TL;DR
- OpenAI paused new sign-ups and upgrades to the $200 ChatGPT Pro plan (Pro 20X) on 10 September 2026, citing “unprecedented” Astra demand straining system capacity. Existing $200 subscribers, the $100 Pro tier, Business, Enterprise, and the API were all unaffected.
- The $200 tier’s 20× usage allowance, the largest flat-rate cap OpenAI sells an individual, gave heavy users enough headroom to strain systems while still paying a flat fee.
- OpenAI has committed tens of gigawatts of new compute via Nvidia, AMD, and Broadcom, with first deployments targeted for late 2026, against a demand spike that took a week, and has given no date for lifting the pause.
In plain English
Astra is OpenAI’s newest and most capable model, able to operate a desktop the way a person would.
ChatGPT Pro $200 (Pro 20X) is OpenAI’s highest consumer usage tier, capped at 20 times Plus.
Metered vs. flat-rate access: metered means paying per unit of use, like OpenAI’s per-token API billing; flat-rate means one fee for a capped allowance, like Pro $200’s monthly limit.
Compute capacity, measured here in gigawatts, is the physical limit on how much AI processing a company’s data centres can run at once, and it only grows as new data centres and chips come online.
On 10 September, OpenAI stopped taking new sign-ups and upgrades for their $200-a-month ChatGPT Pro plan — the tier that promises 20 times the usage of Plus, the biggest flat-rate allowance OpenAI sells to individual subscribers (though existing subscribers keep their access).
From their help page: “As of September 10, 2026, we’re temporarily pausing new sign-ups and upgrades to the ChatGPT Pro $200 plan (Pro 20X). This includes sign-ups and upgrades from Free, Go, Plus, or Pro $100.”
OpenAI product lead Thibault ‘Tibo’ Sottiaux, who owns Codex and ChatGPT, flagged the possibility a day earlier, on X:
“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. Priority will always be to keep excellent service for existing users, but we might have to pause new Pro subscriptions for a bit if this continues.”
“To make sure our current users have an incredible experience and continued access to Astra, we are going to pause subscriptions to our $200 Pro plan. These put the most strain on our systems and we wanted to take the smallest step that allows us to continue giving the broadest access possible. All other plans and the api remain available.”
He added that the company is “working on adding more capacity as fast as we can.”
The resource being rationed here is the headroom inside the $200 tier’s 20x allowance. A subscription is simply a bet that the average user won’t use their full allowance; and that ceiling was generous enough that even OpenAI’s own most demanding users could use far more than the plan assumed they would, especially once a single feature — Astra’s “computer use” mode — made every session use compute at close to metered-API rates, even though customers were still paying a flat fee.
Since the last time OpenAI put a pause on signups (November 2023) they have spent heavily on infrastructure: a strategic partnership announced with Nvidia to deploy at least 10 gigawatts of systems, with Nvidia investing up to $100 billion in OpenAI as each gigawatt comes online; a 6-gigawatt agreement with AMD across multiple GPU generations; a Broadcom collaboration to deploy 10 gigawatts of custom AI accelerators; and the Stargate infrastructure programme with SoftBank and Oracle. All three chip deals target first deployments in the second half of 2026, with the Broadcom rollout running through 2029. None of that capacity turns on overnight. A data centre is a multi-year build; a demand spike is a week. The Pro pause exists because those two timelines don’t match.
OpenAI hasn’t said when the pause lifts, how many people were signing up per day, or how close the $100 Pro tier or Business plans are to needing the same treatment. OpenAI itself calls this “temporarily,” with capacity work underway, though the company gave itself no deadline either. The underlying tension it’s managing — model capability advancing faster than the physical buildout that runs it — doesn’t resolve on OpenAI’s marketing timeline. It resolves when OpenAI adds enough compute capacity.