Microsoft delays default usage-based billing for Copilot to 2027

A shift in AI economics

Microsoft is moving toward a pricing model where users pay for what they actually use rather than just a flat fee. This shift is known as usage-based billing (UBB), which treats AI interactions like a utility where costs are tied to consumption. The company is currently navigating how to balance these costs with existing license structures.

The company recently announced it is delaying the move to make UBB the default for new Microsoft 365 Copilot Business subscriptions. This change will now take effect between January and March 2027, which is the company's third fiscal quarter. This marks the second time the company has pushed back the effective date of this specific policy.

The decision comes as the company looks to expand its total addressable market. By combining seat subscriptions with consumption charges, the company aims to increase its average revenue per user. This model allows for a basic allotment of compute under a license while charging metered fees for more advanced AI uses.

Partner feedback and preparation

The delay was prompted by feedback from the company's roughly 500,000 partners worldwide. These solution providers are currently trying to figure out which AI uses are covered by a standard license and which ones trigger extra costs. They need to ensure that customers do not receive unexpected bills at the end of the month.

Microsoft stated they are adjusting dates to give partners extra time to prepare. This extra time allows providers to help clients stay within their budgets. Many providers are already helping customers determine which specific AI capabilities will trigger consumption charges.

To assist with this transition, Microsoft is offering 30 days of sandbox testing before the policy goes live. A sandbox is a testing environment where users can try out software without affecting live systems. The company is also providing readiness guidance to help partners prepare for the new billing mechanics.

Managing the spending limits

Even with the move toward usage-based billing, Microsoft is maintaining some guardrails for its partners. Solution providers will still have a default spending limit of 4,000 Copilot Credits per user per month across eligible services. This limit helps prevent runaway costs for organizations that are just starting to adopt the technology.

Partners who want to avoid any variable costs have the option to set the spending limit to zero. This allows them to stop usage-based spending entirely for their clients. This flexibility is important for companies that prefer the predictability of a fixed license over a consumption-based model.

Specific offerings like Copilot Cowork, GitHub Copilot Harness, and certain APIs will require usage-based billing. These tools are part of a broader trend where spending is tied directly to the value that AI agents deliver. As these agents take on more complex tasks, the company expects demand for these credits to grow.

The growth of AI consumption

The move toward UBB is not just a policy change but a reflection of how AI usage is growing. During a recent earnings call, CEO Satya Nadella noted that the number of Copilot customers actively consuming credits grew more than 75 percent over the last quarter. This indicates that more businesses are moving beyond basic chat and into high-volume AI tasks.

The company is also seeing significant growth in other areas, such as agentic features on GitHub business. Credit consumption for these specific features has more than doubled recently. This trend suggests that as AI becomes more capable of performing autonomous work, the demand for underlying compute power will continue to rise.

The company's strategy involves helping clients see where they might be wasting licenses. Some providers suggest that if only a small portion of a workforce uses a tool, a consumption model might be more efficient. By moving to a metered system, companies can potentially save money by only paying for the users who are actually performing high-value work.

Global rollout and next steps

The new timelines for usage-based billing and sandbox availability apply to a wide range of international markets. These include countries such as Australia, Brazil, France, Germany, India, Italy, Korea, the Netherlands, Poland, and Spain. This ensures that global partners have a synchronized timeline for preparing their customers for the change.

While the company has delayed the default UBB policy, it has not yet revealed a specific date for the availability of the Partner Center sandbox environments. Partners will need to keep an eye on official channels for these specific dates as they approach the 2027 deadline. The goal remains to provide a smooth transition for the thousands of providers who manage these accounts.

Moving forward, the focus for both Microsoft and its partners will be on transparency. Providers are already differentiating themselves by helping clients understand their AI consumption bills. As the market moves toward token-based pricing, the ability to monitor and cap spending will be a primary concern for IT decision-makers.

How it works

Usage-based billing (UBB) is a pricing model where the cost of a service is determined by how much of it is actually used. In the context of AI, this often refers to "compute" or "tokens," which are the units of processing power required to generate text, images, or perform data analysis. Instead of paying a flat monthly fee for unlimited use, customers pay for the specific volume of work the AI completes.

Microsoft is currently transitioning from a purely license-based model to a hybrid model. In this setup, a company pays a base fee for a seat (the license) which covers standard use. When a user performs more advanced tasks or uses specific high-demand tools, they draw from a pool of credits or are charged extra based on the amount of data processed. This is similar to how cloud storage works, where you pay for the space you occupy.

For partners, this means they must now act as consultants to help clients forecast their AI spending. They need to identify which users will be "power users" who might consume many credits and which users will only need basic features. By monitoring these consumption patterns, partners can help companies avoid sudden budget overruns while still taking advantage of the most advanced AI capabilities available in the Microsoft ecosystem.

Who is affected

The primary group affected by this change is Cloud Solution Providers (CSPs). These are the third-party companies that sell and manage Microsoft services for other businesses. With about 500,000 partners worldwide, these providers are responsible for explaining the new billing mechanics to their clients and ensuring that the costs remain manageable for the end users.

The end customers—businesses of all sizes—are also directly impacted. They will see a shift in how they are billed for Microsoft 365 Copilot Business licenses purchased through these partners. While the delay gives them time to plan, they will eventually need to account for variable costs in their IT budgets rather than relying solely on fixed annual or monthly costs.

Additionally, the change affects specific software products within the Microsoft ecosystem. Tools like Copilot Cowork and GitHub Copilot Harness are specifically highlighted as requiring usage-based billing. This means teams using these specific developer and productivity tools should expect to see their costs fluctuate based on the volume of work their teams perform with these AI systems.

What comes next

The next major milestone for this policy is the window between January and March 2027. During this period, Microsoft expects to turn on usage-based billing by default for new subscriptions. Until then, the company will continue to provide readiness guidance and the promised 30-day sandbox testing period to help partners prepare for the transition.

In the meantime, Microsoft will continue to monitor the growth of its consumption-based AI business. The company is already seeing high growth rates in credit consumption, particularly for agentic features. This data will likely inform how they refine their pricing tiers and what specific features they choose to put behind consumption-based walls in the future.

For the broader tech industry, this move signals a shift toward a more standardized way of pricing AI. As more companies move away from unlimited access, the ability to monitor and cap spending will be a primary concern for IT decision-makers.