Sam Altman Says “Our GPUs Are Melting” Amid Viral Ghibli AI Art Frenzy

But the fun has come at a cost.

OpenAI CEO Sam Altman took to X earlier this week to admit that the company’s servers were buckling under the pressure. “It’s super fun seeing people love images in ChatGPT, but our GPUs are melting,” Altman wrote. He added that the company was working to improve performance, but in the meantime, they’ve implemented strict rate limits — especially for free-tier users, who are now capped at just three image generations per day.

The viral Ghibli-style image trend exploded almost overnight after ChatGPT rolled out its new image feature to a wider audience. Built on DALL·E 3, the image generation tool allows users to describe a scene or upload a photo, which is then transformed using AI into a Ghibli-esque animation. The results have been wildly popular, prompting everything from AI-reimagined family portraits to anime-style pet pictures.

However, the technical strain isn't the only issue confronting OpenAI. The Ghibli craze has opened a Pandora’s box of legal and ethical questions about the use of AI to replicate recognizable art styles — especially those protected by copyright and associated with living artists.

Hayao Miyazaki, the legendary co-founder and director of Studio Ghibli, has long voiced his disapproval of AI-generated art. In a resurfaced clip from a 2016 NHK documentary, Miyazaki called AI art “an insult to life itself.” That quote has taken on new relevance in recent days as critics argue that mimicking the Ghibli style — without Studio Ghibli’s input or consent — crosses a line from admiration into appropriation.

Josh Weigensberg, an intellectual property attorney at Pryor Cashman LLP, weighed in on the controversy. “There’s a legal gray area when it comes to style,” he said. “While style alone may not be copyrightable, if an AI system produces an image that replicates protected characters, scenes, or specific artistic elements, you’re getting into dangerous territory.”

To OpenAI’s credit, the company has implemented certain guardrails. The DALL·E image generator does not allow users to explicitly prompt for the style of a living artist by name. That restriction is part of a broader content policy designed to protect individual creators. However, terms like “Studio Ghibli-style” still yield results, revealing a potential loophole in how the model interprets and applies stylistic cues from its training data.

It remains unclear whether Studio Ghibli itself plans to respond or take legal action. The studio is known for its tight control over its intellectual property, but it has so far remained silent on the current trend.

Despite the controversy, fans continue to flood social media with their AI-generated masterpieces. Some view the trend as a harmless homage, while others believe it risks undermining the value of human-made art. The debate taps into broader anxieties about AI’s role in creative industries — a space traditionally thought to be uniquely human.

OpenAI is now racing to address the dual pressures of massive user demand and growing scrutiny. Altman emphasized that improving efficiency and GPU capacity is a top priority, noting that “scaling up is part of the challenge when innovation goes viral.”

In the meantime, the “Ghibli Tsunami” shows no sign of slowing down. As ChatGPT continues to push the boundaries of what AI can create, it also forces a vital cultural conversation: how do we embrace new technologies while respecting the artists and traditions that inspire them?

One thing is clear, the line between inspiration and imitation has never been blurrier.