This weekend, a soft, rumbling feeling reached a loud consensus: Why the heck was OpenAI structured that way?
In very dumb words, its nonprofit arm had complete control over the for-profit holding company, giving it the ability to oust Sam Altman as CEO without even a day’s heads-up to the well-known Silicon Valley executive or any of the organization’s investors. More technically, as explained to me by stealth startup founder James Rosen-Birch, OpenAI is a tax-exempt charity wrapped around a holding company, which then has a majority stake (with Microsoft as a minority owner) in the for-profit side of OpenAI.
OpenAI’s origin story is now famous: a company that was founded to advance research and development into artificial general intelligence and realized it needed lots of money to do that. It took money from investors and sovereign wealth funds who, of course, wanted returns, so they created a for-profit subsidiary that could issue them equity. That subsidiary remained, however, under complete control of the tax-exempt charity and its board, who, unlike the investors, did not have a financial stake in the business, only the will to see that AI was used for the good of humanity. If this company structure gives you the ick, you are not alone.
Though no one knows precisely what happened, a common theory is that Altman’s for-profit efforts became at odds with the mission-driven nonprofit board.
Internal tensions
In fact, Kimberly Bryant, founder of Ascend Ventures, told us what many in the tech world were thinking: As OpenAI became more popular, struck pretty brand deals, and sought a near-$90 billion valuation, it introduced commercial complexities that the board may have felt were contrary to the company’s stated objectives. Though it is a nonprofit, vision misalignment is hardly unknown in the for-profit sector, leading to conflicts between board directors and CEOs.
“Nonprofit entities inherently prioritize ‘serving the public good’ over maximizing profit, a commitment that faces challenges amid the dynamics of hypergrowth and the diverse objectives of investors,” Bryant told TechCrunch+. “Issues arise when boards become overly controlling, overstep their authority, or act with self-interest . . . such dynamics can not only impede progress but also pose a significant threat, potentially causing severe damage to the organization or company.”
Rosen-Birch said OpenAI’s structure created several overlapping problems and questions, such as whether a for-profit company in a tax-exempt shell really exists for the good of humanity if it doesn’t have to pay for shared public goods and services. “And perhaps most relevant to the problem at hand, how does a board judge whether a company is acting in the interest of humanity? What power do they have to enforce those interests?” he said. “In hindsight, it seems the board was just as confused about [the answer to these] questions as we are.”
The latest update as of publication is that Microsoft, one of OpenAI’s biggest investors, wants Altman to work for them; nearly the entirety of OpenAI has signed a petition to quit if he doesn’t come back, and the fate of the OpenAI board is up in the air.
A cautionary tale
Though OpenAI’s board structure is unique, the fallout from the last few days should serve as a cautionary tale for founders and board members. RareBreed Ventures founder McKeever Conwell said founders rarely pay attention to the structure of their own boards until something grave, like a Silicon Palace coup, is staged against them. “Everybody wants to say boards are broken, but they’re not,” he said. “People just don’t know how to manage their boards.”
Bryant, who was ousted from her last company by the board, said the OpenAI mess could teach startups to “meticulously” choose their board of directors, establish clear expectations, impose term limits, and diligently ensure alignment with the organization’s long-term vision.
“Actively seeking board members who contribute strategic value beyond financial investment is crucial,” she said. “Equally important is the development of comprehensive bylaws and a charter that provides robust guidelines for critical decisions, such as leadership changes and conflict resolution.”
The key word in what Bryant said is “comprehensive.” OpenAI has a charter, its only known guiding document, which is less than 500 words. Rosen-Birch noted that it doesn’t really provide actionable rules or guidance and grants the board only the power to select an executive or just fold the company. “When they were faced with a novel and confusing problem, which we do not yet know the details of, they had to evaluate it against a vaguely defined philosophical mandate and act accordingly at the risk of personal liability,” he said. “The results, as we saw this weekend, were equal parts confusing and disastrous.”
Furthermore, Conwell added that it’s essential for founders to do their due diligence on who they put on their boards. A board having the power to fire its CEO is good for holding one accountable but can be risky depending on the motives of the people on the board. A good board has an odd number of people, according to Conwell — one that equally represents the company, the investors, and a third-party individual for an outside perspective.
“When you have an even number of folks on the board, it’s hard to break even if there is a tie or when you have a board that is contentious from the offset,” he continued. OpenAI had six people on the board and currently has four. People on X (Twitter) debated the impact more investors on the board would have had in helping avoid what happened over the weekend.
