(Bloomberg) — For much of the past year, OpenAI has been pushing ahead with a complicated effort to turn itself from a nonprofit into a more conventional moneymaking business that would be more appealing to investors — and doing so over heated objections from former employees, academics and rivals, including Elon Musk.
But on Monday, OpenAI bowed to public pressure and partly walked back its proposed corporate conversion. While the company still intends to make changes to its for-profit arm to be more investor-friendly, the nonprofit will maintain control of the overall business, effectively preserving the way it currently operates.
The revised plan is not a done deal, however, and it remains unclear whether OpenAI’s concession will be enough to appease critics, state regulators and investors, not all of whom have the same priorities. Any further uncertainty risks slowing down the company and unnerving stakeholders at a time when OpenAI faces heightened competition in the US and China.
As Bloomberg News reported late Monday, OpenAI has so far not received the blessing of a major stakeholder: Microsoft. The software giant wants to make sure that any changes to OpenAI’s structure adequately protect Microsoft’s $13.75 billion investment, according to several people familiar with the matter. The two companies are still actively negotiating the details.
Meanwhile, Musk, who co-founded OpenAI a decade ago, appears intent on continuing his legal crusade against the company for allegedly betraying its founding mission to develop AI to benefit all people. Marc Toberoff, Musk’s lead counsel in pending litigation against OpenAI, said the updated approach “changes nothing.” In response, OpenAI said the fact that Musk, who runs a rival AI startup, is “continuing with his baseless lawsuit only proves that it was always a bad-faith attempt to slow us down.”
OpenAI also needs buy-in from the state attorneys general of California and Delaware. A spokesperson for California Attorney General Rob Bonta said the state’s Department of Justice is reviewing the new proposed plan, while Delaware’s Kathy Jennings said she is “encouraged” by OpenAI’s changes. “I intend to review it for compliance with Delaware law by ensuring that it accords with OpenAI’s charitable purpose and that the non-profit entity retains appropriate control over the for-profit entity,” she said.
For OpenAI, the original proposed restructuring was the culmination of a decade-long evolution from being a nonprofit research lab to a global corporation with a growing suite of AI software that’s projected to bring in $12.7 billion in revenue this year. It also reflected the fact that OpenAI’s path to building better AI was far more costly than expected, requiring the startup to raise tens of billions of dollars in funding.
“We didn’t quite understand when we started how much of fulfilling our mission was going to be about putting AI tools in the hands of people,” OpenAI Chief Executive Officer Sam Altman said on a press call Monday. The new structure, he said, allows OpenAI to “act like a normal company when necessary to serve that mission.”
As it tries to pull off a corporate transition, OpenAI has to weigh potentially competing interests from investors who are eager to maximize their returns and safety advocates concerned about the company prioritizing profits and commercialization above all else. State officials, meanwhile, are focusing on whether OpenAI provided the nonprofit with fair value for its stake in the for-profit entity.
In a letter to employees, Altman said the decision to keep the nonprofit in control followed discussions with civic leaders and the offices of the state attorneys general. “We look forward to advancing the details of this plan in continued conversation with them, Microsoft, and our newly appointed nonprofit commissioners,” Altman said.
A simplified for-profit structure is considered more attractive to investors. Instead, OpenAI is taking a different approach with the hope of achieving the same aim. The company is converting its for-profit division into a public benefit corporation — an entity free to pursue income but with the goal of bettering society — and removing a cap on the financial returns its investors can earn. Previously, any returns that investors made above a certain threshold — 100 times the initial investment for OpenAI’s first round of investors — would go toward the nonprofit.
“As we were looking at various options, we definitely wanted something that works for investors, or at least works well enough for investors that they’re happy to continue to fund us to the degree we think we will need,” Altman said. With the changes, SoftBank Group Corp. is prepared to move forward with its full $30 billion investment in OpenAI as part of a recently announced funding round, Altman said.
Less clear at the moment is what stake the nonprofit will ultimately have in the for-profit and how that stake will be valued. In a blog post, OpenAI said only that the nonprofit would be a “large shareholder” in the public benefit corporation. It’s also unclear if Altman, who famously holds no equity in OpenAI, will receive some in the new structure. A person familiar with the matter said there have been no changes or active discussions on the matter.
The governance structure is a work in progress, too. Under the new plan, the nonprofit will have a separate board from the for-profit, but both will initially be made up with the nonprofit’s current directors. Over time, the nonprofit will be able to select board members for the for-profit.
But it’s still being determined whether the nonprofit will be able to fire OpenAI’s for-profit directors or executives, according to the person, who spoke on condition of anonymity to discuss private deliberations. That’s a crucial detail, given it was a prior incarnation of the nonprofit board that briefly ousted Altman as CEO more than a year ago.
Todor Markov, a former OpenAI employee who now works at rival Anthropic and previously signed onto an amicus brief in Musk’s case against OpenAI criticizing the restructuring, said the plan for the nonprofit’s continued control is “a win for responsible AI development,” but “the details matter.”“We need strong guardrails, not just good intentions,” Markov said in a statement. “We’ll be watching closely to ensure this control remains more than just words on paper.”
More stories like this are available on bloomberg.com
