OpenAI has temporarily suspended new subscriptions and upgrades to its $200-a-month ChatGPT Pro plan, citing unprecedented demand for its new AI model, Astra. The decision highlights the growing pressure on AI companies to meet rapidly increasing demand from users.
Unprecedented demand for its most expensive ChatGPT Pro offering has led OpenAI to take the unusual step oftemporarily halting news signups and upgrades for its highest priced, $200-a-month plan. As usage for its most advanced artificial intelligence model to date, Astra, continues to spike dramatically, demand for compute continues to grow as well.
OpenAI said its decision is intended to protect users currently using the $200-a-month plan while the company works through the increased demand, but isnotaffecting any existing subscribers who pay $200-a-month for ChatGPT Pro.
Why did OpenAI pause $200 ChatGPT Pro?
OpenAI Product Leader Thibault Sottiaux described the company’s priciest Pro subscriptions as being the most demanding on its systems, and decided to temporarily stop adding more users instead of limiting others elsewhere. He told Reuters he hopes, however, that the company can bring in more people to the plan soon. The move follows recent remarks by OpenAI, that it was experiencing unprecedented demand for the company’s flagship model.
Increasing demand at this unprecedented level for its services, is why it’s seeking new, efficient ways to ensure a healthy and predictable experience across all users.
It comes as accessing advanced AI is becoming just as much of a hardware issue as it is software.
Astra boosts demands to new highs.
Launched Sept. 3, Astra is OpenAI’s latest flagship artificial intelligence model. The company is marketing it as capable of advanced operations including complex reasoning and coding, use of a computer, and additional tasks expected by advanced professionals.
OpenAi is positioning it as a key step towards artificial general intelligence.
In part to the announcement, the company is reporting historic usage in individual and business consumers. The pressure on its infrastructure is one challenge the new version of its service, called Astra, presents. The decision signals the next step for the business as it seeks new ways to rapidly expand its computer capacity in an effort to balance user demand. As all top A.I. companies rush to hire hardware engineers, and gobble up computing infrastructure through leases, advanced models will only begin to require exponentially large amounts of processing capabilities if users demand them for the intensive operations that only they can do.
Existing ChatGPT Pro users still have their access
Existing subscribers will not lose access to the $200-per-month plan of ChatGPT Pro. The plan will continue to be renewed according to existing users’ billing schedules. People who canceled or moved out of the $200 plan may face a waiting list for its reintroduction until the company resumes all subscriptions for the service.
The OpenAI service still offers an alternative $100 ChatGPT Pro plan that currently only offers five times the message caps of ChatGPT Plus; the higher $200 Pro plan offers 20 times the usage cap of Plus.
OpenAI continues to maintain service on all of its plans, including API calls, though it has restricted access at higher levels.
What it Means For OpenAi
The decision marks a point of inflection in its growing generative AI industry for its business leaders. The highest price plan caters to those with the strongest compute demands for the model and has been subject to throttling during heavy traffic, and is limited to those subscribed to it. The company, by limiting new individuals from this high-access tier, can manage to lower increased demands on its compute capacity; therefore allowing existing 200$ customers to continue their service without degrading the level.
The maneuver further illuminates how compute has become crucial for A.I.
The development and iteration of AI systems alone, however, can only carry business so far when consumers engage it. This move suggests that providing access will become about securing adequate computer infrastructure.
The AI computation race continues on
OpenAI’s latest move coincides with wider problems in artificial intelligence’s compute demands. The advanced computing power needed both during development and for direct operation for training these large and sophisticated models can greatly impact a company’s bottom line when large customer populations need service on a particular unit at scale. Astra’s growth implies consumers are now using AI for tasks more intense than simple chats, which can increase demands even further.
These demands call for high end processing, a finite number of skilled labor with hardware knowledge and an infrastructure that can scale fast enough to meet user engagement without seeing a decline in quality.
This demand comes as AI companies compete directly for both chips as well as high end silicon where and data centers which are becoming increasingly important and challenging to scale for.
