surpassed 100,000 customers as of July 2026. The updated figure of 120,000 means another 20,000 organizations have joined since that milestone.
Microsoft’s figures show Foundry revenue more than tripled year over year.
The number of customers building agents that use multiple AI models grew fivefold.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Revenue tied to open-source models on Foundry grew tenfold year over year.
Deeper adoption, not just more logos
The number of Foundry customers processing at a run rate of one trillion tokens grew 4x year over year.
More than 17,000 customers are using both Foundry and Microsoft Fabric, the company’s data platform. That figure is up 60% year over year.
A Forrester study found a 327% return on investment over three years for organizations using Foundry. The study attributed the gains mainly to improvements in developer productivity.
How Foundry got here
Foundry reached general availability for core enterprise scenarios in 2026.
Microsoft has positioned the platform around governance and security. The goal is simpler management of AI deployments and less model lock-in, meaning customers are not trapped with one provider’s models if something better comes along.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.

