OpenAI has stopped accepting new subscribers to ChatGPT Pro, its $200 monthly tier, because demand for Astra, the model it shipped on September 3, is straining the systems that run it. Thibault Sottiaux, OpenAI’s product lead for Codex and ChatGPT, announced the pause on X rather than in a press release, which is itself a signal of how quickly the decision came together.
Turning away the highest-priced consumer tier is an unusual choice for a company that has spent two years pitching Pro as the plan for its most demanding users. “We wanted to take the smallest step that allows us to continue giving the broadest access possible,” Sottiaux wrote, framing the pause as protective rather than punitive. The API, along with the lower-priced Go and Plus plans, keeps accepting new users.
The warning came before the pause. On Wednesday, Sottiaux wrote that “demand for Astra is really unprecedented. We’re pulling all the levers possible to sustain the demand, but I’ve not seen anything like it until now and we went through very steep growth before.” He added that the priority would be existing users, with new Pro sign-ups the thing to sacrifice if growth kept climbing.
OpenAI has not disclosed how long the pause will last or how many people are trying to sign up each day, so there is no public baseline for how large this surge actually is. That opacity matters: a company confident the spike is temporary would have every incentive to publish a number that makes the constraint look short-lived. It has not.
The timing points to a real capacity problem rather than a manufactured one. Only last month OpenAI lifted how much Codex users could consume, which suggests the current strain arrived on top of an already-tight system rather than building gradually. Pro is also the tier OpenAI has said puts the most load on its infrastructure per user, so cutting it off first is the fastest lever available for freeing up capacity without touching the free and mid-tier plans that carry the most total users.
Astra rolled out across Pro, Plus, Enterprise, and Business accounts with claimed gains in reasoning, coding, and computer use, and OpenAI has called it the start of the “AGI era,” its own characterization rather than an independent assessment. That branding is doing real work here: a launch framed as a generational leap invites exactly the kind of demand spike that now has the company rationing its most expensive product. The more interesting fact than the marketing line is the tradeoff OpenAI chose. A company optimizing purely for revenue would keep taking $200 checks and let service quality slide. OpenAI instead shut off new money to protect the experience for people already paying it, which says more about how tight frontier inference capacity is right now than any benchmark claim attached to Astra.
For teams evaluating Astra access, the practical read is to lock in a Pro seat only if you can get one before the pause lifts, and to treat any capacity announcement from a frontier lab this month as a proxy for how constrained GPU supply has become industry-wide.
Reported by Sarah Perez for TechCrunch, September 10, 2026.
