OpenAI Faces Computational Crunch: GPT-5 Delayed, New Chips and Features on the Horizon

OpenAI CEO Sam Altman recently revealed that the company is grappling with computational limitations, which has impeded the anticipated release of new AI models. As AI models become increasingly complex, distributing the required computing resources across multiple projects has proven challenging for the Microsoft-supported firm.

In response to these constraints, OpenAI is actively collaborating with Broadcom and TSMC to design proprietary chips aimed at mitigating their computational constraints. Altman confirmed that GPT-5, the company's next large AI model, will not see a release this year. Instead, the company plans to unveil several noteworthy versions later on.

Chief Product Officer Kevin Weil emphasized the necessity for further refinement of their video model, Sora, to ensure its accuracy and safety, as well as to scale its computing needs, before its next major update is rolled out.

Altman further noted that there are no release plans for the company's image model DALL-E and mentioned that the upcoming model update will be worth the wait. OpenAI has also denied any plans to launch a model codenamed Orion this year.

Reddit saw the introduction of an “ask me anything” session and ChatGPT’s new search functionality, enhancing OpenAI’s competitiveness against Google, a subsidiary of Alphabet. This development underscores the ongoing rivalry in the AI domain.

Regarding recent executive departures, VP of Engineering Srinivas Narayanan stated that despite the loss of some partners, the company continues to boast a talented team and has acquired many new members, maintaining its momentum in innovation.