Greg Brockman
Last reviewed
Sources
No citations yet
Review status
Needs citations
Revision
v3 ยท 4,477 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Sources
No citations yet
Review status
Needs citations
Revision
v3 ยท 4,477 words
Add missing citations, update stale details, or suggest a clearer explanation.
Greg Brockman is an American software engineer and entrepreneur who co-founded OpenAI in December 2015 and serves as its president and chairman of the board. He previously was the first chief technology officer of the payments company Stripe, which he joined as employee number four in 2010 after leaving university.[1][2] At OpenAI he initially held the title of chief technology officer, became president after the organisation's 2019 restructuring into a capped-profit entity, and was the executive who led recruiting of the founding research team, built early engineering infrastructure such as OpenAI Gym, and later helped drive the company's large language model and product programmes including GPT-3, GPT-4, ChatGPT and DALL-E.[1][3][4] Brockman briefly resigned during the November 2023 board removal of Sam Altman and returned days later when Altman was reinstated; he then took a three-month sabbatical in 2024 before returning to lead OpenAI's scaling and infrastructure work tied to the Stargate Project.[5][6][7]
Brockman was born on 29 November 1987 in Thompson, North Dakota, a small town east of Grand Forks, and grew up there.[1] He attended Red River High School in Grand Forks, where he was strong in mathematics, chemistry, and computer science. He won a silver medal at the 2006 International Chemistry Olympiad and was a finalist in the Intel Science Talent Search, both rare achievements for a student from a sparsely populated region of the upper plains.[1] His high school competition record reflected interests beyond the chemistry that he formally trained in; teachers and friends quoted in later profiles described his self-directed work on mathematical puzzles and on the early stages of what would become a career-long interest in computer science.[1]
After high school he took a gap year, during which he began serious self-directed programming work after reading Alan Turing's 1950 essay "Computing Machinery and Intelligence." On his own blog he later described the essay as having articulated a question, whether a machine could think, that came to define the rest of his career.[8] One of his first projects was an attempt to build a weather chatbot that he eventually shelved after concluding that the relevant techniques did not yet exist. The exercise nonetheless taught him how to learn programming on his own, without the structured curricula he would later encounter at university.[8]
Brockman enrolled at Harvard University in 2008 with the intent of double-majoring in mathematics and computer science, but left after roughly a year, citing in retrospect a desire for a more research-oriented programme.[1][8] He then transferred to the Massachusetts Institute of Technology, where he focused on programming language theory and worked on a static buffer overrun detection project with friends; he attended only briefly before dropping out in early 2010.[1][8] On his personal blog he described the decision as the result of an unlaunched Palo Alto startup with an unusually strong team reaching out, the team's quality persuading him to leave school within a few months of arrival. The startup was Stripe, founded by his MIT classmate Patrick Collison and Collison's brother John.[8]
Brockman joined Stripe in 2010 as the company's fourth employee, joining founders Patrick and John Collison and engineer Darragh Buckley while the company was still in stealth and called "/dev/payments".[1][2][9] He worked across the company on backend infrastructure, payments integration, banking partnerships, and product engineering. He has described this period as the source of his views on the relationship between engineering and operations: at a payments company, downtime meant immediate loss of customer revenue, and engineering decisions had unusually direct financial consequences.[9]
In 2013 Brockman became Stripe's first chief technology officer, formalising a role he had effectively been playing.[1] During his tenure the company grew from a handful of people to roughly 250 employees across multiple offices on four continents, processing billions of dollars in payments.[1][9] He has written that his responsibilities expanded to cover not just engineering but cross-cutting company problems, and that about six months before leaving he had deliberately worked to remove himself as a bottleneck on any specific area in order to free himself to take on whatever the company most needed.[9]
