OpenAI Faces Major Leadership Exodus Amid Strategic Shift

OpenAI is undergoing a critical strategic transformation.

Author: Chen Junda

Source: Zhixidong

Zhixidongxi reported on July 13 that, over the past two days, the direct confrontation between Apple and OpenAI has become one of the most talked-about topics in the global tech community. Apple has formally sued OpenAI and two former Apple employees, accusing them of stealing trade secrets and poaching key talent, bringing the competition over personnel and patents into the open.

Amid the ongoing escalation of this controversy, OpenAI has also seen a series of major leadership changes. Last week alone, three senior OpenAI executives announced their departures, including Johannes Heidecke, head of the safety system; Joshua Achiam, chief futurist; and Fidji Simo, OpenAI’s second-in-command and CEO of AGI deployment.

This is not an isolated personnel change. Since April this year, at least 11 senior executives and key project leads at OpenAI have departed, and two executives have undergone role adjustments, affecting critical business areas including AI applications, enterprise markets, AI infrastructure, video generation, AI hardware, AI for Science, chips, and security.

These changes first became evident among OpenAI’s C-level executives. In addition to Fidji Simo stepping down due to health reasons, Brad Lightcap, OpenAI’s long-time COO responsible for operations and commercialization, has been reassigned to lead a newly established joint venture focused on the enterprise market; meanwhile, OpenAI’s CFO, Sarah Friar, has seen her role in key financial decisions diminished, and she has even been excluded from certain critical meetings.

In addition, several key leaders have departed, including Joanne Jang, head of OpenAI Labs; Bill Peebles, lead of Sora; Kevin Weil, head of OpenAI for Science; and Srinivas Narayanan, CTO of B2B applications. The three leads behind OpenAI’s compute project “Stargate”—Peter Hoeschele, Shameez Hemani, and Anuj Saharan—have also left and joined Meta.

▲13 personnel changes at OpenAI from April this year until now (compiled by Zhixidong)

For many tech companies rushing to go public, changes in executive leadership shortly before the IPO are not uncommon, but the scale and breadth of the recent adjustments at OpenAI are rare even in Silicon Valley.

These intense personnel changes are more of a consequence than a starting point. Over the past several months, OpenAI has already executed a more critical internal shift, including reprioritizing business goals and consolidating and narrowing resource allocation. Exploratory projects previously spread across multiple cutting-edge areas—such as video generation, AI hardware, and AI research—have been rapidly downgraded or shut down. In contrast, the core focus on commercialization and the enterprise market has been strengthened more than ever before.

Against this backdrop, the departures and role changes among OpenAI’s senior leadership are no longer just ordinary personnel news, but a window into understanding OpenAI’s current direction.

In this transformation at OpenAI, a representative figure is Fidji Simo. She joined in 2025 as CEO of OpenAI Applications (a role later renamed to CEO of AGI Deployment), serving as the core leader responsible for OpenAI’s application layer and commercialization efforts.

Simo was a key figure in driving OpenAI’s shift from a research-focused to a product- and revenue-focused approach, and she played a significant role in this strategic realignment to streamline operations and focus on core priorities.

AGI deployment CEO Fidji Simo

Simo is renowned for her strong commercial acumen. She joined Facebook from eBay in 2011 and rose from head of advertising to Vice President and General Manager of the Facebook App, overseeing core businesses including Feed, Groups, Video, and Advertising. During her tenure, she drove profitability and significant growth for Facebook’s mobile platform.

After leaving Facebook in 2021, she became CEO of Instacart, leading the company to consecutive profitability and a successful IPO, breaking a more than one-year drought of large tech IPOs on U.S. stock markets at the time.

▲Fidji Simo

Whether at Facebook, Instacart, or OpenAI, Simo’s core mission has always been to “make money,” and there are only two ways to do so: increasing revenue and reducing costs.

To enhance its self-sustaining capabilities, OpenAI needs to convert its technology into sustainable profits. Simo personally brought in a group of executives with extensive commercialization experience:

Dave Dugan: Former head of Meta’s advertising business, responsible for advertising sales;

Denise Dresser: Former CEO of Slack, appointed Chief Revenue Officer;

Ashley Alexander: Former Head of Product at Instagram, responsible for the health business;

Glen Coates: Former Head of Core Products at Shopify, responsible for the app platform.

▲The "Executive All-Star Team" introduced by Fidji Simo

Meanwhile, Simo has been deeply involved in the fundraising efforts previously led by Altman, frequently attending investor roadshows and becoming the primary point of contact between OpenAI and the outside world. Since August last year, OpenAI’s CFO, Sarah Friar, has reported directly to Simo instead of Altman—an unusual arrangement, as the CFO typically reports directly to the company’s CEO.

