From a small research lab “free from financial obligations” to one of the most valuable private businesses in the world, the company that ushered in the generative AI era has evolved a lot over its decade of existence.
Now, OpenAI is on the verge of more change. Its longtime strategic partnership with Microsoft seems to be fraying. The company’s long drift into for-profit-dom is coming to a head. And as the company still at the center of the generative AI boom, OpenAI’s structural pivots and alliance shifts have implications for the rest of the industry.
“This is going to be a very, very interesting next 12 to 18 months because of all these dynamics,” Ritu Jyoti, group VP at market intelligence firm IDC, told Tech Brew.
History of change: Having gathered top research minds in the AI space from day one, OpenAI has always encompassed a series of competing and evolving visions for what the next generation of AI should look like.
That was one of the takeaways Tyler Johnston, executive director of the nonprofit Midas Project, arrived at as he spent the past several months sifting through more than 200 news articles, legal filings, employee testimonials, and other information about OpenAI’s history. The group recently published the result of this work as a research report called the OpenAI Files in partnership with advocacy group Tech Oversight Project.
“OpenAI is a company that has existed on this insanely accelerated lifespan,” Johnston told us. “It’s just astonishing how much has happened in the 10 years that it’s existed, and how many different visions of the organization have, at various times, flourished or been squashed based on how convenient they were to the leadership at the time.”
Backed by big-name investors and co-founded by Sam Altman and Elon Musk, among others, OpenAI immediately made a splash upon its launch in 2015 as a nonprofit. The lab had already assembled a laundry list of top researchers, with the not-for-profit purpose as part of the recruiting pitch, according to a Wired feature at the time.
“We just talked about every specific structure and thought this one had the most advantages,” Altman told Vanity Fair of the nonprofit mission. “I think that the misaligned incentives [of profit-seeking] would be suboptimal to the world as a whole.”
Some nonprofit: OpenAI once again grabbed headlines in 2019 when it declared it would not release the full second iteration of its generative pre-trained transformer model (GPT-2) due to “concerns about malicious applications of the technology.” The model was capable of producing “synthetic text samples of unprecedented quality.”
A month later, OpenAI converted to a capped profit structure, where returns beyond 100 times an investment would be passed to the nonprofit board. A few months after that, Microsoft invested $1 billion to kick off a “multiyear partnership.”
“This was brilliant, by Microsoft, by [CEO] Satya [Nadella],” Vasant Dhar, a data science professor at NYU’s Stern School of Business and author of a forthcoming book on the history of AI, told Tech Brew. “He just saw this thing coming, you know, from much further away than most people did, and he clinched a really solid deal.”
Frenemy era? But as OpenAI perhaps continues down its path toward full for-profit status—despite some pushback and potential setbacks—experts we talked with say Microsoft and OpenAI were destined to drift apart eventually. The companies have reportedly clashed over their revenue-sharing agreement and OpenAI’s proposed acquisition of coding platform Windsurf. (The deal ultimately fell through when Google swooped in with a shock acquihire of top Windsurf employees.) OpenAI has signed competing cloud agreements, and Microsoft has rolled out competing models.
Microsoft and OpenAI said in a joint statement emailed to Tech Brew that they “have a long-term, productive partnership” and are “optimistic we will continue to build together for years to come,” a statement also provided to The Wall Street Journal, Reuters, and others. In a January blog post, Microsoft reiterated that its partnership would continue with OpenAI through 2030, while also softening its cloud exclusivity terms.
Part of the tension, too, hinges on language around the original agreement, which ends when OpenAI achieves what its board determines to be AGI, which it defines as “a highly autonomous system that outperforms humans at most economically valuable work.” But AGI does not have a single agreed-upon definition, and that’s reportedly caused rifts as well.
“That is fundamentally a very, very slippery thing to base your benchmark on, especially as over the past 50 years, we’ve kind of continuously redefined what AGI looks like in practice,” Rowan Curran, principal analyst with Forrester Research, told us.
While the pressure campaign to keep OpenAI a nonprofit appeared to notch a win earlier this year, Johnston argued that the framing of the story was misleading.
“In my mind, OpenAI is very conveniently describing control in such a way that they retain a lot of the freedom that they wanted…while still acting as if they had made a concession,” Johnston said.
But an eventual Microsoft-OpenAI split could also be a boon for the overall ecosystem and the companies that are using this technology as it spurs new competition, Jyoti said.
“It would diversify the AI providers—it will not be just centered around Microsoft and OpenAI. It will create some new alliances,” Jyoti said. “It will give developers and enterprises more choice, but it can also [add] more complexity.”