When OpenAI launched Astra last week — a model whose advanced abilities prompted OpenAI president Greg Brockman to call it the start of the “AGI era” — it only gave a handful of enterprise customers access to give the company time to scale its compute capacity.
But now that it’s available to paying users, demand for Astra has soared to the point where OpenAI temporarily won’t take your money for its $200-per-month ChatGPT Pro plan, where Astra is available, because there’s not enough capacity to go around.
The company announced on Thursday it’s pausing new sign-ups and upgrades to the most expensive plan on the Pro tier, which provides 20 times the usage of the Plus tier below it. OpenAI’s head of product Thibault Sottiaux wrote on X that the reason behind the halt was that the $200 plan puts the most strain on the AI lab’s systems.
“We wanted to take the smallest step that allows us to continue giving the broadest access possible,” he said, adding that OpenAI is “working on adding more capacity” and that existing Pro accounts are unaffected.
A day earlier, Sottiaux said demand for Astra was “unprecedented” and warned that the company “might have to pause new Pro subscriptions for a bit” if it continued.
“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 wrote on X.
Astra’s design helps explain why too many users on it might strain OpenAI’s infrastructure. Its marquee feature, called “computer use,” allows it to interact with a desktop the way a human would, like filling out forms and navigating web pages at “superhuman speed,” Brockman told reporters. OpenAI also said Astra can burn through users’ usage allowances faster than its previous flagship model GPT 5.6 Sol.
Compute constraints
This isn’t the first time OpenAI paused new subscriptions to paid tiers due to lack of capacity. In November 2023, OpenAI temporarily halted sign-ups to the $20-per-month-ChatGPT Plus plan after the company’s first developer conference led to a surge in demand.
“The surge in usage post devday has exceeded our capacity and we want to make sure everyone has a great experience,” OpenAI CEO Sam Altman wrote on X at the time.
Since then, the company has grown its capacity. OpenAI said its available compute rose from 0.2 gigawatts in 2023 to about 1.9 GW in 2025, a 9.5-fold increase over two years.
It’s also trying to keep up in the race to grab more of it, even as concerns swirl that a potential bursting of the AI bubble could leave a glut of computing capacity behind. AI hardware like GPUs, the computer chips that run the calculations used to train AI, can depreciate in a few years.
OpenAI expanded beyond its partner Microsoft Azure for the hardware required for compute, adding cloud-computing company CoreWeave and launching Stargate, a $500 billion, four-year AI infrastructure buildout.
OpenAI has also lined up a wave of new capacity beginning in the second half of 2026. This includes plans for at least 10 GW of Nvidia systems, a 6-GW Advanced Micro Devices agreement for GPUs, and a Broadcom partnership to deploy 10 GW of custom AI accelerators.
OpenAI did not respond to Fortune’s request for comment.