Are AI subscriptions getting tighter?

Has anyone else noticed this?

A few months ago, I was using AI exactly the way I use it today.

Same workflows.

Same type of tasks.

Roughly the same amount of usage.

Yet lately I'm hitting rate limits far more often than before.

The strange part is that I can't point to any major change in my behavior.

I'm not suddenly generating 10x more content.

I'm not running dramatically more complex prompts.

I'm not spending significantly more time in the platform.

But the limits seem much easier to reach.

Which raises an interesting question:

Are AI companies changing the economics of access now that these tools have become essential for many users?

Many revolutionary companies have followed a familiar pattern:

1. Deliver massive value.

2. Attract users at scale.

3. Become embedded in people's workflows.

4. Gradually optimize for monetization.

Uber was one example.

And lately I'm starting to wonder whether the AI giants are entering the same phase.

Not because the technology is getting worse.

But because the economics are changing.

As companies like OpenAI and Anthropic mature and move closer to public-market expectations, the incentive structure inevitably changes.

Growth is no longer enough.

Profitability starts to matter.

And when profitability starts to matter, power users stop being a success metric and start becoming a cost center.

That would naturally lead to tighter rate limits, more aggressive usage controls, and a gradual shift away from flat subscriptions toward token based pricing models.

I can't prove that's what's happening.

But it would explain why many of us feel like we're getting less access today than we did a few months ago despite paying the same subscription fees.

Curious to hear from others:

Have you noticed AI subscription limits becoming easier to hit recently?

Or is your experience exactly the same as it was 3–6 months ago?

---

From the comments

One thing I forgot to mention. Fast Mode uses more credits. Higher thinking modes use more compute. Naturally, users gravitate toward both because they promise better results without sacrificing speed.

It feels like AI companies are gradually training us to consume more compute while making the tradeoff feel like a feature rather than a cost.