Update: 01 Apr 2025

It appears that the company’s GPUs are being overloaded by OpenAI‘s newly released picture generator. Here is a detailed explanation of why GPUs are the essential component of AI systems. OpenAI appears to have encountered a new dilemma in the controversy surrounding AI that mimics Studio Ghibli‘s aesthetic.

While praising the reaction to the most recent image generator, OpenAI CEO Sam Altman claimed that the company’s GPUs were melting. The business is restricting the use of the functionality, Altman continued. Essentially, the business is being compelled to restrict the quantity of photos that ChatGPT produces. According to Altman, the free tier will have the capacity to generate three photographs every day.

Seeing how much visitors enjoy photographs in ChatGPT is a lot of pleasure. Our GPUs, however, are melting. While we try to make it more efficient, we will temporarily implement some rate constraints. I hope it won’t take long! Altman stated in his X post that ChatGPT’s free tier will soon receive three generations every day. After the image generator’s capabilities were overwhelmingly successful, Altman replaced his display graphic with a Studio Ghibli-esque representation of himself.

What is GPU..?

Altman’s claim that GPUs are melting due to the new picture generator is not amusing. ChatGPT, DALL-E, and even driverless automobiles are examples of generative AI systems that significantly rely on enormous amounts of processing power. The graphics processing unit, or GPU, is at the center of this computational capability. Understanding how GPUs operate and why contemporary AI systems depend so heavily on them is necessary to comprehend why they are melting.

We learned in elementary school how the central processing unit (CPU) functions as the computer’s brain. It is well known that the CPU can manage a wide range of functions, including loading apps, playing music, and operating web browsers. The CPU is capable of multitasking, just like the human brain (there is no analogy here). However, something more is required with AI systems. Artificial intelligence (AI) systems must operate iteratively, which means they must perform the same mathematical problems repeatedly—possibly a million times—in contrast to a typical computer, which completes numerous little jobs quickly. Massive volumes of data are needed for this, and they must be collected simultaneously. The GPU enters the scene at this point.