Microsoft Copilot pricing explained: What businesses need to know before buying AI

Before the June 2026 pricing changes, Copilot usage was measured in premium request units (PRUs). PRUs tracked users’ engagement with advanced Copilot AI features and did not apply to standard code completion or Next Edit Suggestions (Github Docs, 2026).

Under this model, each subscription tier included a limited number of PRUs per month. Once that limit was reached, users were automatically shifted to a cheaper Copilot model with reduced capabilities, rather than being charged additional usage fees. This was known as a “fallback experience” (Rodriguez, 2026).

This structure made costs more predictable for many users because subscriptions came pre-built with fixed usage allowances. However, it also limited access to more advanced AI functionality after those monthly thresholds were exceeded. 

On June 1, 2026, billing for Copilot products transitioned to a usage-based system powered by GitHub AI Credits, the new standard unit of measurement for AI usage (Rodriguez, 2026). 

To understand how this new pricing structure works, it helps to first know how AI models process and store information using tokens. 

When you submit a prompt to an AI model, the system breaks the content down into smaller units of data called tokens (input tokens) (Microsoft Copilot Blog, 2026). The AI’s response is also measured in tokens (output tokens), along with any context it reuses or stores (cached tokens). Under the new pricing model, token usage is translated into AI credits, which are deducted from an organization’s or user’s available monthly credit balance. The number of credits consumed depends on the AI model being used and the amount of processing required. More advanced or resource-intensive models tend to consume more credits because they need more tokens. Simpler AI models consume fewer credits (Github Docs, 2026). 

Microsoft has yet to establish a straightforward conversion between tokens and dollars. This means businesses may have difficulty estimating the costs of specific Copilot functions. It can be helpful for Copilot users to start thinking about their AI interactions differently, by token rather than by request (Desai, 2026).  

Subscribers receive a monthly allotment of AI credits based on the dollar amount they pay for their subscription tier. The current conversion rate is $0.01 USD = 1 GitHub AI Credit (as of June 1, 2026). So, a $10-per-month individual subscription will come with 1,000 AI credits each month (Kunder, 2026). 

Things look a little different for Business and Enterprise plans, where AI credits will be pooled across an organization rather than assigned to individual users. That means a company with 100 Copilot Business licenses (billed at $19/month per user as of June 1, 2026) will have all users share from a collective bucket of 190,000 AI credits per month. This model allows for variation among teams, since some employees may rely heavily on Copilot while others use it only occasionally (Github Docs, 2026).

Once an individual or organization exhausts its monthly credits, additional AI usage will require purchasing more credits or upgrading the plan tier. Unlike the previous pricing model, there will no longer be a fallback experience that automatically shifts users to a lower-tier AI model (Rodriguez, 2026). 

AI is everywhere right now, but one subtopic that has been under more recent scrutiny is its profitability. Many, if not all, AI companies are still trying to determine sustainable pricing models for products that require enormous computing resources.

As AI companies attempt to generate profit while still scaling rapidly, pricing structures across the industry will continue to evolve. Free and subscription-based models are giving way to hybrid approaches that combine licensing with usage-based billing.

At the same time, AI technology has become dramatically more advanced in recent years. Newer models require significantly more computing power, cloud infrastructure, and energy consumption than earlier chatbot-style tools (Rodriguez, 2026). 

$0/month

A basic chatbot available via web browser or mobile app, the free version of Copilot can help with general productivity tasks, like summarizing articles and drafting emails. It’s also included in Word, Excel, PowerPoint, and Outlook for eligible Microsoft 365 accounts.  

$10/month per user

Designed for individual consumers, CoPilot Pro offers an enhanced chatbot experience, plus full integration of AI capabilities into all Microsoft Office apps. 

$39/month per user

Copilot Pro+ provides access to premium AI models and advanced AI features for individual users. Under the old model, this subscription tier came with unlimited PRUs. After June 1, it provides subscribers with 3,900 AI credits each month.

Note: Sign-ups for Copilot Pro and Copilot Pro+ was paused until the new usage-based pricing model went into effect on June 1 (Ahmad, 2025).

$30/month per user

Microsoft 365 Copilot allows individuals to integrate AI across their Microsoft 365 applications such as Word, Excel, Outlook, Teams, and PowerPoint. 

$19/month per user

Specifically built for small and mid-sized businesses, Microsoft 365 Copilot Business can implement AI capabilities across Microsoft 365 applications for up to 300 users. It’s a more affordable option than the Enterprise subscription tier, but it only works as an add-on to existing Microsoft 365 Business licenses.  

$39/month per user

Microsoft 365 Copilot Enterprise provides AI functionalities across Microsoft 365 applications with additional compliance, governance, and security capabilities for larger organizations.  

Starting at $200/month

Allows organizations to build custom AI agents, workflows, and automation experiences tailored to business processes (Microsoft Copilot Blog, 2025). 

Before purchasing Copilot licenses or AI add-ons, organizations should establish a clear AI strategy centered around specific needs and measurable outcomes. When small businesses adopt AI with intention, it can absolutely improve productivity and efficiency. It just requires ongoing oversight and initiatives to keep costs under control (Byteiota, 2026). 

To help customers prepare for the updated pricing model, GitHub introduced new billing reports and usage preview features that provide subscribers with a detailed breakdown of their monthly AI consumption. Reviewing historical usage data can help organizations identify which teams, tools, and workflows are generating the most value.

Under usage-based pricing models, monitoring the consumption of your monthly AI credits is crucial. Organizations can reduce the risk of unexpected costs by setting monthly spending budgets and configuring automated alerts when usage approaches predefined thresholds. While reaching those limits may temporarily restrict access to premium AI capabilities, these safeguards can help businesses maintain predictable budgeting and avoid large surprise invoices.

Many organizations may discover they are not using AI as extensively or as strategically as they initially expected. As pricing models evolve, it’s worth evaluating whether AI is delivering meaningful value for every workflow where it has been deployed. In some cases, the cost of using AI for routine or low-impact tasks will be too high under the new usage-based pricing model. Manual processes may be a more practical and cost-effective option.

Agentic AI with certain automations in place can quickly rack up hidden costs because agents may continuously run in the background without direct human oversight. Without guardrails in place, these automations can consume AI credits around the clock.

As AI capabilities mature, organizations should become more intentional about where and how they use these tools. If employees have grown accustomed to relying on AI for every small task, now is a good time to establish clearer usage guidelines. Encouraging teams to reserve AI for high-impact work can help maximize return on investment while keeping costs under control.

Businesses that adopt AI without a clear strategy may find themselves paying for licenses and usage that provide little measurable value. Proactively managing these consumption-based pricing changes will put businesses in a better position to control costs, forecast budgets accurately, and realize the full value of their AI investments.

With Microsoft’s shift toward consumption-based AI pricing, understanding how Copilot subscriptions, AI credits, and tokens work is now more important than ever. And while some specifics are still developing, it’s a good idea to stay apprised of any updates and perhaps more reserved in your AI usage until we understand the full scope of pricing.