OpenAI Freezes $200 ChatGPT Pro Sign-Ups as AI Compute Runs Short

OpenAI has suspended new subscriptions and tier upgrades for its premium $200-per-month ChatGPT Pro service, while existing Pro account holders maintain uninterrupted access. The tier provides 20 times the usage allowance of the $20-per-month Plus subscription and is the most computationally intensive offering in OpenAI’s consumer and enterprise lineup.

The restriction follows a rapid deployment of autonomous “agentic” artificial intelligence that is pushing global computing infrastructure toward its operational limits. OpenAI head of product Thibault Sottiaux called pausing sign-ups for the top pricing tier the “smallest step that allows us to continue giving the broadest access possible” as technical teams work to bring additional server capacity online.

OpenAI’s recently launched Astra model requires substantially more real-time computational power than traditional text-based conversational tools. Days before the subscription freeze, Sottiaux described demand for Astra as “unprecedented,” saying internal growth metrics had exceeded every previous spike in the company’s history.

At launch, OpenAI President Greg Brockman framed Astra as marking the beginning of the “AGI era.” The advanced model includes “computer use” functionality, allowing it to visually interpret desktop interfaces, navigate web pages, and fill out complex forms autonomously at what Brockman called “superhuman speed.”

Unlike standard large language models that produce static text responses, Astra processes continuous visual feeds alongside multi-step reasoning loops. Those workloads consume hardware resources at a significantly higher rate than previous flagship systems such as GPT 5.6 Sol, allowing individual users running agentic workflows to exhaust backend compute allocations much faster than on earlier model generations.

OpenAI’s active operational compute footprint has grown from approximately 0.2 gigawatts in 2023 to nearly 1.9 gigawatts by 2025, a 9.5-fold expansion in available power. Even with that increase, next-generation AI software continues to outpace immediate hardware supply.

The company has expanded beyond its primary computing partner, Microsoft Azure, by entering cloud hosting agreements with specialized infrastructure provider CoreWeave. OpenAI also launched “Stargate,” a four-year, $500 billion infrastructure buildout intended to create massive custom data center campuses.

The current freeze recalls an earlier capacity crisis in November 2023, when OpenAI temporarily stopped sign-ups for the $20-per-month ChatGPT Plus tier. That interruption followed the company’s inaugural developer conference, after a surge in consumer and enterprise interest overwhelmed server capacity and Chief Executive Officer Sam Altman announced a public pause until infrastructure could catch up.

Server deployments depend heavily on advanced graphics processing units (GPUs) and specialized accelerators, which face tight global manufacturing constraints and rapid physical depreciation cycles. OpenAI has secured commitments for another wave of compute infrastructure scheduled to begin deployment in the second half of 2026.

Those supply agreements include Nvidia Systems procurement plans for at least 10 gigawatts of high-performance GPU clusters, an agreement with Advanced Micro Devices (AMD) to secure 6 gigawatts of graphics processing hardware, and a Broadcom partnership to design and deploy 10 gigawatts of custom AI accelerator chips.

Until the multi-gigawatt installations become operational, OpenAI remains dependent on real-time capacity management to balance resources among its free, Plus, Pro, and enterprise user tiers. OpenAI did not respond to requests for comment about current capacity timelines or estimated dates for resuming ChatGPT Pro sign-ups.