The End of the All You Can Eat Buffet
GitHub is finally admitting what everyone in the AI industry already knew: unlimited AI code generation for a flat monthly fee is a Ponzi scheme of computing power. Starting June 1, Copilot users will be charged based on actual token consumption, breaking the current model where a quick autocomplete and a massive autonomous coding session cost the same. This is a brutal but overdue reckoning with the physics of inference costs.
The new currency? AI Credits, pegged to token usage with variable rates depending on model complexity. Expect sticker shock when you realize a single GPT-5.5 powered debugging session could burn through your monthly allocation. GitHub claims this move “reduces the need to gate heavy users,” which is corporate speak for “we’re bleeding money on your agentic workflows.”
The Agentic AI Tax Is Coming Due
The real story here is the cascading cost crisis across the major AI providers. Leaked internal documents suggest GitHub’s costs have nearly doubled since January, coinciding directly with the rise of agentic assistants like Openclaw that run endlessly consuming tokens. This isn’t just about Copilot. Anthropic has already started charging enterprise clients for full compute costs and briefly throttled Claude Code usage on Pro plans.
This trend signals the end of the generously subsidized AI era. The AI giants have their growth metrics but are burning cash to keep the illusion of affordability alive. Now they’re coming for your wallet. If you thought your $20/month subscription covered heavy use, think again. The days of unlimited AI access are over, and the token counters are coming for every developer who bought into the cheap coding utopia.
Source: Arstechnica
