Last reviewed
Jun 3, 2026
Sources
11 citations
Review status
Source-backed
Revision
v1 · 1,963 words
Add missing citations, update stale details, or suggest a clearer explanation.
The 2026 OpenAI executive departures were a run of high-profile leadership exits from OpenAI over the course of 2025 and into 2026, capped by a single-day triple departure on April 17, 2026. On that Friday, Kevin Weil, the company's former chief product officer who had moved to lead its science research effort, Bill Peebles, the head of the Sora video team, and Srinivas Narayanan, the chief technology officer of enterprise applications, all announced they were leaving. The exits landed as OpenAI shut down its consumer Sora app and folded its science group, part of what the company internally described as shedding "side quests" to focus on enterprise customers and a forthcoming "superapp." [1][2][3]
The departures drew attention partly because of their timing and partly because of what they said about the shape of OpenAI in 2026. The company that built ChatGPT had spent two years chasing splashy consumer products. By the spring of 2026 it was retrenching around paying business customers, coding tools, and a smaller set of bets, and a number of the people who had led the consumer experiments were on their way out. By that point only a handful of OpenAI's eleven original co-founders were still at the company.
The three exits were announced within hours of one another, mostly through social media posts and internal notes, which is part of why they read as a cluster rather than three unrelated resignations.
Kevin Weil had joined OpenAI in 2024 as chief product officer after running product at Instagram, Twitter, and Planet Labs. In 2025 he handed off the product role and took over OpenAI for Science, an internal research group pursuing AI tools for scientific discovery. His departure came one day after that team shipped GPT-Rosalind, a model aimed at accelerating life sciences research and drug discovery. In his farewell note Weil called it "a mind-expanding two years" and argued that accelerating science would be one of the most positive outcomes of the technology. He did not announce a specific next destination. [1][3]
Bill Peebles joined OpenAI in 2023 and became the public face of Sora, the text-to-video system. He co-authored the diffusion-transformer research that underpinned the product and helped orchestrate the launch of the standalone Sora app in late 2025, which briefly shot to the top of Apple's App Store. In his note to the team, Peebles called building Sora "the honor and adventure of a lifetime" and credited it with igniting "a huge amount of investment in video across the industry." He also offered a pointed argument about how research labs should be run: "Cultivating entropy is the only way for a research lab to thrive long-term," writing that the kind of work that produced Sora needs space away from a company's mainline roadmap. He did not name where he was going next. [2][4]
Srinivas Narayanan, who had spent roughly three years at OpenAI as chief technology officer of enterprise applications, told colleagues internally that he was leaving to spend more time with family, according to reporting by Wired relayed in coverage of the exits. Narayanan had helped ship ChatGPT and the company's API and had grown the applied engineering organization from a few dozen people into a much larger operation. [1][3]
| Name | Role at OpenAI | Departure announced | Stated destination |
|---|---|---|---|
| Kevin Weil | Former chief product officer; led OpenAI for Science | April 17, 2026 | Not disclosed |
| Bill Peebles | Head of the Sora video team | April 17, 2026 | Not disclosed |
| Srinivas Narayanan | CTO of enterprise applications | April 17, 2026 (internal) | Not disclosed; cited family |
None of the three publicly named a new employer at the time of the announcements, and the reporting did not confirm where they landed. That is worth stating plainly, because a lot of the prior 2024 to 2025 OpenAI departures came with an immediate new venture attached, and these did not.
The exits cannot really be separated from what was happening to Sora. OpenAI had announced in the weeks before that it was discontinuing the consumer product. According to OpenAI's own support documentation, the Sora web and app experiences were set to be discontinued on April 26, 2026, with the Sora API following on September 24, 2026. [5][6]
The reasons reported were mostly about money and focus. The Sora app was reportedly losing on the order of $1 million per day in compute costs, and its user base had fallen sharply from a peak of roughly a million to under half that. OpenAI wanted to redirect that compute toward coding tools and enterprise customers, a strategy that looked a lot like the path rival Anthropic had taken. There was also collateral damage: Disney had reportedly committed around $1 billion to a Sora partnership, and according to reporting the company learned the app was being shut down less than an hour before the public announcement. The deal died with the product. [3][6][7]
Internally, OpenAI framed the moves as cutting "side quests," a phrase that recurred in the reporting. The consumer Sora app and OpenAI for Science were both treated as bets the company no longer wanted to carry as standalone efforts. OpenAI for Science, the group behind a discovery platform called Prism and the GPT-Rosalind model, was absorbed into other research teams rather than wound down entirely, according to Weil's own announcement. The throughline was a shift away from consumer moonshots toward enterprise revenue and a planned superapp that would fold ChatGPT, Codex, and other tools into a single package. [1][2][3]
The financial backdrop helps explain the urgency. OpenAI was reportedly running at well over $25 billion in annualized revenue, with enterprise making up more than 40 percent of that, while still projecting losses in the range of $14 billion for 2026. In that context, a consumer video app bleeding a million dollars a day was an easy thing to cut. [3]
The April exits were the latest entries in a long list. OpenAI had been losing senior people steadily for two years, and the pace did not let up in 2026.