The OpenAI board had Altman (who was fired); the company’s president, Greg Brockman (who was removed from the board and then resigned after Altman was fired); its chief scientist, Ilya Sutskever; Quora founder Adam D’Angelo; Geosim Systems CEO Tasha McCauley; and Helen Toner, director at Georgetown University’s Center for Security and Emerging Technology. (HBO, are you paying attention?)
Jason Schloetzer, an associate professor at Georgetown University’s McDonough School of Business, told us that a majority of boards should have a thorough process before ousting a CEO, including having evidence that could stand up in court. In its initial statement, the OpenAI board said that Altman had not been “consistently candid” with them but did not go into further detail.
Schloetzer said a thorough investigation should happen in cases such as this, especially since, in typical cases, a CEO could sue for breach of an employment contract. Emmett Shear, OpenAI’s interim CEO, promised an investigation, but perhaps that effort should have come before, not after, a situation arises, along with having clear terms as to why a CEO could be ousted in the first place. (Shear has now suggested he would resign if the board doesn’t tell him why they ousted Altman.)
It’s worth noting that in its initial statement, the board did say it came to a decision after a “deliberative review process,” though it was quite vague on what that meant. There are plenty of scenarios where a board might not be able to come out and say what happened, for example, in cases of pending litigation. Given, however, the way the board has handled Altman’s ousting, it’s not surprising that the tech community universally panned how the OpenAI board handled the move to the point where even Sutskever said he “deeply regrets” his involvement in what happened.
Boards are supposed to represent the interests of the shareholders and act as their representatives to ensure that the corporation’s actions are copacetic. Typically, when a founder is pushed out, it’s because the board feels they’ve made a grave mistake that jeopardizes the company or because the board feels they no longer have the skill set to run the company.
The OpenAI board has yet to clarify why they ousted Altman, but it is clear they realized his removal was hurting shareholders’ value, prompting an ill-fated attempt to reinstate him. It didn’t help that they told their shareholders they had ousted him only minutes before the world found it. “The rapid reversal was likely a byproduct of both pressure from investors and agitation inside the company,” Rosen-Birch said.
“It’s crucial for board members to recognize conflicts are inevitable in any organization, but their resolution can shape the organization’s future for better or worse,” Bryant said. “Navigating conflicts requires a thoughtful, empathetic, and strategic approach, which includes engaging with trusted counsel, moving forward with transparency and accountability, neutralizing bad actors on the board, and upholding high ethical standards when making these decisions.”
New conversations
This fallout did inspire, however, more conversations about board structures. Luckily, startups typically operate as a C-Corp, which gives them standardized governance structures and well-established norms. Their board members will also be more invested in the businesses. The OpenAI fiasco is not likely to spread.
“For founders who want to play with corporate structure: Don’t use corporate structure as a form of marketing, and be really strict in how you define how decisions are adjudicated and incentives are aligned,” Rosen-Birch said.
There often isn’t enough information about boards for startups compared to the resources that can help, for example, fundraising. Conwell has worked with founders with little to no idea how to manage or work with boards and said that board members can be jealous, manipulative or good team players. To learn more about prospective board members, he said it’s important to call founders and ask them what they’ve heard. Often, their thoughts on an individual will not be found in tweets or a Medium post “but in a phone call,” Conwell said.
There are traps to avoid, however. Schloetzer said an unfortunate by-product of this OpenAI drama might be founders attempting to use this situation to justify pushing for dual-class share structures or stacked friendly boards as a way to prevent a similarly shocking event from happening to them. A dual-class share structure occurs when a company uses more than one class of stock with disparate voting rights, often allowing a group of insiders to receive superior voting rights per owned share, granting them a large say over corporate decisions.
Meanwhile, a friendly board exists when a company appoints a majority of directors who simply do whatever the leadership wants. “These situations have been shown to generally insulate management from market pressures and pay themselves excess compensation, both of which are indicative of lower governance quality,” Schloetzer said.
Overall, it’s essential for directors and founders to understand the importance of engaging skilled counsel experts in corporate governance early in the initial stages of board formation and throughout the company’s growth, Bryant said. This is especially useful when critical decisions need to be made. “These situations are extremely costly, and the only ones who win at the end of a legal battle are the attorneys,” she said.