Source: Microsoft's GitHub Copilot, the $10/month service used by 1.5M+ people, loses an average of $20+ per month per user and as much as $80 for some users
Microsoft, Google and others experiment with how to produce, market and charge for new tools — Tech companies are touting …
Context & Ripple Effects
GitHub positioned Copilot as a $10-per-month developer tool when it moved out of preview, after more than 1.2 million people tried the preview. This report shows that the early subscription price did not map cleanly to the cost of serving users.
The loss gap also provides context for GitHub's later plan to move from request-based to token-based billing as operating costs rose: flat-rate access leaves the provider carrying the cost of unusually intensive use.
First-order effects
- Microsoft is subsidizing Copilot usage at the reported price, with heavier users creating substantially larger losses than the average subscriber.
- Copilot users retain a simple, low fixed monthly price for now, while Microsoft absorbs the variation in inference demand across the user base.
Second-order effects
- The reported economics pressure Microsoft to distinguish light from heavy usage through limits, higher tiers, or usage-based charges rather than relying solely on one flat subscription.
- Other AI tool vendors face a similar pricing test: broad adoption is less valuable if incremental model-serving costs outpace subscription revenue.
Third-order effects
- AI developer tools may move away from conventional SaaS-style flat pricing toward plans that expose or constrain inference consumption, particularly for high-intensity workflows.
- If costs remain variable by user and task, competitive differentiation will increasingly depend on cost per useful coding task, not just model capabilities or headline subscriber counts.
The trend: Generative-AI software is being forced to align familiar subscription pricing with inference costs that rise unevenly across users.