What Ilya Saw: Mira Murati’s Screenshots About Sam Altman
Before we get to today’s column, OpenAI is continuing its trend of diversifying its cloud compute providers away from Microsoft. Monday morning, the ChatGPT maker announced that it has signed a seven-year, $38 billion cloud deal with Amazon Web Services.
OpenAI will immediately start using compute from AWS, with the goal to deploy all capacity before the end of next year, with the option to use even more compute into 2027 and beyond. That's good news for AWS, bringing it into line with its smaller cloud firms Microsoft and Google, both of which are already dealing with OpenAI.
On to today’s column…
It’s tempting to file away the events of November 2023—l’affaire Altman—into the bowels of ancient history, given everything that has transpired since then.
But all of the main characters in the firing and rehiring of OpenAI CEO Sam Altman are still very much around and in positions of power. They’re also still providing new information about what, exactly, happened in those frenetic days—and that information sometimes runs counter to the recollections of the other characters.
For instance, Rocket on Friday reported on an October deposition of OpenAI co-founder Ilya Sutskever, who was on the OpenAI board and led the firing of Altman but has been mum about the episode ever since. Sutskever is now CEO of his own AI developer, which is worth about $32 billion on paper, and he likely also is a billionaire on paper from his OpenAI stake alone.
The phrase “what Ilya saw” has been a running meme on X since Sutskever pushed out Altman, with people in and around the artificial intelligence field speculating that Sustekver saw the beginnings of sentient AI and that it prompted him to act against Altman. I also reported at the time that Sutskever was instrumental in a secret breakthrough, Q*—later called Strawberry and now simply known as “reasoning”—which added fuel to the speculative Ilya meme.
If there’s one thing Sutskever made clear to lawyers representing Elon Musk in the deposition, however, it’s that much of the evidence he presented to the board to push out Altman came from OpenAI’s then-chief technology officer, Mira Murati.
Specifically, he said he used a number of screenshots from Murati to produce a memo to OpenAI’s independent directors to convince them to fire Altman. (The three independent directors had asked him to pull together the memo to present evidence about Altman after they spoke to him about the issue and his own distrust of Altman, he said.)
The screenshots and other information from Murati included allegations that Altman consistently lied to colleagues, undermined executives and pitted them against each other.
There’s also a reference to Murati’s allegation that Altman retaliated against her in “subtle” ways for giving him critical feedback.
“I fully believed the information that Mira was giving me,” Sutskever said.
Interestingly, Murati, in a note to OpenAI staff in March 2024, tried to distance herself from the firing, writing, “I never reached out to the board to give feedback about Sam” and that she only did so when “individual board members reached out directly to me for feedback about Sam.”
OK, but Sutskever’s comments suggest he might not have acted against Altman without Murati, though he had been thinking about the right time to push out Altman for an entire year before going to the board, he said.
Upon Altman’s firing, Murati took the interim CEO title only to backtrack and support Altman about a day later, which seems to have left the board scratching its head. (She abruptly left OpenAI in September 2024 and now has her own AI firm, worth $10 billion on paper, and hired a number of OpenAI researchers.)
In another notable moment from the deposition, Sutskever mused about what kind of person should be in charge of artificial general intelligence that can handle most of the world’s economic work.
Sutskever said he’s come around to the view that it “would be like choosing between different politicians.”
He added: “That‘s how the world seems to work. I think it’s very—I think it‘s not impossible, but I think it’s very hard for someone who would be described as a saint to make it.” He added, without elaborating: “I think it's worth trying.” (Sutskever also said the board’s firing process was “rushed.”)
Since his firing, Altman has indeed shown he is a gifted politician and leader, managing AI talent battles, deals with rival firms such as Nvidia and AMD, and settling a longtime tussle with Microsoft, which last week agreed to let OpenAI restructure itself so it can go public.
As for Musk, the xAI CEO’s lawsuit against OpenAI produced the Sutskever deposition. Musk, like Altman and Sutskever, also co-founded OpenAI and Musk funded it when it was only a nonprofit. His suit claims Altman and other OpenAI leaders betrayed him by starting a for-profit unit controlled by the nonprofit board where Altman and Sutskever (but not Musk) were directors.
The case could go to trial early next year. Selfishly, we hope it produces even more revelations about what went down in 2023. According to court documents, Murati was scheduled to sit down for her own deposition with Musk’s lawyers Sunday.
OpenAI’s Altman: Go Ahead, Short Our Stock!
OpenAI CEO Sam Altman sounded exasperated when Altimeter Capital founder—and OpenAI shareholder—Brad Gerstner asked him the question that Gerstner said was “hanging over the market”: how a company generating $13 billion in revenue this year would pay for the $1.4 trillion in computing capacity that Altman has said the company is on the hook for.
“Brad, if you want to sell shares, I’ll find you a buyer…I just—enough,” Altman said on Gerstner’s podcast.
“There are not many times that I want to be a public company, but one of the rare times it’s appealing is when those people are writing these ridiculous, ‘OpenAI is about to go out of business,’” Altman said. “I would love to tell them they could just short the stock and I would love to see them get burned on that.”
That’s an unusual reason to want to go public, a prospect Altman said he assumed “would happen someday,” though he dismissed a recent Reuters report that it could take place in 2026 or 2027.
Altman, in goading short-sellers, is keeping company with his frequent antagonist Elon Musk. The Tesla CEO has openly mocked investors who bet that stock in the electric car maker would plunge. At one point Musk even sold red satin “short shorts.”
The OpenAI CEO said the naysayers had it wrong because OpenAI is “doing well more revenue” than $13 billion and he suggested it could even get to $100 billion in revenue in 2027.
The revenue pace indicates the startup will hit its own projections of $13 billion in total revenue this year—and could move up by a year its expectations for hitting $100 billion in revenue, which we’ve previously reported it anticipated reaching in 2028. (For context, it hit $12 billion in annualized revenue, or the last month’s revenue times 12, in July, up from $6 billion in January.)
Altman acknowledged one attraction to an IPO would be to give individual investors a chance to own shares in the company. It could also raise a lot of money to pay for its computing needs.
But “it would be better to fund the company with revenue growth,” he said.—Laura Mandaro