Microsoft Transitions to Consumption-Based Pricing for AI Copilot Cowork

By Patricia Miller

Jun 16, 2026

2 min read

Microsoft updates Copilot Cowork pricing to a consumption-based model, offering flexibility while exploring a new AI model as an alternative.

Microsoft is shifting its pricing model for Copilot Cowork, its autonomous AI solution integrated into Microsoft 365. The company is moving away from a flat per-user charge to a consumption-based system structured around what it calls Copilot Credits. This switch signals a strategic move to offer more flexibility in pricing based on actual usage rather than a fixed monthly fee.

Copilot Cowork debuted in March 2026 as a feature included in the standard Microsoft 365 Copilot add-on, which typically costs around $30 per user each month. Initially, users could access Copilot Cowork without incurring extra costs. The AI tool is designed to manage complex workflows using corporate data platforms like SharePoint and OneDrive, allowing it to execute tasks independently, thus minimizing the need for continuous human intervention.

The new credit-based model aligns with current consumption metrics reflected in other Microsoft products such as Copilot Studio and the Work IQ API services. For IT departments responsible for extensive Microsoft deployments, this update provides a method for tracking AI-related expenses more effectively throughout the corporate ecosystem.

Because Copilot Cowork is tailored to handle long-lived workflows, it can require varying amounts of computational resources. A flat subscription fee fails to accurately represent the costs associated with running an AI agent that might sift through substantial document libraries as opposed to one that quickly summarizes emails.

In addition to the pricing model change, Microsoft is exploring the incorporation of a Microsoft-hosted variant of DeepSeek, potentially named DeepSeek V4. This initiative aims to provide a cost-effective alternative to the Anthropic Claude models currently used by Copilot Cowork. DeepSeek’s AI models, available through Azure AI Foundry since early 2025, offer flexibility in deployment configurations, including both serverless and provisioned options for developers.

If Microsoft proceeds with this rollout, expected shortly after June 16, 2026, it would mark a significant step in integrating a Chinese-developed AI model into a mainstream productivity suite.

This change presents advantages and challenges for enterprises. Initially, companies can adopt Copilot Cowork at a lower cost, enabling gradual scaling based on its performance. However, the unpredictable costs associated with usage-based models raise crucial considerations, especially when appropriate controls are not implemented.

Introducing DeepSeek brings to the forefront issues of data sovereignty and security, primarily affecting organizations in sectors such as government contracting, finance, and healthcare, which operate under stringent regulations. Hosting DeepSeek on Microsoft’s infrastructure could alleviate some of these security concerns by ensuring data does not traverse Chinese servers.

By diversifying its AI model offerings, Microsoft enhances its bargaining position, preventing dependency on a single AI provider. This strategic approach positions Microsoft to cater effectively to enterprise needs while managing cost and performance expectations in the evolving AI landscape.

Important Notice And Disclaimer

This article does not provide any financial advice and is not a recommendation to deal in any securities or product. Investments may fall in value and an investor may lose some or all of their investment. Past performance is not an indicator of future performance.