He announced his departure from Stripe on his blog on 6 May 2015, writing that he had "topped out the S-curve for what I could learn here about starting a company" and wanted to pursue another project of his own while he still had the freedom to do so.[9] He noted that Stripe continued to thrive after his exit, eventually reaching more than 450 employees without him.[9] In a complementary post titled simply "Greg," Sam Altman, then a personal acquaintance, described Brockman as "the most productive engineer" he knew, praising his recruiting ability, openness to feedback, and willingness to take on hiring, culture, and operational work in addition to engineering. Altman in that post also flagged what he called Brockman's role as "chief optimist," the cofounder responsible for maintaining team morale during downturns.[10]
After leaving Stripe, Brockman approached Altman, then president of Y Combinator, about his goal of working on safe artificial general intelligence. Altman immediately endorsed the idea and mentioned that he had been independently thinking about a "YC AI lab" project; the conversation, on Brockman's account, was decisive in moving him from idea to action.[8][28] Altman invited him to a dinner in August 2015 with Elon Musk, Ilya Sutskever, and several other researchers at the Rosewood Sand Hill hotel in Menlo Park, California, where the group discussed forming a non-profit AI research lab as a counterweight to capability concentrations inside large technology companies.[4][11]
Over the following months Brockman led recruiting of the founding research team. He has described an intense period of personal outreach in which he met candidates one at a time, often over multi-day visits, persuading researchers from Microsoft Research, DeepMind, Google Brain, and several universities to join a then-unnamed organisation with no track record.[4][28] Altman later wrote that this recruiting effort was, in his judgement, the single most important factor in OpenAI's early viability.[10]
OpenAI was publicly announced on 11 December 2015. The launch announcement listed Brockman as chief technology officer and named Sam Altman and Sutskever as the other top leaders. The eleven-person founding cohort also included Andrej Karpathy, John Schulman, Trevor Blackwell, Vicki Cheung, Durk Kingma, Pamela Vagata, Wojciech Zaremba, Musk, and Altman, with an initial $1 billion in committed funding from Musk, Altman, Peter Thiel, Reid Hoffman, Jessica Livingston, Amazon Web Services, Infosys, and YC Research.[11][1] The team formally began work on 4 January 2016, with Brockman's San Francisco apartment serving as the first office until the group moved into space in the city's Mission District.[8][28]
In April 2016 Brockman was the lead author of the technical report introducing OpenAI Gym, a toolkit and benchmark suite for reinforcement learning research.[12] Gym provided a standard Python interface for reinforcement learning environments and a curated collection of benchmarks; it quickly became one of the most widely cited reinforcement learning libraries of the late 2010s and was the project through which many OpenAI engineers first contributed code.[12] Brockman has cited Gym as one of his earliest concrete contributions to the field, written largely while still self-teaching machine learning.[25]
He then led engineering on Universe, released in December 2016, which extended the Gym idea to arbitrary desktop programmes. Universe wrapped any application by driving it through a VNC remote-desktop connection and reading screen pixels back as observations; the initial release included roughly a thousand environments drawn from Flash games, browser tasks, and commercial titles such as slither.io and Grand Theft Auto V. The project was intended to make any computer programme a potential reinforcement learning testbed without requiring source code access.[13]
As Universe wound down in 2017 the team pivoted into the Dota 2 project, which became OpenAI Five. Brockman was among the engineering leads, helping build the distributed reinforcement learning infrastructure used to train the agents on a large cluster of GPUs and CPU workers. He has subsequently said the Dota 2 work was where he first internalised the empirical scaling regularities that would shape later OpenAI research, specifically that gains continued as compute, batch sizes, and model sizes grew.[25] OpenAI Five played professionals at The International 2018 (losing two showcase matches) and in April 2019 defeated the world champion team OG in a televised exhibition; Brockman published a blog post on his own site marking the OpenAI Five Finals event.[14]
