Your team probably uses GitHub Copilot for the same reason it uses Slack. Not because the CTO evaluated seventeen alternatives. Because half the developers started paying $10/month from their own wallets, got hooked, and told their manager "we need this for everyone." The enterprise contract followed the personal habit.

That pipeline — developer tries tool → loves tool → demands tool at work → company buys tool — is how every dominant developer product won its market. VS Code killed paid editors by being free. Docker entered companies through individual experiments. Git replaced SVN because developers refused to go back. The pattern is old, reliable, and depends on one thing: an accessible individual on-ramp.

On April 20, GitHub removed the on-ramp.

The news, briefly

We already covered the mechanics: GitHub paused all new individual signups for Copilot's paid tiers. Agentic AI sessions — where the model runs entire multi-step coding workflows without human intervention — cost more per user than the subscription collects. Microsoft is reportedly moving to token-based billing to stop the bleeding.

This article isn't about the economics. It's about the flywheel GitHub just snapped in half.

How developer tools actually win

Enterprise software procurement is theater. The RFP process, the vendor demos, the compatibility matrices — all of it exists to formalize a decision that was already made in someone's terminal six months earlier.

VS Code didn't win because Microsoft shipped a compelling pitch deck to CTOs. It won because individual developers downloaded it, installed their extensions, and told their teammates. By the time the "official editor" conversation happened, VS Code already owned 60% of the floor. Docker didn't need enterprise compliance dashboards in 2014. Kubernetes didn't have a sales team when Google open-sourced it. The enterprise contract is always the trailing indicator. The leading indicator is what a developer installs on a Saturday afternoon to build a throwaway project.

This is the flywheel: accessible free/cheap tier → individual adoption → muscle memory → workplace demand → enterprise deal. Every link depends on the previous one. Cut the first link and the chain doesn't degrade gracefully — it stops.

What GitHub actually paused

A junior developer starting a side project this week cannot sign up for Copilot Pro. The free tier still exists, but it offers basic autocomplete with no agentic workflows — no autonomous multi-file sessions, no agent mode. It's a demo, not a tool.

So that developer picks something else. Claude Code, Cursor, Windsurf — the specific name barely matters. What matters is that whatever they learn becomes the muscle memory they carry forward. Six months from now, when their team lead asks "what AI coding tool should we standardize on?", they won't say Copilot. They'll say whatever they've been reaching for since that Saturday afternoon.

Multiply by a few thousand developers per quarter. Now fast-forward to Q1 2027, when CIOs are locking AI budgets. The procurement committee asks engineering: "What are people already using?" And the answer is no longer Copilot-by-default.

GitHub's competitors understand this perfectly. Cursor charges $20/month but its door is open. Claude Code lives in the terminal, which filters for command-line developers, but those are exactly the senior engineers whose opinions carry weight in tooling decisions. Neither is a clean Copilot replacement. Neither needs to be. They just need to be the thing developers reach for when the Copilot door is locked.

The price of pausing

Forrester analyst Charlie Dai summarized the cost problem in TNW on April 21: "Cost structures built for lightweight assistance no longer hold." Fair. Agentic AI is expensive. Flat-rate pricing for unpredictable compute is a losing bet.

But there's a cost on the other side of the ledger that doesn't show up in the cloud bill. Copilot still leads workplace adoption — 29% per JetBrains' April 16 AI Pulse survey — but competitors are closing the gap fast. Claude Code and Cursor each sit at 18% adoption and climbing, and Claude Code pulls a 91% satisfaction rate against Copilot's more modest scores. That adoption lead was built entirely on the bottom-up flywheel. GitHub is now protecting margins by dismantling the mechanism that built its market share.

Maybe token-based billing fixes the unit economics. Maybe the pause lasts weeks, not months. But developer habits form fast and change slowly. Every week the door stays locked is a week of muscle memory forming somewhere else.

The CTO's 2028 problem

Here's where it lands. Some CTO in 2028 will sit in a planning meeting, ask the team what AI coding tool to standardize on, and hear a name that isn't Copilot. They'll wonder when the switch happened. It happened on a Saturday, in a dorm room, when an intern hit a signup wall, shrugged, and installed something else. That intern is now a senior engineer, and they just picked the company's default for the next three years.

GitHub didn't pause a subscription tier. It paused the only go-to-market strategy that ever actually worked for developer tools — and did it during the quarter when enterprise AI budgets get locked in.

The economics forced the decision. The consequences don't care about the economics.