Four Red Flags Behind the Attempted Coup at Open AI
The Have-Your-Cake-and-Eat-it-too Corporate Structure, The PR-Stunt “Capped-Profit” Model, The Largest Investor Without a Board Seat, The CEO with No Equity Stake (Or A Secret One)
In the last ten days, Sam Altman was fired as CEO of Open AI by its board, accused of lying, replaced twice, and promptly hired by its largest investor. Nearly all of the company’s employees threatened to quit. The board -- two of whose members have now been replaced -- appears to have considered every possible alternative before ultimately hiring Altman back.
Now that the dust has settled, many questions remain about what this means for the future of one of the world’s most important technology companies and whether AI Doomers or AI Accelerationists now have the upper hand.
But this chaos didn’t happen in a vacuum. Buried amidst the product launches and the consumer hype were at least a handful of bright red governance flags that facilitated the events of the last two weeks.
🚩 Red Flag No. 1: Open AI’s Have-Your-Cake-and-Eat-it-too Corporate Structure
Much has been made about Open AI’s byzantine corporate structure, where its non-profit parent organization owns and controls a for-profit subsidiary.
The best recounting of the events that led to the creation of these multiple entities comes from Semafor, who describes that in early 2018, Elon Musk, one of OpenAI’s co-founders, parted ways with Open AI, then a nonprofit, as a result of misaligned visions. Up to that point, Musk had funded the organization to the tune of $100 million and, as a result of parting ways, reneged on a promise to donate $1 billion. With the loss of its largest donor, Altman and team created a for-profit entity in order to raise enough money to pay for the computing power necessary to pursue the most ambitious AI models.
Mixing for-profit and nonprofit structures isn’t uncommon, and complex corporate structures — usually to avoid taxes, create more flexible financing options, or obscure ownership — are pretty standard. In fact, many of Open AI’s defenders have pointed this out.
But simplifying Open AI down to a nonprofit that owns a for-profit misses the issue at the root of the company’s recent governance turmoil.
Open AI’s structure -- unlike many of the best-known examples of nonprofits that own for-profits -- is the corporate legal embodiment of trying to have its cake and eat it too-- the governance of an organization driven solely by a mission with the financing of an organization driven by traditional things like growth, market share, and profit.
Open AI’s problem was that its board had no fiduciary duty to the company or its shareholders. Its mission of “building safe and beneficial artificial general intelligence for the benefit of humanity” is wholly misaligned with how the rest of the company operates. In order to attract top talent, the company offered competitive pay packages with large amounts of equity, and in order to fund the hiring of that talent and the computing power needed to train its models, the company raised a reported $13 billion from an array of institutional investors and companies such as Microsoft, its largest investor.
Many companies espouse missions that are above profit. But Open AI’s board embodied that ethos. Open AI’s company charter vaguely outlines a scenario under which the board might essentially shut down the company. At least three of Open AI’s board directors, the three outside directors -- had no financial incentive in the company’s success -- including no equity. Their duty is to humanity, not the company. The problem wasn't a mix of for-profits and nonprofits; the issue was a board with entirely different incentives than employees and shareholders.
The best-known examples of corporations owned by nonprofits are philanthropic entities, such as foundations, that use some proceeds of their ownership stakes for charitable giving or other socially-motivated activities. In those cases, the philanthropic entities are typically established once the companies are mature and profitable, and usually at the behest of a controlling shareholder. The stewards of these philanthropic entities are almost always aligned with other shareholders, all of whom are financially incentivized for the equity value to rise.
🚩Red Flag No. 2: The PR-Stunt “Capped-Profit” Model
When Open AI created the for-profit company to entice investors, it created a “capped profit” company to thread the needle between mission and profit.
As reported in TechCrunch at the time, “profits from any investment in the OpenAI LP (limited partnership, not limited profit) will be passed on to an overarching nonprofit company, which will disperse them as it sees fit. Profits above a 100x return that is. In simplified terms, if you invested $10 million today, the profit cap will come into play only after that $10 million has generated $1 billion in returns.”
Sounds noble? Not so much.
If we take Open AI’s words literally, its model doesn’t cap “investment returns,” it caps “profits.” The difference is meaningful. In Silicon Valley, early investors often generate returns over 100x their original investment. The goal of doing so is baked into the model of venture capital, where the vast majority of investments are expected to fail, with the remaining winners generating large enough returns to make up for the failures.
So, it's conceivable that Open AI’s earliest investors will ultimately generate a return on their investment greater than 100x. What’s far less likely is that OpenAI generates profits above 100x the amount of invested equity.
