OpenAI's ChatGPT $100 Plan Is Not For You

Sam Altman wants you to think he did you a favor.

On Thursday, he posted on X that OpenAI is launching a $100-per-month subscription tier, and made sure to include five words that should give everyone pause: "by very popular demand." Classic Altman. Frame a business decision as an act of generosity. The truth, as it usually is with OpenAI's pricing moves, is less flattering.

The new tier slots between the $20 Plus plan and the $200 Pro plan, offering five times the Codex usage of Plus. On the surface, fine, a $180 jump between adjacent tiers is genuinely awkward, and developers hitting rate limits on Plus had nowhere to go but a plan that costs as much as a car payment. But filling a product gap and responding to user demand are two different things. One is a product decision. The other is a revenue decision dressed in customer-first language.

Altman's post said nothing about the other side of the ledger. OpenAI is burning through money at a scale that makes most tech companies look fiscally conservative. 

Despite raising $40 billion in its latest funding round, yes billions, the company continues to operate at a significant loss, the $40 billion disappears faster than you'd think when you're running the most compute-hungry models in the world at consumer prices. Compute costs compound. And Codex, which now has three million weekly users, up fivefold in three months, is almost certainly one of the most expensive things OpenAI runs.

Also read: Sam Altman’s Home Hit in Molotov Attack as Threats Reach OpenAI HQ

The people most willing to pay more are already self-selecting through heavy usage and OpenAI knows exactly who they are. The $100 tier doesn't create new demand, it monetizes demand that was being subsidized by the $20 plan. Which is exactly why the limited-time promotion baked into the launch is worth paying attention to. 

Subscribe before May 31st and you get 10x Codex usage instead of 5x. Get users hooked at elevated capacity, let workflows form, then normalize the lower limits once the promotion expires.

A spokesperson told TechCrunch the new plan was designed to offer developers better value than Claude Code, and that framing alone tells you something. Anthropic has had a $100 tier for a while now. The company that essentially created the modern AI industry just spent months without a response to a competitor's pricing tier, then quietly introduced one and called it user demand. That's not confidence. That's catch-up.

Also Read: Anthropic Accidentally Exposes Claude Code Source Code in npm Release    

They might even slowly, not dramatically, trim the capabilities of the $20 plan similar to what they did with Codex at its launch by restructuring limits in Plus to support more sessions per week instead of longer sessions per day. 

None of this makes OpenAI uniquely cynical. Every subscription business eventually faces pressure to segment users and extract more from the highest-value ones. The uncomfortable part is that the company that spent years insisting it was building AI "for the benefit of humanity" keeps framing commercial decisions as community-driven ones. Users had been asking for better rate limits on the $20 plan, more transparency around throttling, an explanation for why the $200 Pro plan quietly vanished from the pricing page. 

And if you can't afford any of it, the free tier now runs ads. Those requests are still pending. The revenue optimization, however, is right on schedule.

OpenAI used to at least pretend by saying "for the benefit of humanity". But now they don't bother.