In March 2019 OpenAI restructured into a "capped-profit" entity, OpenAI LP, with the original non-profit serving as parent. The structure capped investor returns at 100 times their investment and was justified by the leadership as necessary to raise the capital required for ever-larger compute budgets.[15] Brockman was closely involved in drafting the OpenAI Charter (released April 2018) and in managing the transition; he later told MIT Technology Review that he had "spent a lot of time really poring over them to get them exactly right" when discussing the charter's clauses.[15] He became president of OpenAI as part of the restructuring, with Sutskever assuming the chief scientist role.[1][15] He also became chairman of the board, a position he would hold until November 2023.[17]
Beyond OpenAI itself, Brockman served on the board of the Stellar Development Foundation, the non-profit behind the Stellar blockchain network, between 2017 and 2018, writing a blog post on his motivations for joining.[26]
Brockman remained president and chairman of the board while OpenAI released the GPT-2 language model (2019), GPT-3 (June 2020), the Codex code model (mid-2021), DALL-E (January 2021) and DALL-E 2 (April 2022), and ChatGPT (November 2022).[1] He has said that during this period his focus shifted from purely engineering management to a combination of engineering, product, and partnership work, including the deepening relationship with Microsoft that began with a $1 billion investment in 2019 and grew to multi-billion-dollar commitments through 2023.[15]
He frequently served as a public spokesperson and live demonstrator for these systems. On 14 March 2023 he led the live-streamed launch demo of GPT-4, showing capabilities such as image input, instruction following on a US bar exam question, and on-the-fly generation of a working website from a photograph of a hand-drawn sketch on a napkin. The demonstration was widely covered as evidence of qualitative gains over GPT-3 generation models.[1] He has written on his personal blog about the products' technical and societal implications, including a 2022 post arguing that systems like GPT-3, Codex, and DALL-E 2 had crossed a utility threshold for everyday tasks and that more accessible interfaces were the appropriate next step.[16] ChatGPT, released on 30 November 2022, was that interface in practice and grew to 100 million weekly active users by November 2023.[17]
On 17 November 2023 the OpenAI board, which then included Sutskever, Adam D'Angelo, Tasha McCauley, and Helen Toner, removed Altman as chief executive and informed Brockman that he was being removed from the board chair role but would continue as president.[5][17] According to a public timeline Brockman posted on X, Sutskever sent him a message at 12:19 p.m. Pacific time requesting a call, then sent a Google Meet link at 12:23 p.m. where the news was delivered. The board's public statement said Altman had been "not consistently candid in his communications" with the directors and that the board no longer had confidence in his ability to lead.[17]
Brockman wrote: "Sam and I are shocked and saddened by what the board did today." Within hours he posted the message he had sent to OpenAI employees: "based on today's news, i quit." He added that he was "super proud of what we've all built together since starting in my apartment 8 years ago" and that he remained committed to the mission of building safe AGI.[5][17] Mira Murati, who had been OpenAI's chief technology officer, was named interim chief executive. The simultaneous removal of both Altman and Brockman, the two most public faces of the company, created what reporters described as an immediate governance vacuum.[17]
Over the following weekend Microsoft chief executive Satya Nadella announced that Altman and Brockman would join Microsoft to lead a new advanced AI research group. Inside OpenAI, more than 700 of about 770 employees signed an open letter threatening to follow them unless the board resigned and reinstated Altman; Brockman himself stated that roughly 95 per cent of staff signed.[17] According to The Wall Street Journal's reporting later cited by multiple outlets, Brockman's wife Anna personally appealed to Sutskever to reverse his position; Sutskever publicly recanted his vote on 20 November 2023, saying he regretted his participation in the board's actions.[18][17]
On 21 November 2023 OpenAI announced that Altman would return as chief executive and Brockman would return as president, with a new initial board chaired by former Salesforce co-chief executive Bret Taylor and including Larry Summers and Adam D'Angelo. Neither Altman nor Brockman ever formally joined Microsoft, and the five-day episode is sometimes referred to in coverage as "Shipt-mas" or "the OpenAI weekend."[17] An internal review by the law firm WilmerHale was published on 8 March 2024 and concluded that prior board action did not arise from concerns regarding product safety, finances, or statements to investors; OpenAI simultaneously announced that Altman and Brockman would continue to lead the company and that Altman would rejoin the board alongside three new directors.[19] Reporting by Axios described how Altman and Brockman had "won the narrative" against the directors over the weekend, in part through public timelines and statements posted on X.[27]
