In a live broadcast on October 28 (local time), OpenAI CEO Sam altman revealed that the company is aiming to develop a legitimate AI research center, a system that can automatically carry out large-scale scientific research projects by 2028.
Previously, in September 2026, OpenAI was expected to launch a version of the AI research assistant at the internship level.
The announcement was made on the day OpenAI completed its transition to a public welfare company model, marking a major turning point in its exit from the nonprofit framework.
This makes OpenAI more flexible in mobilizing capital and expanding its global computing infrastructure, according to altman.
OpenAI's Science Director Jakub Pachocki said that this "AI researcher" is not humans, but a system capable of self-planning, analyzing data and conducting experiments on a scale beyond the capabilities of traditional research groups.
"We believe that deep learning systems can reach super intelligence in less than a decade," Pachocki said.
OpenAI is betting on two strategic directions: algorithmic innovation and increased experimental computing time, which is to allow the model to spend more time thinking about complex problems.
According to Pachocki, the current models can handle tasks in about 5 hours and achieve the same results as the top candidates at the International Mathematical Olympiad.
To speed up progress, OpenAI has committed to investing in 30 gigawatt of infrastructure, equivalent to a $1.4 trillion financial obligation, to serve new-generation AI data centers.
Along with that, the nonprofit OpenAI Fund still holds a 26% stake, responsible for managing research direction and investing 25 billion USD in AI application projects in medicine and technology safety.
According to altman, this is a framework that combines public and commercial interests, helping OpenAI maintain its mission of developing AI for humanity while accelerating towards the stage where AI can explore science on its own, a leap that can change human history.
