OpenAI Faces GPU Shortage, Delays GPT-4.5 Rollout

OpenAI's CEO Sam Altman recently revealed that the company has run into a significant roadblock: it is currently out of GPUs, causing delays in the rollout of its new model, GPT-4.5. In a post shared on X, Altman emphasized the model's extensive demand for hardware, stating that it requires “tens of thousands” more GPUs before the broader ChatGPT user base can fully access its features. The rollout will be staggered, with GPT-4.5 first becoming available to subscribers of ChatGPT Pro starting Thursday, while ChatGPT Plus users will gain access the following week. The pricing for GPT-4.5 is notably steep; OpenAI is charging $75 per million tokens inputted into the model and $150 for each million tokens it generates. This pricing structure is significantly higher than that of GPT-4, with costs increasing by 30 times for input and 15 times for output. The increase in expenses has raised some eyebrows, leading to discussions about how this may indicate the challenging nature of creating and running such large models. Altman noted that while OpenAI has experienced rapid growth, this expansion has led to unexpected GPU shortages. He stated, “This isn’t how we want to operate, but it’s hard to perfectly predict growth surges that lead to GPU shortages.” He also mentioned that the company is striving to address these challenges by developing its own AI chips and constructing a large network of data centers to bolster its computing capacity in the future. In summary, OpenAI is navigating the complexities of scaling its capabilities amid a hardware shortfall, impacting the launch pace of its advanced AI models while also adapting its strategies for long-term sustainability.