On 5 August 2024 Brockman announced on X that he was taking a sabbatical "through end of year," describing it as the first significant break since OpenAI's founding nine years earlier. He wrote that "the mission is far from complete; we still have a safe AGI to build."[20] His departure followed an unusually difficult internal period: the public dispute over the disbanding of the Superalignment team in May 2024, the resignations of Sutskever and Jan Leike, and shareholder questions about OpenAI's pace of safety work.[6]
He returned to the company on 12 November 2024, posting that his "longest vacation" was complete and that he was "back to building OpenAI" with a focus on technical infrastructure problems.[6] During his absence OpenAI raised a $6.6 billion funding round at a $157 billion valuation in October 2024 and lost several senior executives including Mira Murati (chief technology officer), John Schulman (co-founder, head of post-training, who left for Anthropic), and Barret Zoph (vice president of research). On his return he was widely interpreted as a stabilising figure for an engineering organisation that had absorbed multiple high-profile departures.[6]
After his return Brockman established and now leads OpenAI's Scaling group, which is responsible for the infrastructure underlying training of frontier models and the systems that serve ChatGPT to end users.[7] The group consolidates training-stack engineering, datacenter capacity planning, supply chain, and serving infrastructure under a single organisation.[7]
On 21 January 2025 OpenAI, SoftBank, Oracle, and MGX announced the Stargate Project, a new joint venture intended to invest up to $500 billion over four years in AI infrastructure in the United States, with OpenAI holding operational responsibility and SoftBank holding financial responsibility. The announcement event was held at the White House with US President Donald Trump, Altman, SoftBank chief executive Masayoshi Son, and Oracle chairman Larry Ellison; Brockman was present as part of the OpenAI delegation, and OpenAI's blog post on the day named him among the executives leading the effort.[21]
Reporting by Fortune in November 2025 described Brockman as OpenAI's "builder-in-chief" and quoted him saying the Scaling group oversaw "everything from how we train our frontier models to how we run ChatGPT for millions of people."[7] Through 2025 he led negotiations with chip suppliers, including a multi-year supply agreement with AMD announced on 6 October 2025, under which OpenAI agreed to deploy 6 gigawatts of AMD Instinct GPUs across multiple generations, with the first 1 gigawatt of MI450 chips set to come online in the second half of 2026.[29] AMD also issued OpenAI warrants for up to 160 million shares, roughly 10 per cent of the company, vesting against deployment milestones, and AMD's share price closed about 24 per cent higher on the day of the announcement, adding more than $60 billion in market value.[29][7] On the deal, Brockman said: "Building the future of AI requires deep collaboration across every layer of the stack. Working alongside AMD will allow us to scale to deliver AI tools that benefit people everywhere."[29] AMD chief executive Lisa Su credited the agreement to Brockman's "maniacal focus on ensuring there's enough compute in this world."[7] He oversaw the build-out of Stargate data centre sites including the flagship Abilene data center in Texas and additional locations announced through the year in Ohio, New Mexico, Wisconsin, and Michigan.[7][22]
In May 2026 internal restructuring put Brockman in formal charge of OpenAI's product organisation, merging ChatGPT, Codex, and the developer API under a single agentic platform; an internal memo from Brockman wrote that OpenAI would "invest in a single agentic platform and merge ChatGPT and Codex into one unified agentic experience for all." The change consolidated reporting lines under Brockman just days before Google's annual I/O conference and was interpreted by industry analysts as a refocusing of OpenAI's product strategy around an agentic ChatGPT.[23][3]
Brockman became engaged to Anna in 2018, after about a year of dating, and they held a civil ceremony in the OpenAI office on a workday in November 2019. He posted on X that the ceremony was officiated by Ilya Sutskever, with a robotic hand serving as ring bearer; a larger wedding followed at a later date.[24] Anna Brockman became publicly involved in the November 2023 board crisis when, according to The Wall Street Journal's later reporting, she personally appealed to Sutskever to reverse his vote against her husband and Altman.[18] The couple live in San Francisco.[7]
