From Flat Subscription to Usage-Based Pricing Model
Starting June 1, GitHub Copilot billing will move from a largely flat subscription to a usage-based pricing model tied to AI consumption. Copilot plans will still be sold at familiar entry points—USD 10 (approx. RM46) for Copilot Pro, USD 19 (approx. RM88) for Business, and USD 39 (approx. RM181) for Pro+ and Enterprise—but those figures now function more like an access fee than an all-inclusive bundle. The core change is that premium-model usage will be metered separately via GitHub AI Credits, turning Copilot into a service where costs scale with how heavily teams rely on advanced AI assistance. GitHub frames this shift as a response to rising inference costs and developer frustration with rigid limits, while also acknowledging that it has been absorbing significant AI overhead. In practice, developers will see a tighter link between their day-to-day AI coding assistant costs and how they actually use Copilot.
How GitHub AI Credits Work Inside Copilot
Under the new system, Premium Request Units (PRUs) are being retired and replaced with GitHub AI Credits, directly pegged to token usage. One AI Credit equals one cent of spend, and credits are consumed based on input, output, and cached tokens, which vary depending on the underlying model and the complexity of a query. Crucially, GitHub says code completions and Next Edit Suggestions remain unlimited for all paid plans and do not draw from the credit pool, so the heaviest metering will fall on premium models and more advanced, agent-like workflows. Fallback options to cheaper models, previously available when PRUs were exhausted, will disappear. Instead, once AI Credits run out—and any admin-defined budget limits are reached—access to premium capabilities can be throttled or cut off entirely. This design explicitly ties advanced AI behavior to incremental spend, while preserving a baseline of unmetered assistance for everyday coding.
Cost Implications for Heavy Users and Teams
For heavy Copilot users, the key implication is that AI coding assistant costs will no longer be capped at the monthly seat price. A developer or team that leans heavily on premium models, long-running agents, or frequent, complex queries can now push total spending well beyond the listed USD 10 (approx. RM46), USD 19 (approx. RM88), and USD 39 (approx. RM181) plan fees. GitHub has not yet published a detailed overage schedule, leaving procurement and engineering leaders guessing how sharply bills might climb once default credits are exhausted. That uncertainty complicates internal chargeback, budgeting approvals, and guardrail design. Managers will need faster visibility into credit consumption, clearer policies on when premium assistance is appropriate, and decisions about which workflows justify higher AI spend. Individual developers face similar trade-offs: experimentation becomes more deliberate when each intensive session may translate into tangible overage costs.
Comparing the New Billing Model with the Old One
Previously, Copilot felt closer to a traditional SaaS subscription: users paid a flat per-seat fee and received a bundle of entitlements, such as Premium Request Units, with clear plan-based limits. Pro+ already hinted at stratification, offering more than five times the capacity of Pro, but the billing logic still revolved around plan tiers rather than granular usage. With the shift to a usage-based pricing model, the subscription price acts as an entry ticket while AI Credits determine how far developers can push advanced features. There is no longer a simple guarantee that paying for a higher tier locks in predictable, all-inclusive access each month. Instead, real-world cost depends on behavior—how often developers invoke premium models, how long they run Copilot agents, and how aggressively they test AI-driven workflows. This blurs the line between licensing and operational expense, turning Copilot into a metered cloud-like service.
What Developers Should Do Before June 1
Ahead of the June 1 transition, teams should treat Copilot like any other metered cloud resource. That starts with understanding which workflows truly require premium models versus those that are well served by standard completions, which remain unmetered on paid plans. Engineering leaders may want to identify power users, estimate how often they lean on chat-style or agent-driven assistance, and set initial budget caps or org-wide guidelines around AI usage. Because GitHub has not provided a full overage schedule, conservative guardrails and close monitoring during the first months will be essential. Organizations should also prepare to adjust internal policies as they learn how quickly AI Credits burn down in practice. For individual developers, the new GitHub Copilot billing approach means being more intentional: reserving intensive AI sessions for high-value tasks and tracking whether the productivity gains justify potential increases in monthly spend.