Copilot just 9x’d Sonnet and 27x’d Opus and teams have no idea

Understanding the Shifts in AI Service Pricing: What Teams Need to Know

In recent weeks, a subtle yet significant change in the AI industry’s pricing model has gone largely unnoticed—until now. GitHub’s quietly updated multiplier table for its Copilot subscription plans signals a shift toward more accurate, sustainable pricing models for large language models (LLMs). This development marks a pivotal moment in the ongoing evolution of AI service economics.

Contextual Overview

For organizations leveraging AI tools like GitHub Copilot, understanding the underlying billing structure is crucial. Copilot offers users a monthly allotment of “premium requests,” which determine how much they can utilize advanced AI models. Previously, the rate at which this quota depletes was governed by multipliers assigned to different models. For example, the Opus 4.6 model operated at a 3x multiplier, effectively limiting usage relative to its value. Recently, that multiplier was increased dramatically to 27x. Similarly, Sonnet 4.6 moved from 1x to 9x.

The Underlying Issue

These multiplier adjustments are only surface indicators of a deeper financial phenomenon: the increasing disparity between the cost of compute resources and what end-users are billed. AI companies have historically subsidized this gap, absorbing significant infrastructure costs while offering relatively low prices to consumers. However, this model is no longer tenable.

Key Factors Driving the Change

1. Growing Compute Demands: AI models like Claude Code, which support long-context sessions and complex agentic workflows, utilize vastly more tokens per user—sometimes 10 to 100 times more than basic chat completions. As a result, infrastructure requirements escalate, pushing the limits of current data center capacity. Building and scaling this capacity can take 18 to 24 months.

2. Rising Operational Costs: Since January, compute expenses for services such as GitHub Copilot have nearly doubled weekly. Major providers like Microsoft and Anthropic have been absorbing these costs to maintain service stability, but their margins are shrinking.

3. Transition to Honest Pricing: The recent multiplier increase to 27x suggests a move toward pricing structures that more accurately reflect actual resource consumption. It’s a wake-up call for users and organizations alike.

Implications for Organizations

The current model assumes widespread, often unrestricted access to powerful AI models. Many enterprises have integrated Copilot as a standard benefit, with little oversight or governance of individual usage. Employees deploy Opus and other models extensively for coding, review, boilerplate actions, and one-line completions—often without corporate awareness of the cumulative cost.

Upcoming Changes and Expectations

Effective June 1, GitHub plans to implement full usage-based billing. The recent multiplier hike serves as a warning: the era of flat-rate, unlimited access is ending. Organizations should anticipate detailed billing and tracking that link costs directly to specific user or team activity. This shift compels technical leaders and finance teams to reconsider their AI budgets, workflows, and governance policies.

Industry-Wide Transition

This evolution is not unique to GitHub. Major AI providers—including OpenAI, Anthropic, and others—are adjusting their pricing strategies to align more closely with the actual costs of compute infrastructure. Heavy, agentic usage—such as extensive workflows, long sessions, and resource-intensive applications—may soon be met with more deliberate cost management and governance.

The Bottom Line

The “free lunch” of unlimited AI access is drawing to a close. Organizations that rely heavily on AI models must proactively review and adjust their defaults, governance policies, and budget forecasts. Acting before June 1 will help mitigate unexpected expenses and foster more sustainable AI use practices.

In summary, these recent pricing adjustments underscore a vital shift toward transparency and sustainability in AI services. Staying ahead of these changes is essential for organizations aiming to optimize their investments and maintain efficient workflows in an increasingly cost-conscious AI landscape.