The Tokenpocalypse Was Always Coming

GitHub Copilot just changed how it charges you. Not a small adjustment. The whole model. Instead of paying a flat monthly rate for your AI pair programmer, you are now on a per-token meter.

Per. Token.

Uber, meanwhile, burned through its entire annual AI budget in four months. Not "they overspent by a lot." Four months. The company eventually imposed spending caps on employee AI usage. That is what out-of-control token costs look like in practice.

A Reddit user coined the term "Tokenpocalypse" to describe what is coming. I am formally adopting it.

Here is what is actually happening. AI companies spent three years getting you hooked on cheap compute. They subsidized usage, burned venture capital, competed on raw capabilities to win market share. Free tiers everywhere. Flat subscriptions priced at obviously unsustainable levels. The strategy was never complicated: form the habit, then figure out the business.

Now they are figuring out the business.

Anthropic is reportedly preparing an IPO. OpenAI is in the same position, actively working toward profitability before going public, per TechCrunch. When a company files an S-1, it has to show unit economics that survive scrutiny from people who are professionally skeptical. "Unlimited AI for twenty dollars a month" does not survive that scrutiny.

So GitHub reprices. Others will follow.

Meanwhile, a senior OpenAI employee is telling people internally that "chat is dead." Which is a fascinating thing to say about the core product of a company valued at over a hundred billion dollars. What they mean, spelled out by the executive who oversees core product and platform, is that ChatGPT is becoming a "super app." A personal agent that handles "everything in your life, be it personally or at work." OpenAI is still building this, apparently rolling it out in "coming weeks."

Read: they want to be present for more of what you do. More compute. More billing surface.

I get it. Running the most computationally expensive consumer product category in history is genuinely expensive. We have already watched the government start buying equity in OpenAI, which tells you something about how serious the money question has gotten. This is not villain behavior. This is what happens when an industry has to reconcile its costs with its revenues.

But the shape of what is coming matters.

The companies with IPO pressure are repricing first. If you are locked into an enterprise contract for another 18 months, you have some runway. If you are on a month-to-month consumer plan, you are in the front of the line. The metered billing changes are not random. They are precisely calibrated to extract more from high-usage customers first while keeping the platform sticky for everyone else.

Uber hit the wall fast because nobody tracked usage. That is the thing about AI adoption in most companies: it happened because individual teams adopted it, not because procurement got involved. Then one quarter you are explaining to finance why your cloud spend tripled.

So. Some practical things.

Start tracking your usage before someone else tracks it for you. Every major AI platform has usage dashboards now. Turn them on. Find out whether the value you are getting actually maps to the tokens you are burning.

Treat your current flat-rate subscription as a temporary discount. Build your team's habits assuming the AI budget is a real constraint, not a rounding error.

And if "chat is dead," know what is replacing it. An all-in-one super app that handles your personal and professional life is not cheaper than a chatbot. It is the same compute, more integrated, priced accordingly. Check out what ChatGPT actually does today before getting excited about what it promises to do tomorrow.

The tokens were always going to cost something. The party is not over. Just more expensive.

Want more takes on where AI is actually headed? Subscribe to About.chat Weekly at newsletter.about.chat and we will send it to you every week.