OpenAI pauses ChatGPT Pro sign-ups as Astra demand strains compute

DQChannels Bureau

14 Sep 202611:54IST

OpenAI has temporarily paused new sign-ups and upgrades for its USD 200-per-month ChatGPT Pro plan as strong demand for its latest AI model Astra puts additional pressure on computing capacity.

The move comes shortly after Astra became available to paying users following an initial limited rollout for enterprise customers. Its growing use is creating capacity challenges, particularly on the highest-priced ChatGPT plan, which offers substantially higher usage limits than the Plus tier.

According to information shared by OpenAI’s product leadership, the pause is aimed at managing demand while maintaining service quality for the wider ChatGPT user base. Existing Pro subscribers will continue to have access while OpenAI works to add more capacity.

Astra puts more pressure on AI infrastructure

One of the main reasons for the higher compute demand is Astra’s wider set of capabilities. A key feature is “computer use”, which allows the AI system to interact with computer interfaces, navigate webpages and complete tasks that would normally require direct human input.

Astra can also consume usage allowances faster than earlier flagship systems. Complex or longer tasks therefore require more computing resources.

This points to a problem that AI providers are increasingly facing. Better models do not simply mean better answers. They can also mean much higher infrastructure requirements.

As advanced AI systems move from demonstrations into everyday work, providers need more GPUs, data-centre capacity and high-performance computing resources to serve users at scale.

The issue is not entirely new for OpenAI. In November 2023, the company temporarily stopped new ChatGPT Plus sign-ups after usage increased sharply following its first developer conference.

Since then, OpenAI has expanded its computing resources considerably. Available compute reportedly grew from around 0.2 gigawatts in 2023 to approximately 1.9 GW in 2025.

Yet the latest pause shows how quickly demand can again put pressure on capacity when a powerful new model reaches a wider audience.

The infrastructure race continues

OpenAI has been expanding its computing footprint through several infrastructure partnerships as AI workloads grow. The company has moved beyond its earlier dependence on Microsoft Azure, adding CoreWeave and pursuing large-scale infrastructure through its Stargate initiative.

More capacity is expected to come online from the second half of 2026. OpenAI has also reportedly committed to large deployments involving Nvidia GPUs, AMD accelerators and custom AI chips developed with Broadcom.

The challenge goes beyond simply adding hardware. AI companies have to balance computing demand with the cost, availability and long-term economics of specialised infrastructure.

For businesses and technology professionals, the development offers a useful reminder of what sits behind every AI interaction. A model may appear to be a software product, but its ability to respond quickly and reliably depends on a large physical infrastructure running underneath it.

The temporary restriction on new ChatGPT Pro sign-ups shows that even major investments in AI infrastructure may not be enough to remove capacity bottlenecks when demand rises quickly.

As AI systems become more capable and take on computer-based tasks, the next stage of competition may therefore depend on two things: how intelligent the models become and how efficiently their providers can scale the computing infrastructure needed to run them.

What keeps a TSC Auto ID partnership going for nearly two decades?

What is changing for Sennheiser channel partners in India's AV market

RFID business in India gains momentum as channel sees new opportunities