During a recent Reddit AMA, OpenAI CEO Sam Altman acknowledged that limited compute capacity is a significant factor in delaying new product releases. Altman explained that the increasing complexity of AI models demands extensive computational resources, leading to difficult decisions on how to allocate their available capacity for various projects.
Reports indicate that OpenAI has struggled to secure sufficient infrastructure to train and operate its generative models. To address these challenges, OpenAI has reportedly been collaborating with Broadcom to develop a custom AI chip that could help meet these demands, though this chip is not expected until 2026.
Due to these resource limitations, some advanced features, like ChatGPT’s Advanced Voice Mode, remain constrained. Although OpenAI initially demonstrated this feature with visual capabilities in April, Altman indicated that these vision features won’t be available in the near future. According to Fortune, the demonstration was expedited to coincide with Google’s I/O developer conference, though some within OpenAI felt the technology wasn’t fully ready for release.
In response to questions about upcoming tools, Altman noted that there is currently no set launch date for the next version of DALL-E, OpenAI’s image generator. Similarly, OpenAI’s video-generating tool, Sora, is facing delays due to technical and computational challenges. These limitations, combined with efforts to ensure safety and manage impersonation risks, have slowed Sora’s progress, especially as it faces competition from video-generation tools like those from Luma and Runway.
OpenAI’s focus remains on enhancing the capabilities of its existing o1 “reasoning” models and their successors. At a recent developer event, OpenAI previewed future features, including improved image understanding. Altman hinted at significant releases later in the year, but none will be branded as GPT-5.
