AI Companies Shift to Usage-Based Billing Amid Rising Inference Costs

On Monday, GitHub announced that all Copilot subscriptions will move to usage-based billing on June 1. The change replaces the former “premium request units” with “GitHub AI Credits,” each credit equal to one dollar of the subscriber’s monthly payment and consumed per token processed. A $10-per-month plan therefore provides $10 in AI credits; any token usage beyond that incurs additional charges. The shift follows an earlier notice from Microsoft that “significant strain” on servers forced the end of free trials for new Copilot accounts earlier this month.

Futurism

AI providers have long offered cheap or free access to large language models, absorbing the high inference costs of compute-intensive tasks. The rise of AI agents and continuous-coding sessions has amplified demand, straining infrastructure. Microsoft, as GitHub’s owner, oversees the Copilot policy shift. GitHub’s product chief product officer Mario Rodriguez leads the pricing overhaul. Anthropic, creator of the Claude code model, has recently tightened rate limits during peak hours, while Google’s Antigravity coding environment imposed weekly usage caps earlier in 2024.

Futurism

The shift aims to make AI services financially sustainable by matching revenue to actual compute consumption. Aligning pricing with token usage preserves service reliability while protecting provider margins. For developers, the change introduces variable costs that could deter extensive AI-assisted coding if expenses rise sharply, potentially slowing broader AI adoption in software development.

Futurism

GitHub said the change aligns Copilot pricing with actual usage and represents an important step toward a sustainable, reliable service. Mario Rodriguez explained that the previous model could not be maintained, noting that a brief chat and a multi-hour coding session can cost the same amount. Anthropic and Google have publicly limited usage to manage demand, prompting alarm among users who fear that tighter limits on Claude Code and Antigravity may curb productivity. Public disclosures do not reveal exact token-price rates beyond the credit system, leaving the scale of extra charges unclear.

Futurism

• “A quick chat question and a multi-hour autonomous coding session can cost the user the same amount,” — Mario Rodriguez, GitHub product chief product officer

Futurism

• “no longer sustainable.” — Mario Rodriguez, GitHub product chief product officer

Futurism

• “On Monday, it announced that all GitHub Copilot plans will be moved to usage-based billing, charging customers based on the number of tokens their AI tasks consume.” — GitHub announcement

Futurism

• “aligns Copilot pricing with actual usage and is an important step toward a sustainable, reliable Copilot business and experience for all users.” — GitHub announcement

Futurism

• “significant strain” — Microsoft notice to Copilot customers

Futurism

GitHub will track token consumption after June 1 and may adjust credit pricing. Anthropic and Google are expected to refine usage caps as demand evolves, while developers weigh cost-benefit trade-offs for AI-assisted coding.

Futurism