OpenAI needs to seriously reconsider its pricing and capacity strategy.
Right now, many developers are willing to pay premium prices because GPT-5.5 is arguably the best AI model available for real-world work. But that advantage won’t last forever.
In the next 6 to 12 months, Chinese AI labs are likely to release models that rival GPT-5.5 in coding, reasoning, and agent workflows. If those models are offered at a fraction of the cost, developers will move. History shows that developers are loyal to performance and value, not brands.
The biggest risk for OpenAI is not that someone releases a smarter model.
The biggest risk is that someone releases a model that is 95% as good for 10% of the price.
At that point, most software teams won’t care about small benchmark differences. They’ll care about economics.
My view is simple:
A $100/month plan with significantly higher limits would make far more sense than aggressively restricting usage. Heavy users, founders, builders, and engineers are the people creating the next generation of products on top of these models. They are also the people most likely to leave if capacity becomes the bottleneck.
When AI reaches a certain capability threshold, usage limits become more important than raw intelligence gains.
If developers feel constrained while competitors offer similar performance with fewer restrictions, switching becomes inevitable.
The AI race is no longer just about who has the smartest model.
It’s about who gives builders the best platform to build on.
My concern is that the jump from $25 to $100 only delivers about 5x capacity.
For developers running coding agents, long reasoning sessions, and real production workloads, a $100 plan should feel like a major upgrade.
Personally, I think the structure should be closer to:
Plus ($25) = 1x
Pro ($100) = 10x
Pro Max ($200) = 20x
The reason is simple: the AI landscape is changing fast.
A year ago OpenAI’s lead was massive.
Today, Chinese labs and open-source models are closing the gap at an incredible pace.
If competing models reach 90–95% of GPT-5.5’s capability while offering more generous usage limits, many developers won’t optimize for benchmark scores.
They’ll optimize for productivity and cost.
OpenAI still has the best platform in my opinion.
But keeping builders on the platform may become more important than squeezing capacity out of the pricing tiers.
Honestly, I wasn’t complaining when the May limits were in place.
At $100/month, the value felt reasonable and I could work without constantly thinking about usage.
Since those limits were reduced, the experience feels noticeably worse for heavy users.
I understand OpenAI has costs and can’t offer unlimited inference forever.
What I find difficult to reconcile is that Sam Altman has repeatedly said that the cost of AI should decrease dramatically over time, even mentioning cost reductions on the order of 10x over the years.
From a user perspective, that trend isn’t showing up in the available capacity.
If inference costs are falling that quickly, I would expect higher limits and more value for power users, not less.
OpenAI is still ahead today, but competitors are improving very quickly. If model quality converges while usage becomes more restrictive, developers will naturally start evaluating alternatives.
I’m not looking for unlimited access. I’m looking for enough capacity to build products efficiently without constantly managing limits.
Well, I don’t think we’ve got to that inflection point yet. Presently prices are climbing, especially ahead of IPOs - demand has outstripped supply and datacentre build projects are all behind schedule mainly due to power limitations - apparently there is built capacity that can’t be switched on because the electricity networks can’t supply them.
I’m sure prices will come down once supply catches up.
Come on, guys. We have more AI power available to individual developers today than billion-dollar tech companies had 5 or 6 years ago, and we’re arguing over an extra $100 a month? That is not a cost problem. That is a perspective problem.
It takes time for costs to go down, but they do. As a long time chatgpt subscriber that is very perceptible, but not in the short time.
Also, considering the unstable market on GPU and energy supply we currently are, I see the continuous resets and bonuses as openai’s best effort on doing what you asked without compromising on uncertain future costs that could make the product bleed money.
Considering how things have been going, I think it is fair to assume limits will raise when it is possible.
For example, 200 dollars pro used to be only 7x plus usage, now it is 10x.
Ah, just noticed that OpenAI realises what mess they got themselves into.
When I cancelled my subscription I was greeted with this friendly offer, and took it.
I advise you all do the same - cancel and use one free.
In the IPO world this is called “market signal”.
Some of my products running at 900%, and there is nothing extraordinary about that. I think you meant something like 700%? And even that it’s not something unheard of.
Well, with all due respect, he did say that they had already reduced the cost of compute 10x in the interview that he’s misquoted in, in the screenshot.
I’ve also likely misquoted him but it was in a past tense and he then said something to the effect that they weren’t sure if they could do it again…
Business Insider could be sued for that headline, tbh.
There are few people I listen to in this world, and I don’t instantly agree with what those people say…
But Sam is certainly one of those people my ears perk up when I hear him speaking.