Simo is also recruiting deputies for her senior executives; for example, in January this year, she hired Cynthia Gaylor, former CFO of DocuSign, to oversee the company’s finance and investor relations under Friar.

To cut costs, Simo lowered the priority of side projects, including shutting down the Sora app, scaling back the Atlas browser, and reducing efforts in smart hardware. These decisions may be closely related to the departures of several team leads at OpenAI in April:

Joanne Jang, Head of OpenAI Labs

Joanne Jang leads OpenAI Labs, which primarily explores AI interaction and model behavior, with hardware being one of its core focus areas. She embodies OpenAI’s internal “research lab culture,” emphasizing long-term exploration and future product forms over short-term commercial returns.

However, hardware恰好 falls within Simo’s side project list. Her departure is, in a sense, a microcosm of how the “research exploration” faction within OpenAI has been marginalized under the current commercialization strategy.

▲Joanne Jang (Source: LinkedIn)

Bill Peebles, lead of the Sora team

Another key figure affected is Bill Peebles. Peebles was involved throughout virtually the entire development process of Sora from 0 to 1 and is truly the soul of the Sora project.

In his resignation letter, Peebles bluntly expressed his views on OpenAI’s reduction of side projects: “In life, it’s easy to fall into the trap of focusing only on the most important things, but for a research lab, maintaining diversity in exploration is the only path to long-term success.”

▲ Bill Peebles wearing a jacket with the word "Sora" printed on it (source: Bill Peebles' personal website)

Kevin Weil, Vice President of OpenAI and Head of the OpenAI for Science team

Kevin Weil’s shift better reflects the adjustment in OpenAI’s power structure following Simo’s arrival. Weil previously served as OpenAI’s Chief Product Officer and was a key figure bridging research with product execution. However, since Simo joined in 2025, their responsibilities have begun to overlap significantly.

In October last year, he stepped down as Chief Product Officer at OpenAI to focus on OpenAI for Science. In April of this year, this team of approximately ten people was integrated into the Codex team, and the AI research application Prism, which the team had been responsible for, lost its status as an independent product.

▲ Kevin Weil (Source: LinkedIn)

Hannah Wong and TBPN

Simo is still driving a transformation of OpenAI’s public facade. Following the departure of Chief Communications Officer Hannah Wong earlier this year, Simo spearheaded an unexpected decision: to directly acquire the tech podcast company TBPN and appoint its founder as a communications advisor to reshape OpenAI’s external narrative.

This move was almost absurd, as it was announced in early April, leading employees to initially think it was a delayed April Fools' joke.

The significance of this move lies in the possibility that OpenAI may no longer be content merely with technological leadership, but now seeks to actively control market narratives, media influence, and public perception. Some current employees believe that Simo’s arrival has shifted the internal power balance at OpenAI from the research team to the product team, with power potentially having moved from Altman to Simo.

Simo’s influence at OpenAI has been steadily growing, making her sudden departure all the more unexpected. She suffers from postural orthostatic tachycardia syndrome, a condition that commonly causes fatigue and dizziness. According to insiders, she has occasionally missed meetings over the past few weeks, just as OpenAI was urgently addressing the new threat posed by Anthropic.

Anthropic has surpassed OpenAI in both valuation and ARR. Although OpenAI’s new Chief Revenue Officer, Denise Dresser, has questioned the accuracy of Anthropic’s data, the latter’s rapid growth in the B2B market is undeniable.

Simo’s current health condition may no longer keep pace with the intense and fast-paced business momentum at OpenAI. Simo’s responsibilities have been distributed among President Greg Brockman, Chief Financial Officer Sarah Friar, and Chief Strategy Officer Jason Kwon.

Compared to major competitors like Google and Anthropic, OpenAI still holds an advantage in user numbers in the consumer market. However, it must be acknowledged that the AI industry as a whole has yet to identify any other large-scale, effective consumer monetization model beyond subscriptions. OpenAI’s much-anticipated advertising business has faced strong user backlash, with advertisers also questioning its limited effectiveness.

Therefore, OpenAI is aggressively expanding into the enterprise market, particularly in programming—a domain primarily dominated by Anthropic. Brad Lightcap, who recently transitioned from COO to head of special projects, was previously OpenAI’s top sales leader in driving enterprise market growth.

Brad Lightcap, former COO and Head of Special Projects at OpenAI

Lightcap, aged 34, graduated from Duke University and worked as an investment banking analyst at J.P. Morgan before joining the tech company Dropbox as a strategic finance analyst. In 2016, his path crossed with Altman’s when he became an investor at the incubator YC, where Altman was then CEO.

