The explosion of AI content creation models has led both OpenAI and Google to simultaneously apply new limits to two tools that are attracting global attention: the Sora video creator and the Nano Banana Pro image creation model.
The restrictions were introduced during the US holiday season, when users tend to experiment with AI the most.
OpenAI announced that the free Sora app is only allowed to create a maximum of 6 videos per day.
Users who pay for ChatGPT Plus and Pro are not affected and can still buy additional video likes if needed.
Bill Peebles, head of Sora, shared on X that the GPU is operating at a very high load.
OurPU is heating up and we want to allow as many people as possible to listen to Sora, said Bill Peebles.
Although the term "PCU melting" is often mentioned by the community, experts say thatPCUs in data centers often do not melt, thanks to advanced cooling systems.
CEO Sam altman once compared the same situation in the Ghibli AI wave, then admitted that the statement was just to emphasize the level of overload.
On Google's side, nano Banana Pro users who are not registered can currently only create or edit up to 2 photos per day with Gemini 3 Pro.
Google's support page also clearly states that the limitations of Gemini 3 Pro may change over time, depending on usage volume.
Previously, the Nano Banana model running on Gemini 2.5 allowed creating up to 100 images per day.
The simultaneous tightening of limits shows that the amount of computing needed to create AI videos and images is becoming increasingly huge.
Modern multi-modal models require processing power far beyond document-only models, causing energy consumption and operating costs to increase sharply.
Although causing inconvenience to a part of free users, AI companies said the new limit helps ensure stable service quality, while preventing the system from falling into a state of overload when demand increases dramatically.