The earlier wave is the better known one. Co-founder and chief scientist Ilya Sutskever left in 2024 and started Safe Superintelligence. Longtime chief technology officer Mira Murati departed in late 2024 and launched Thinking Machines Lab, a public benefit corporation, in February 2025. Co-founder John Schulman left in August 2024 to join Anthropic. Chief research officer Bob McGrew and research vice president Barret Zoph left around the same time as Murati. The list of 2024 to 2025 departures also includes chief communications officer Hannah Wong, chief people officer Julia Villagra, and others; by some counts at least a dozen senior leaders left in 2025 alone. [3][8]
Early 2026 brought another reshuffle, this one shaped partly by health. Fidji Simo, who had been hired to run applications and product, took medical leave in April 2026, citing a relapse of postural orthostatic tachycardia syndrome, a neuroimmune condition. Chief marketing officer Kate Rouch stepped back after breast cancer treatment, with Gary Briggs stepping in on an interim basis. Chief operating officer Brad Lightcap moved off the COO role to lead a special projects team of forward-deployed engineers reporting to Sam Altman, with his former duties shifting to Denise Dresser, a former Slack chief executive who had joined as chief revenue officer only months earlier. President Greg Brockman absorbed product oversight during Simo's absence, taking on the ChatGPT, Codex, and Sora portfolios alongside his existing role. [9]
| Departed or shifted | Role | When | Where they went |
|---|---|---|---|
| Ilya Sutskever | Co-founder, chief scientist | 2024 | Safe Superintelligence |
| Mira Murati | Chief technology officer | Sept 2024 | Thinking Machines Lab (Feb 2025) |
| John Schulman | Co-founder | Aug 2024 | Anthropic |
| Bob McGrew | Chief research officer | 2024 | Not specified |
| Barret Zoph | VP of research | 2024 | Joined Murati's lab |
| Fidji Simo | Head of applications/product | April 2026 (medical leave) | Remained, on leave |
| Brad Lightcap | Chief operating officer | April 2026 | Special projects, internal |
| Kate Rouch | Chief marketing officer | April 2026 | Stepped back after treatment |
OpenAI was founded in December 2015 by eleven people: Elon Musk, Sam Altman, Ilya Sutskever, Greg Brockman, Trevor Blackwell, Vicki Cheung, Andrej Karpathy, Durk Kingma, John Schulman, Pamela Vagata, and Wojciech Zaremba. A decade later, only three are still at the company: Altman as chief executive, Brockman as president, and Zaremba as a lead researcher. Musk resigned from the board in 2018; Sutskever and Schulman left in 2024; the rest departed earlier. Some tallies count only Altman and Brockman as remaining, depending on how Zaremba's research role is treated, but the widely reported figure is three of eleven. [10][11]
That number is the part that tends to stick. A company can lose a chief product officer or a video lead and keep running. Losing nine of eleven founders over ten years is a different kind of statistic, and it is the one critics point to when they argue that OpenAI in 2026 is a substantially different organization from the research nonprofit that started in 2015.
Read together, the 2026 departures look less like a crisis and more like a company narrowing its focus, although the two can be hard to tell apart from the outside. The people leaving in April 2026 were largely the ones who had led consumer and research moonshots: Sora, OpenAI for Science, the splashy app launches. The people staying or gaining power, including Brockman and the newly elevated Dresser, were aligned with enterprise revenue and the superapp.
There is a tension in that story that is worth sitting with. OpenAI built its brand and much of its talent pipeline on ambitious, money-losing bets like Sora, and Peebles' parting argument about "cultivating entropy" was essentially a warning that cutting those bets too aggressively could hollow out the research culture that made the company interesting in the first place. At the same time, a firm reportedly losing $14 billion a year and preparing for a possible public offering has obvious reasons to stop funding a video app that bleeds a million dollars a day. Both things can be true. Whether shedding the side quests sharpens OpenAI or quietly drains it of the people who take the big swings is the open question the 2026 exits leave behind.