▲Brad Lightcap

In 2018, Altman asked Lightcap to help find someone to manage OpenAI’s finances; at the time, OpenAI had only about forty employees and no clear business model. Lightcap interviewed more than twenty candidates, but none were willing to take on this “hot potato.” He then decided to take the role himself, becoming OpenAI’s first finance employee and serving as its CFO until 2022.

With the surge in popularity of ChatGPT, OpenAI needed to turn it into tangible revenue. Lightcap took on this mission, transitioning to COO of OpenAI to oversee operations, finance, and strategic partnerships. During his tenure, Lightcap rapidly expanded OpenAI’s original market development team of just 50 people to over 700, building a comprehensive team encompassing sales, customer success, developer relations, and strategic partnerships.

Moreover, Lightcap was the driving force behind OpenAI’s global expansion. He led the initiative to establish offices in Brazil, India, Australia, Japan, and other regions, aiming to build local teams capable of deploying AI worldwide. He believes: “OpenAI’s responsibility is not only to build tools, but also to become the world’s foremost expert in deploying them.”

▲Brad Lightcap attended an event at OpenAI's Japan office

During the 2023 November upheaval at OpenAI that shook the tech world, Lightcap served as the stabilizing force for the organization. He revealed to Bloomberg that during the most turbulent 48 hours, he personally reached out to more than 40 key clients, calmed employee concerns, and ensured OpenAI did not suffer major client losses during the power vacuum.

About a year ago, his responsibilities expanded further to focus on business strategy, key partnerships, infrastructure, and operations. Today, faced with Anthropic’s strengths in programming and the enterprise market, his primary focus has shifted to aggressively driving deep integration of enterprise-grade products, aiming to fill the gap OpenAI has in tools for “deeply embedding into enterprise systems,” such as AI programming assistants.

The adjustment in April, in a sense, freed Lightcap from day-to-day operations at OpenAI, allowing him to focus on expanding into the enterprise market—potentially his strongest area.

The special project overseen by Lightcap is DeployCo, a joint venture established by OpenAI and several U.S. private equity funds, with a post-investment valuation of $10 billion. As its name suggests, DeployCo’s mission is to promote and deploy OpenAI’s enterprise products within the companies invested in by these funds. Investors hold significant influence over software and AI budget decisions at their portfolio companies, enabling OpenAI to more easily acquire these clients.

B2B Application CTO Srinivas Narayanan

OpenAI’s personnel changes in the enterprise market don’t end there. In April, Srinivas Narayanan, CTO of OpenAI’s B2B applications, announced that he would be concluding his more than three-year career at OpenAI this week.

Narayanan was a key figure in OpenAI’s transition from a model company to a platform company, primarily responsible for building the engineering infrastructure and driving the technical commercialization of the OpenAI API, and previously reported to Lightcap.

When asked about his reason for leaving, Narayanan cited OpenAI’s recent and upcoming products.

▲Srinivas Narayanan

In addition to the changes in COO and the CEO responsible for AGI deployment, the subtle relationship between OpenAI’s CFO, Friar, and Altman is also worth noting.

Friar joined OpenAI in June 2024, and her arrival was seen as a landmark event signaling OpenAI’s preparation for an IPO. Prior to joining OpenAI, Friar served as CFO and CEO of Square and Nextdoor, respectively, leading both companies through their IPOs and bringing extensive capital markets experience. During her tenure at these companies, she also spearheaded acquisitions worth billions of dollars.

However, Friar and Altman’s collaboration at OpenAI appears to have been strained: Altman aims to complete OpenAI’s IPO in the fourth quarter of this year, but Friar has reportedly told her colleagues at OpenAI that she does not believe the company is ready for an IPO this year, as the necessary procedures and organizational preparations have not yet been completed.

▲ Friar and Altman together (source: Bloomberg)

This CFO is also concerned about Altman’s plan to spend $600 billion over the next five years. She is unsure whether OpenAI will need to invest heavily in AI servers in the coming years, or whether it can sustain this spending commitment if its revenue growth slows.

Some insiders revealed that Altman has begun excluding Friar from certain discussions related to OpenAI’s finances, such as communications with OpenAI’s top investors regarding server expenses. One participant noted that Friar’s absence was awkward, given her prior involvement in communications on the same topic.

Another person who attended the senior OpenAI meeting with Altman also found it unusual that Friar was not invited, as the meeting involved discussions of significant financial decisions.

Altman and Friar have vastly different personalities and backgrounds. A person closely involved with both executives revealed that Friar’s job was difficult: “She worked for a founder with grand ambitions who wanted to push the boundaries as far as possible on capital spending.”

