OpenAI was founded as a nonprofit with a mission to build safe artificial general intelligence for the benefit of humanity. For a while, that structure made sense. But in 2019, the company made a discovery that changed everything: scaling up AI models—with more data, compute, and parameters—led to predictably stronger results.
The insight was formalized in a 2020 paper titled “Scaling Laws for Neural Language Models,” and it reshaped OpenAI’s trajectory. That same year, the company released GPT-3, a model 100 times larger than GPT-2. Microsoft invested. Venture capitalists piled in. Inside the company, employees began to see Sam Altman as the one who could turn a nonprofit breakthrough into a world-changing—and highly profitable—business.
And yet OpenAI remained a nonprofit company. Seen in that light, yesterday’s announcement that OpenAI’s for-profit arm will become a “public benefit company” (PBC) is no big surprise.
Under the newly proposed structure, OpenAI will continue operating as a for-profit AI business housed within a nonprofit parent. (Altman said last year he wanted to free the for-profit from the nonprofit parent.)
“We made the decision for the nonprofit to retain control of OpenAI after hearing from civic leaders and engaging in constructive dialogue with the offices of the Attorney General of Delaware and the Attorney General of California,” OpenAI board member Bret Taylor said in a blog post Monday. The change is that the for-profit part will now be a “public benefit corporation” and no longer a “capped profit” entity.
Now there’s no limit on how much OpenAI shareholders—including investors and employees—can earn. Dropping the capped-profit model was also a condition of OpenAI’s last two funding rounds. In the most recent (and largest), lead investor SoftBank stipulated that OpenAI adopt a new corporate structure by the end of 2025. Investors are willing to bet big on OpenAI, but they want the potential for big returns.
Altman and others at OpenAI have said that bringing in revenue has become more important with the realization that building progressively better models will require massive investments in infrastructure and computing power.