Open AI has raised $13 billion from equity investors. To generate profits greater than 100x the amount of capital the company has raised, it would need to generate more than $1.3 trillion in profits, which, according to Gergely Orosz on Twitter, is nearly twice as much profit as Apple, the world’s most valuable company, has generated since being founded in 1976.
This is why the capped-profit model is much less altruistic than it sounds. First, it is unclear to what extent this capped profit model is legally binding or subject to future changes. Open AI existed for less than five years before abandoning its nonprofit beginnings. Who is to say this capped profit model couldn’t be changed in the next five years?
Second, even if the capped profit stays put, Open AI would likely need to be the most cumulatively profitable company in history before any profit gets capped.1
🚩 Red Flag No. 3: The Largest Investor Without a Board Seat
The most shocking revelation of the surprise firing was that Microsoft, Open AI’s largest investor. held no formal board seat or board observer rights. With more than $10 billion invested, a 49% ownership in OpenAI’s for-profit subsidiary, and a partnership that involves deep product integrations, Microsoft and Open AI are inextricably linked.
Corporate investments in startups are often made gingerly, as both sides recognize that the signaling effects are different and more substantial. But under normal circumstances, it would be heresy for a 49% investor not to have any role in the boardroom.
In a 2022 survey, 90% of corporate venture investors said they took either a board director seat, board observer role, or both.
I’ve seen no reporting on why Microsoft had no seat on the board, but Open AI’s ability to maintain some independence and some remaining conflicts of interest probably played a part. Most likely, however, the aforementioned first red flag -- Open AI’s odd corporate structure -- prevented Microsoft’s seat at the table.
On the Monday after the attempted coup, Microsoft CEO Satya Nadella, in a spate of media interviews, said that "something has to change" in regards to the board of directors, but stopped short of saying what changes he would like to see — including whether or not Microsoft should have its own seat at the table.
I’d be shocked if Microsoft doesn’t soon end up with a seat on the newly revised board.
🚩 Red Flag No. 4: The CEO with No Equity Stake (Or A Secret One)
In the hours after Sam Altman was ousted as CEO, amid a series of cryptic tweets that the tech world was watching closely, Altman tweeted, “if I start going off, the openai board should go after me for the full value of my shares”
It was a surprising tweet because people were hanging on Altman’s every word and because Altman very famously owns no equity in Open AI. As reported by Semafor, “Altman also made an unusual decision for a tech boss: He would take no equity in the new for-profit entity, according to people familiar with the matter. Altman was already extremely wealthy, investing in several wildly successful tech startups, and didn’t need the money.”
In a clip of Altman’s congressional testimony in May that went viral the weekend after his firing, Altman had, without equivocation, made it clear that he owned no equity in Open AI:
Sen. John Kennedy (R-LA):
Cool. Are there people out there that would be qualified? We’d be happy to send you recommendations for people out there. Yes.
Okay. You make a lot of money. Do you?
Sam Altman:
I make no… I get paid enough for health insurance. I have no equity in OpenAI.
Sen. John Kennedy (R-LA):
Really? Yeah. That’s interesting. You need a lawyer.
Sam Altman:
I need a what?
Sen. John Kennedy (R-LA):
You need a lawyer or an agent.
Sam Altman:
I’m doing this cuz I love it.
https://x.com/andrewmichaelio/status/1658501431366647809?s=20
Theories abound as to how Altman owns a sizable chunk of AI equity despite statements to the contrary.
In a September episode of the All-In podcast, the hosts speculate that Altman, perhaps through a foundation, actually owns the “residual” equity in Open AI once its investors are capped out.
So, does Altman own any part of Open AI or not? I suspect he does, but either answer is a red flag. He either owns equity in the company and is disingenuous when he says he doesn’t, or he doesn’t own equity, which would again illustrate a fundamental misalignment with the company’s employees and investors.
To be clear -- I’m extremely bullish on Open AI and an avid -- paying -- user of its products. It’s not hard to believe that the company’s current reported $90 billion valuation is probably a steep discount to what it will be worth one day.
But, the company’s external positioning -- from its weird corporate structure to “capped-profit” -- has always felt off. From a governance perspective, it was riddled with red flags before the events in recent weeks.
Open AI lost $540 million in 2022, according to reports. Despite having raised $13 billion, I think the company is far from done raising capital. I asked ChatGPT to estimate how much cumulative profit Saudi Aramco -- likely the most cumulatively profitable company ever in absolute terms -- has generated since its inception in 1933. ChatGPT estimated Saudi Aramco cumulative profit at $2.7 trillion. Assuming ChatGPT’s estimate is correct, only if Open AI raises another $14 billion in equity will it need to be the most cumulatively profitable company ever to reach its 100x profit cap.