OpenAI CEO Sam Altman issued an exhausted plea in the early hours of Sunday, March 30, 2025, as the company’s servers buckled under an overwhelming surge of users generating Studio Ghibli-style images with ChatGPT’s latest model.
“Can y’all please chill on generating images? This is insane, our team needs sleep,” Altman posted at 5:02 a.m. ET on X, revealing just how late he and his staff have been working to keep the platform running.
The frenzy began on Wednesday, March 26, when OpenAI rolled out its upgraded image-generation capabilities with the GPT-4o model—a release it called “the most advanced images yet.” What followed was an internet-wide art spree, as users began transforming photos, film stills, and even memes into dreamy, hand-drawn visuals reminiscent of Hayao Miyazaki’s iconic Studio Ghibli style.
By Friday, social media feeds were dominated by pastel skies, wide-eyed characters, and anime-styled nostalgia. One X user quipped,
“It’s like the internet turned into a Miyazaki movie overnight.”
But the viral success came at a cost.
OpenAI’s infrastructure, powered by GPU-heavy servers, began crashing under the strain. Altman said the image generation load was “biblical” in scale—so intense that even paid users on ChatGPT Plus experienced slowdowns. On Thursday, he admitted the servers were “melting” and pleaded with users to “slow down.”
To cope, OpenAI temporarily limited free-tier users to just three image generations per day, a significant cut from the previous unlimited access.
“We haven’t caught up since launch,” Altman explained in a post to Paul Graham on Sunday, noting that many staff members had been working nonstop since Wednesday.
The internet, of course, had fun with Altman’s plea. One user, LaDoger, mocked,
“Fire your team and generate a new team.”
Altman replied:
“No thanks. I’ve got the best team in the world.”
Despite technical hiccups, the tool itself is earning rave reviews. Users praised its ability to capture the signature Ghibli aesthetic—soft lighting, warm palettes, and whimsical detail—often better than previous AI models. But not all attention has been positive. Some critics raised copyright concerns, questioning whether the model was trained on Studio Ghibli artwork without permission. Others warned OpenAI about alienating its users if it continues restricting access.
“If you keep nerfing the model, people will stop,” one user posted.
Altman replied:
“We’re gonna do the opposite—but still, chill out a bit.”
The Ghibli image boom may be just the beginning. YouTube co-founder Chad Hurley hinted that OpenAI’s next leap could be into video generation, suggesting another major wave of user demand is coming soon. Altman has alluded to bigger plans too, including building the “biggest website in the world” and eventually reaching AGI (artificial general intelligence). But the CEO also acknowledged the toll on his team.
“It’s just hard,” he said.
“We’re building it from scratch, and it’s massive.”
This isn’t the first time Altman has spoken candidly about burnout. In 2023, he told he often lost sleep over whether ChatGPT was released too early, worried it might spark AI consequences he couldn’t control.
“It’ll come sooner if you stop using our GPUs for images,” Altman joked to one user asking about AGI timelines.
For now, OpenAI is pushing updates to ease the load. But with user excitement still high, Altman’s main request is simple: give his engineers a break.