In November last year, in an interview with The Wall Street Journal, Friar said, “I did receive a lot of prompts and nudges from Sam, but they motivated me more than they caused anxiety.” At the time, she also said that an IPO was “not currently on the agenda,” as OpenAI was still working to “get the company into a state of continuous scaling.”

Friar's concerns are not unfounded. The Information, citing informed sources, reported that OpenAI's gross margin last year fell below expectations due to substantial compute expenses.

Due to funding difficulties, large-scale computing projects with deep involvement from OpenAI, such as "Stargate," have been forced to slow down, with subsequent expansion plans for the data center campus in Texas, USA, and the construction plan for the data center campus in the UK both put on hold.

▲ "Stargate" project data center campus in Texas

In mid-April, the three team leads responsible for OpenAI’s Project Stargate left and joined Meta. With its deep pockets, Meta may take over the subsequent expansion of OpenAI’s Project Stargate in Texas to meet Meta’s AI computing needs.

After slowing its own project development, OpenAI shifted to leasing more servers from Microsoft, Oracle, Amazon, and other cloud service providers. Friar agreed that deals with cloud partners are a way to keep the balance sheet lighter.

Friction between Friar and Altman is not uncommon in Silicon Valley; similar cases include the conflict between Airbnb’s CEO and CFO from 2017 to 2018, which ended with the CFO’s departure. Interestingly, Airbnb’s founder and CEO, Brian Chesky, maintains an informal advisory relationship with Altman.

In addition to the leaders of the core businesses, OpenAI has recently seen significant personnel changes across other product and business lines.

Chief Futurist Joshua Achiam

In July this year, Joshua Achiam, a founding member and chief futurist at OpenAI, posted a farewell message on X, concluding his nearly nine-year tenure at OpenAI.

Within OpenAI, Achiam is widely regarded as a core figure steadfastly defending the organization’s safety mission, but he has frequently publicly criticized the broader AI safety community, sparking considerable controversy.

In his resignation letter, Achiam stated that he wishes to continue working on AI safety beyond frontier AI companies.

OpenAI’s safety research team has actually been in a state of constant turmoil. In 2018, Achiam clashed with Elon Musk, then a key investor and co-founder of OpenAI, over safety-related issues.

In 2024, OpenAI’s Superalignment team, originally led by OpenAI co-founder and chief scientist Ilya Sutskever, also disbanded, and OpenAI co-founder John Schulman, who succeeded Sutskever as head of the Superalignment team, soon departed as well.

Clive Chan, a veteran of OpenAI's in-house chip project

In June this year, Clive Chan, OpenAI's self-developed chip and "Employee #002," announced on X that he had left OpenAI and joined Anthropic.

Clive Chan witnessed the entire journey of OpenAI’s in-house chip project, from its early formation to its current progress, but he says he has never been able to shake the urge to once again climb a new mountain from the base—leading him to join Anthropic.

Anthropic is also advancing its own chip development project, and Clive Chan’s addition brings valuable firsthand experience.

Johannes Heidecke, Head of OpenAI's Security System

According to Wired, last week, Johannes Heidecke, head of OpenAI’s safety system, informed OpenAI employees that he is about to leave the company.

Heidecke joined OpenAI's safety team in 2021 and assumed the role of head of safety systems in 2024, succeeding Wang Li, the former head of safety systems at OpenAI.

Heidecke’s departure occurred after OpenAI restructured to integrate its safety and research teams. According to a memo sent by OpenAI’s Chief Research Officer, Mark Chen, to OpenAI employees, OpenAI’s safety team will now report to Mia Glaese, Vice President of Research, who has also been promoted to Vice President of Research and Safety; Saachi Jain, who has prior experience leading safety-related teams, will serve as interim head of safety systems and report to Mia Glaese.

Regarding this adjustment, Mark Chen explained that the pace of model training and release has significantly accelerated, making security coordination challenges greater than ever, thus requiring tighter integration.

Simo’s temporary departure, Lightcap’s reassignment, Friar’s marginalization, and the successive exits of frontline technology and product leaders paint a picture of an OpenAI reshaping its power structure and resource allocation. Today, OpenAI resembles a high-speed commercial machine more than a loose, idealistic research lab.

For OpenAI, the tension between short-term commercialization and long-term technological breakthroughs will not be fully resolved by these adjustments. Cutting "side projects" ahead of the IPO may boost current efficiency and improve the balance sheet, but it could also reduce their buffer against future uncertainties.

Achieving a dynamic balance between capital constraints, intensifying competition, and technological exploration will directly impact OpenAI’s competitive position and growth potential in the next phase.

This article is from the WeChat public account "Zhi Dongxi" (ID: zhidxcom), authored by Chen Junda, and is published with permission from 36Kr.