GitHub Copilot Changes Billing Model: From Fixed Subscription to Tokens, Developers Push Back

Gaia Banfi
GitHub Copilot cambia modello di fatturazione dal canone fisso ai token, le proteste degli sviluppatori
In this article
Want similar results?
Discover how LumenONE can transform your customer management.
Learn More
Back to Blog

Since 1 June 2026, GitHub Copilot has abandoned its fixed-rate subscription system in favour of a token-based consumption billing model. The transition has sparked widespread discontent among developers, particularly those working in small teams or independently, and has fuelled a lively debate on Reddit and X.

How the New System Works

Under the old model, users paid a fixed monthly fee regardless of their actual usage. The new system charges based on the number of tokens consumed during work sessions — a logic similar to that of OpenAI or Anthropic APIs, applied however to a tool that many had chosen precisely for its cost predictability.

Developer Reactions

The most heated comments come from those who have seen their costs increase dramatically. One Reddit user reported paying $29 a month under the old plan and projecting a spend of around $750 a month under the new system. Another shared screenshots of costs jumping from $50 to approximately $3,000.

Not all the feedback has been critical. Some developers pointed out that the difference between low and high spenders comes down to usage habits: those who work with unstructured iterations and let agents run for hours consume far more tokens than those who use Copilot as a support tool for actual work.

“The only party responsible is Microsoft. Microsoft provided this billing system and continued to make it easier and easier to burn tokens in enormous quantities on individual premium requests, which can run for hours or days generating tens or hundreds of sub-agents.”

The Context: What the Old Model Cost Microsoft

The change raises questions about the losses Microsoft accumulated in the preceding period. The old fixed-fee model, with users making ever more intensive use of the tool, had a financial sustainability that was difficult to assess from the outside. Several users have openly questioned how many resources Microsoft was effectively subsidising.

The move to a consumption model therefore marks a strategic turning point: large companies with budget planning capacity can absorb the variability, while individual developers and small teams face potentially unpredictable costs.

At the time the original article was published, Microsoft had not responded to TechCrunch’s requests for comment.

Original article: techcrunch.com

Gaia BanfiLumenIA
I help Italian companies understand and adopt artificial intelligence in a concrete, safe, and measurable way.

You might be interested

See all
    Modelli di intelligenza artificiale italiani quali sono e cosa fanno
    • News

    Italian Artificial Intelligence Models: What They Are and What They Do

    In recent years Italy has developed a number of artificial intelligence language models trained with particular attention to the Italian language. The topic involves universities, research centers and companies, both…

    ⏱ 2 minuti di lettura
    Intelligenza artificiale negli studi professionali l AI arriva prima della sicurezza dei dati
    • News

    Artificial Intelligence in Professional Firms: AI Arrives Before Data Security

    In the offices of lawyers and accountants, artificial intelligence is used to search for rules and rulings and to summarise documents: case files, deeds, financial statements and payslips. These are…

    ⏱ 3 minuti di lettura
    AI Act dal 2 agosto 2026 cosa cambia davvero per le aziende tra obblighi GPAI e trasparenza
    • News

    AI Act from 2 August 2026: what really changes for companies between GPAI obligations and transparency

    From 2 August 2026 a new phase of the AI Act, the European regulation on artificial intelligence, comes into force. From this date the transparency obligations of Article 50 and…

    ⏱ 3 minuti di lettura