What Actually Changed With GitHub Copilot's Pricing

For years, the deal was simple. You paid a flat monthly fee, and GitHub Copilot was just there in your editor, helping you write code. The bill didn't move. You knew what you owed.

That's over now. GitHub flipped its Copilot pricing to a usage-based model, charging by how much work the AI does rather than how many times you ask it to do something. Instead of counting your requests, the system now counts tokens — and that's a completely different math problem for anyone who leans on Copilot all day.

Here's the part that stings. Under the old per-request setup, one prompt was one prompt, whether the AI churned out two lines or two hundred. Now the heavy lifting gets metered. A single chunky task that used to count as one cheap request can quietly eat a big slice of your monthly budget. And for a lot of developers, that's translating into bills that are 10 times higher than what they were used to. Sometimes more.

Why People Are Watching Their Credits Vanish

The change landed barely a day before users started sounding the alarm, and the speed of the burn is what's freaking people out. This isn't a slow creep. It's a sprint.

One developer reported tearing through more than half their monthly credits in a single day. Another did the quick mental math and realized that a month that normally costs them around 60% of their allowance had already chewed up nearly 20% on day one — meaning they were on track to blow way past their limit before the month was even close to done. Someone else watched their entire monthly token budget evaporate in under half a workday. Even the more optimistic users weren't immune; one person who felt fairly upbeat about the new model still admitted to burning through over 70% of their credits in the first 24 hours.

Then there are the projection numbers, and these are the ones that really land. Plenty of users ran their typical workloads through GitHub's own pricing estimator to see what they'd be paying going forward. One developer who'd been comfortably spending about $39 a month suddenly found themselves looking at a potential bill of nearly $1,800. Think about that jump for a second. That's not a tier upgrade. That's a different financial planet.

The Bigger Story: The Cheap-AI Era Is Ending

Here's what's really going on underneath all of this, and it's bigger than one product.

For the past few years, subscription pricing for AI tools has been the norm — and it's been heavily subsidized. Companies ate the real cost to get people hooked, to drive adoption, and frankly to hide just how expensive running these models actually is. It worked. We all got used to powerful AI for a flat, almost suspiciously low fee.

But 2026 is when the bill comes due. As AI agents drive usage way up, the platforms and model developers are no longer willing to swallow the difference. They want to get paid for what they're actually delivering. This isn't a GitHub-only move, either. Anthropic shifted Claude Enterprise subscribers over to token-based billing back in April, and Microsoft followed the same path with Copilot this week. The pattern is clear: usage-based pricing is becoming the default, and the days of treating frontier models like an all-you-can-eat buffet are numbered.

What Developers Are Doing About It

So how are people responding? A couple of different ways.

Some are simply looking for the exit. Frustrated by the new costs, a portion of users say they'll jump to alternative tools rather than watch their bills balloon. When the price tag changes this dramatically, loyalty gets tested fast.

Others are taking a more measured route — they're rethinking how they use the AI rather than abandoning it. The idea is to be deliberate. Instead of firing off prompts casually all day, you use the AI in a very focused way, reserving it for the tasks where it genuinely earns its keep. It's a discipline shift more than a tooling shift. Whether that's enough to keep costs sane is something each developer is going to have to figure out for their own workflow.

What This Means Going Forward

Look, whether you switch tools or just tighten up your habits, the takeaway is the same. The wild, low-cost stretch of frontier AI from the major developers looks like it's coming to an end, and it's ending faster than most people expected. The subsidies that made these tools feel almost free are getting pulled, and what's left is the actual cost of the compute.

There's a wrinkle worth sitting with, too. Several of these companies have big public offerings on the horizon. How a sudden wave of pricing shock and user frustration plays into those plans is anyone's guess right now. The move toward usage-based billing makes the revenue look healthier on paper — but it also risks pushing away the very developers who built these tools into daily habits in the first place.