Brockman maintains a personal blog at blog.gregbrockman.com hosted on the Svbtle platform, where he has published essays including "My path to OpenAI," "Leaving Stripe," "#define CTO OpenAI," "Stellar board," and "How I became a machine learning practitioner," the last describing his decision after roughly three years at OpenAI to retrain himself as a hands-on machine learning engineer over a nine-month course of self-study.[16][9][25][28][26] He is an active poster on X under the handle @gdb, where he uses the platform to make many of the company-related announcements that are then summarised by media.[20][24]
| Years | Role | Organisation |
|---|---|---|
| 2010 to 2013 | Engineer (employee number 4) | Stripe[1] |
| 2013 to 2015 | Chief Technology Officer | Stripe[1] |
| 2015 to 2019 | Co-founder and Chief Technology Officer | OpenAI[1][11] |
| 2017 to 2018 | Board member | Stellar Development Foundation[26] |
| 2019 to November 2023 | Co-founder, President, Chairman | OpenAI[1][17] |
| November 2023 (3 days) | Resigned; announced future role at Microsoft | Microsoft (announced, not joined)[17] |
| November 2023 to August 2024 | Co-founder and President | OpenAI[17][20] |
| August to November 2024 | On sabbatical | OpenAI[20][6] |
| November 2024 to present | Co-founder and President; lead of Scaling group | OpenAI[6][7] |
Brockman occupies an unusual position among the OpenAI co-founders: he is neither the chief executive (Altman) nor a research scientist of Sutskever's stature, but he has been credited in journalistic profiles and in Altman's own writing with much of the engineering and organisational backbone of the company.[7][10] His work at Stripe gave him experience scaling production systems through hyper-growth, and contemporaries describe him as bringing that operational discipline to the construction of OpenAI's training stack, its safety and review processes, and most recently the gigawatt-scale data centre programme behind Stargate.[7][22]
The 2023 board crisis, in which more than 700 employees signed a letter threatening to leave unless he and Altman were reinstated, also illustrated the degree to which OpenAI's internal cohesion was bound up in his personal authority. Reporting after the episode quoted multiple OpenAI staff describing Brockman as "the heartbeat" of the engineering organisation.[17][7] His sabbatical in 2024 and subsequent return were interpreted in similar terms: the announcement that he was leaving for the rest of the year prompted immediate speculation about further executive departures, and his post-return work was largely cast as a stabilising effort.[6]
A through-line in his publicly stated views, on his blog and in interviews, is the centrality of compute and engineering. He has argued that algorithmic insight is rarely the binding constraint on AI progress, and that the more important question is how much hardware, electricity, and software engineering an organisation can bring to bear. The Scaling group and Stargate Project are concrete expressions of that view.[7][28]
Profiles of OpenAI's transition to a capped-profit structure (2019) and to a more conventional for-profit (announced 2024 to 2025) have drawn criticism from outside commentators who argue that the change marked a departure from the lab's original non-profit charter. Brockman, as a principal drafter of that charter, has been a target of such critiques, although he has publicly defended the structure as necessary to fund the compute required to pursue AGI.[15][7]
His decision to resign within hours of Altman's firing in November 2023, and the rapid coordination with Altman and Microsoft that followed, has been variously described as decisive loyalty and as effective pressure on the board; reporting by Axios noted the speed with which the pair "won the narrative" over the directors.[27] Critics of the resolution have argued that the board's original concerns, whatever their merit, were never publicly aired in detail and that the WilmerHale review did not, by design, address the underlying disagreements over governance and disclosure that prompted the action.[17][19]
The November 2025 Fortune profile also flagged the financial scale of his Stargate commitments, noting that OpenAI had committed roughly $1.4 trillion in long-term infrastructure spending against reported annualised revenue of approximately $13 billion at the time, and that the durability of those commitments depended on continued rapid revenue growth and on the absence of a macroeconomic shock to the company's lenders and partners.[7] Brockman has publicly defended the strategy as the appropriate response to the scale of compute required by the next generation of frontier models.[7]