What happened to Sam Altman?

On Thursday night, Sam Altman showed up at a warehouse in Oakland to talk to a room full of artists about generative AI. It was a busy day; just hours before, he was onstage at the Asia-Pacific Economic Cooperation Summit in downtown San Francisco speaking to world leaders.

At the Oakland warehouse, Altman was his usual self, cheering OpenAI’s work and coming across as relaxed, according to someone in the audience. He couldn’t linger, though. At about 7:40PM he told the room that he was running late to a meeting and bounced.

It would be his last public appearance as the CEO of OpenAI. On Friday, he was suddenly fired by the company’s board of directors for not being “consistently candid in his communications.” Employees found out about the news at the same time as everyone else. Even Microsoft, which has plowed billions of dollars into the company, found out only moments before.

At an emergency internal all-hands meeting Friday afternoon, I’m told that nothing more was said about why Altman was pushed out, though I expect him to elaborate on that soon. Employees, many of whom Altman was personally involved in recruiting to OpenAI, are shell-shocked, to say the least.

OpenAI’s corporate structure is highly unusual given that it started as a nonprofit and then created a for-profit subsidiary that has since become the fastest-growing tech startup in history. Unlike traditional companies, the board of OpenAI isn’t tasked with maximizing shareholder value. Its stated mission is to ensure the creation of “broadly beneficial” artificial general intelligence, or AGI, whatever that means. No one on the board holds equity in OpenAI, and Altman has been clear that, aside from indirect investments, he doesn’t either.

So, who exactly fired Altman? After his departure and the accompanying resignation of co-founder and board chair Greg Brockman, OpenAI’s board consists of: co-founder and chief scientist Ilya Sutskever; Quora CEO Adam D’Angelo; former GeoSim Systems CEO Tasha McCauley; and Helen Toner, an alum of Open Philanthropy and the director of strategy at Georgetown’s Center for Security and Emerging Technology.

Like Kara Swisher, I’ve heard that Sutskever, who leads OpenAI’s research team — and not the applied side of the house that brings products to market — was instrumental in the ousting of Altman and that this all went down very recently. Just last week, Altman revealed a plan to create a store of custom AI bots, pitting the company’s interests as a budding consumer platform against the developers it is already selling its AI tools to.

Altman was recently talking to Jony Ive about doing something in AI and consumer hardware, though no company has been formed and I don’t know if the two are still talking. I do know that their talks were serious enough to gauge interest in potential recruits, though, which suggests that Altman has been itching to do something else for a bit.

Logic says that, ultimately, this was a battle over two warring factions. Sutskever, who has been steadily beating the drum about the dangers of superintelligent AI, seems to have won. That suggests we are about to see OpenAI enter a decidedly different, less commercial era.

With the main part of the US v. Google antitrust trial now over, the biggest takeaway is that Apple takes a staggering 36 percent of ad revenue from searches in Safari, which is remarkably a higher take rate than the App Store.

This fact is in direct contradiction with Google’s overarching argument being that Apple chooses it because it’s the best search engine. You don’t need to give up 36 percent of your revenue if you’re simply the best. You do it because you’re afraid of losing critical distribution and can afford it. Not only has this deal ossified the search engine market for over a decade, but it has kept Apple from seriously competing in search.

Given how critical the iPhone is as a computing platform, I sincerely hope Judge Amit Mehta decides to end this anti-competitive arrangement once and for all.

Notable updates to topics I’ve covered in previous issues:

Back in early August, I wrote about the creation of the San Francisco Compute Group, which was planning to offer short-term H100 rentals to startups and academics who can’t afford what the large cloud providers charge.

This week, one of the project’s leaders, Evan Conrad, announced that the company is officially online and “the lowest-cost provider of H100 training clusters in the world.” (If you live in San Francisco, you may spot a flier for the new service while you’re out and about like my friend Casey Newton did.)

My notes on what else happened in the tech industry recently:

Some interesting career moves I noticed this week:

I’ll be back the week after Thanksgiving. In the meantime, send me your feedback, story ideas, and ideas for how to improve this newsletter next year. Thanks for subscribing.