The timing could hardly be more awkward — or more strategic.
On June 1, GitHub Copilot officially switched from flat-rate subscription pricing to token-based billing via "GitHub AI Credits," a change that has triggered one of the most vocal developer backlashes in recent memory. One day later, on June 2, Microsoft Build 2026 opens at Fort Mason Center in San Francisco, with the company widely expected to unveil its homegrown MAI coding model — a cheaper-per-token alternative that could reshape the competitive dynamics.
The back-to-back sequence has not gone unnoticed. Developers on Reddit, Hacker News, and X are calling the Copilot billing change "a joke," while anticipation builds that Microsoft's own model will arrive as a lifeline for users priced out of the new system.
What Changed: AI Credits Replace Flat-Rate Pricing
As of June 1, all Copilot plans — Free, Pro ($10/mo), Pro+ ($39/mo), Business ($19/user/mo), and Enterprise ($39/user/mo) — now bill based on GitHub AI Credits, where 1 credit equals $0.01 USD. Credits are consumed by actual token usage, with per-model rates that vary depending on which AI model a developer selects.
Key details:
- Code completions and Next Edit Suggestions remain unlimited — not metered
- Chat, agentic workflows, and code review now consume credits
- Credits do not roll over month to month
- Enterprise customers get promotional extra credits through August 2026
GitHub CPO Mario Rodriguez justified the shift in a company blog post by pointing to the rising compute demands of agentic coding: "Today, a short chat question can cost the user just as much as an autonomous coding session lasting several hours."
The Backlash: "What a Joke"
The developer community's reaction was swift. On GitHub's official discussion thread, the change accumulated 400+ comments and nearly 900 downvotes within days.
- One Reddit user reported their Copilot Pro+ bill jumping from $29/month to roughly $750/month
- Another showed a projected increase from ~$50/month to ~$3,000/month
- TechCrunch ran the story under the headline: "'What a joke': GitHub Copilot's new token-based billing spurs consternation among devs"
The core complaint is that a single agentic coding session — the kind Copilot's latest features actively encourage — can consume $30 to $40 in credits, three to four times a Pro subscriber's entire $10 monthly allotment. Critics argue that Microsoft built features that burn through tokens, then penalized users for using them as intended.
"Microsoft provided this billing method and they kept making it easier and easier to burn through massive numbers of tokens… while spawning dozens or even hundreds of sub-agents." — Reddit user
Defenders counter that the old flat-rate model was unsustainable — that heavy users were essentially being subsidized by lighter ones. Either way, the "all-you-can-eat" era of AI coding assistance for individuals is over.
Enter the MAI Coding Model
Microsoft Build 2026 opens today, and the biggest anticipated announcement is Microsoft's own coding model — part of the broader MAI model family that already includes MAI-Image-2.5, MAI-Voice-2, and MAI-Transcribe-1.5.
What makes the MAI coding model strategically important:
- Cost-efficiency: Expected to be significantly cheaper per token than third-party models, directly addressing the pricing backlash
- Native integration: Optimized for VS Code, GitHub, Azure, and Microsoft's enterprise toolchain
- Enterprise-ready: Built with Microsoft's compliance and security requirements baked in
- Strategic independence: Reduces Microsoft's reliance on OpenAI models for its core coding product
The MAI coding model doesn't need to beat every frontier benchmark — it needs to be good enough and dramatically cheaper, running efficiently on Microsoft's own Azure infrastructure. For organizations already in the Microsoft ecosystem, this creates a compelling value proposition.
What This Means for Developers
For individual developers, the short-term calculus is painful. Anyone using Copilot for heavy agentic workflows faces significantly higher costs starting now. Alternatives like Cursor, Claude Code, and OpenCode are getting renewed attention as developers re-evaluate their tooling.
For enterprises, the change brings predictability challenges — AI usage cost is now a line item that needs monitoring alongside cloud compute. But the promotional credits through August provide a transition window, and the MAI coding model could reduce costs later this year.
For the market, the message is clear: AI coding tools are entering their infrastructure-pricing phase. The era of flat-rate unlimited access is over, replaced by metered billing that reflects actual compute consumption. The winners will be the providers who can offer the best price-to-quality ratio — and Microsoft Build 2026 is shaping up as the company's answer to that challenge.
Sources: GitHub Blog, TechCrunch, MLQ.ai, Windows Forum, The Decoder
Cet article a été initialement publié sur The Agent Report.
Top comments (0)