Compute

Copilot's first metered
month ends with
five-figure invoices and
a Gartner warning

GitHub's June 1 switch to token billing closes its first cycle today, with agentic developers reporting 10x–50x cost jumps, a Slash employee burning $81,267 in a week, and Gartner forecasting AI coding will out-cost developer salaries by 2028.

By Felix Aroyan Compute & infrastructure · June 30, 2026

GitHub Copilot’s first full month of token-metered billing closes today, and the early invoices look less like a developer tool and more like a cloud-compute bill in 2014. Agentic users are reporting costs running 10x to 50x their old flat-rate subscriptions, with individual projections jumping from $29 to $750, and from $50 to $3,000. The cycle that began June 1 has, in 30 days, converted a predictable SaaS line item into a variable-cost problem the industry hasn’t fully priced yet.

The mechanics are simple, and that’s the trouble. A single agentic task can trigger five to thirty separate model calls; GitHub’s own May 2026 research notes that agentic coding can consume roughly 1,000 times more tokens than a single-turn query. One million output tokens from GPT-5.4 nano costs about $1.25 through Copilot, while the same volume from GPT-5.5 runs roughly $30. That 24x spread sits inside an interface that, until last month, charged a flat fee and hid the model choice behind a dropdown.

The receipts have piled up fast. A Copilot Pro+ subscriber burned 8% of a 7,000-unit monthly allocation in two hours. One user reported a single change request costing more than $6. A “build a Minesweeper game” prompt routed through Claude Haiku 4.5 ate 94 credits; another complex prompt consumed 171. Uber’s CTO burned $1,200 in a two-hour internal demo, and per TechCrunch the company has since capped tool spending at $1,500 per employee per month. Nicolas Brilliante, an employee at the $1.4 billion fintech Slash, spent $81,267 in tokens in a single week.

Gartner made the structural call on June 24: AI coding token costs will surpass the average developer salary by 2028. Senior Principal Analyst Nitish Tyagi’s team already finds 23% of tech leaders spending $200–$500 per developer per month on tokens, and 6% spending above $2,000. ComputerWeekly’s read of the same data lands in the same place, which is that the unit economics of agentic development are being discovered, not designed.

GitHub is cushioning the landing. Business plans get $30 per user per month in promotional credits, Enterprise gets $70, and a $100 Copilot Max tier sits above both. The credits expire September 1. A GitHub spokesperson has pointed users to the new tooling; the GitHub forum, Reddit, and X have pointed back with screenshots.

The FinOps Foundation’s February survey caught the shift early: 98% of practitioners now actively manage AI spend, up from 31% two years earlier. Token budgets are becoming a line of business inside Microsoft’s Experiences and Devices division and everywhere downstream. The metered era of developer tooling didn’t arrive with a keynote. It arrived with an invoice.

Sources