The AI subscription bait-and-switch is happening right on schedule. GitHub Copilot is ending its flat-rate pricing and switching to consumption-based metered billing, and developers are pissed.
The pattern is familiar to anyone who's watched SaaS companies mature: offer an attractive flat rate to get users hooked, wait until the tool becomes essential to their workflow, then switch to metered billing that costs heavy users significantly more. It's the playbook that made AWS billions, and now it's coming for AI coding assistants.
Under the old model, GitHub Copilot cost $10-20 per month per developer, unlimited usage. Under the new model, you pay based on how many completions you accept, how much code the AI generates, and other usage metrics. For developers who rely heavily on the tool, costs could easily triple or quadruple.
Developers on social media are threatening to flee to alternatives like Cursor, Codeium, or open-source options. Some are dusting off their old coding workflows, determined to prove they never needed AI assistance in the first place. Others are just angry about the principle of the thing.
Here's what's really happening: GitHub (owned by Microsoft) is learning what their actual costs are. Running AI inference for millions of developers generating code all day is expensive. The flat-rate pricing was likely subsidized to gain market share. Now that they have it, they're adjusting prices to reflect actual costs.
From a business perspective, this makes perfect sense. From a user perspective, it feels like betrayal.
The deeper issue is that AI coding assistants occupy an awkward middle ground in developer tooling. They're not essential like your IDE or version control—you can code without them. But once you've adapted your workflow around them, going back feels like coding with one hand tied behind your back.
That creates dependency. And dependency creates pricing power.
Microsoft knows this. They've been playing the developer tools long game for decades. Give away free tools, get developers hooked, then monetize through enterprise licenses, cloud services, or—now—metered AI usage.
