OpenAI's GPU shortage forces company to release GPT-4.5 in phases
28 Feb 2025

OpenAI, the leading artificial intelligence research lab, is facing a major hurdle in launching its latest model, GPT-4.5.

The company's CEO Sam Altman has revealed that they are "out of GPUs," forcing a phased rollout of the new model which boasts improved performance, "a deeper world knowledge" and "higher emotional intelligence."

This lack of hardare has required prioritizing access for different user groups and planning additional GPU acquisitions.

GPT-4.5: A 'giant' and 'expensive' model
Model description

The new model, GPT-4.5, is "giant" and "expensive," according to Altman. It will need "tens of thousands" of GPUs more before more ChatGPT users can access it.

Despite the GPU shortage, OpenAI has opted to first give access to ChatGPT Pro subscribers starting Thursday, with ChatGPT Plus customers getting access next week.

OpenAI's pricing strategy for GPT-4.5
Pricing details

Due to its size and complexity, GPT-4.5 comes with a hefty price tag.

OpenAI is charging $75 per million tokens (roughly 750,000 words) fed into the model and $150 per million tokens produced by it.

This pricing marks a major increase from the costs of OpenAI's previous model, GPT-4o, being 30 times the input cost and 15 times the output cost of that model.

OpenAI's future plans to combat computing capacity issues
Future strategies

Altman admitted the company's rapid growth and subsequent GPU shortage in a post on X.

He promised that OpenAI intends to add "tens of thousands of GPUs next week" to make the rollout possible for Plus tier users.

To avoid similar problems in the future, OpenAI plans to create its own AI chips and build a massive network of data centers over the years.