Mark Zuckerberg
Last reviewed
May 31, 2026
Sources
23 citations
Review status
Source-backed
Revision
v1 ยท 2,236 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Last reviewed
May 31, 2026
Sources
23 citations
Review status
Source-backed
Revision
v1 ยท 2,236 words
Add missing citations, update stale details, or suggest a clearer explanation.
Mark Zuckerberg is an American technology entrepreneur who co-founded Facebook in 2004 and serves as chairman, chief executive officer, and controlling shareholder of Meta Platforms, the company that owns Facebook, Instagram, and WhatsApp. Since the early 2010s he has directed one of the largest corporate investments in artificial intelligence, including the research lab Facebook AI Research (later renamed FAIR), the open-weight Llama family of language models, and tens of billions of dollars in annual spending on AI data centers. In 2025 he reorganized Meta's AI work around a goal he calls superintelligence, creating a unit named Meta Superintelligence Labs and recruiting researchers from rival companies in a widely reported talent contest. [1][2][3]
Zuckerberg has become one of the more visible advocates for releasing AI model weights openly. He has argued that open models will spread the benefits of AI more widely and keep the technology from being controlled by a small number of firms, a position that has drawn both support and criticism from researchers, policymakers, and competitors. [4][5]
Mark Elliot Zuckerberg was born on May 14, 1984, in White Plains, New York. He attended Phillips Exeter Academy and then enrolled at Harvard College in 2002, where he studied psychology and computer science. [6][7]
On February 4, 2004, while a sophomore, he launched a social networking site then called Thefacebook with fellow students Eduardo Saverin, Andrew McCollum, Dustin Moskovitz, and Chris Hughes. The site spread quickly across universities, and Zuckerberg left Harvard during his second year to work on the company full time. He became its chief executive and has held that role since. Facebook held its initial public offering in 2012. [6][7]
On October 28, 2021, Zuckerberg announced that the parent company would change its name from Facebook to Meta Platforms, a rebrand he tied to a long-term bet on immersive computing he called the metaverse. The renamed company kept Facebook, Instagram, and WhatsApp as products. Through a dual-class share structure that gives his stock extra voting weight, Zuckerberg retains majority voting control of the company. As of late 2025 Forbes estimated his net worth at roughly 220 billion dollars, placing him among the wealthiest people in the world. [6][8]
Meta's formal investment in artificial intelligence dates to December 2013, when Zuckerberg and the company recruited New York University professor Yann LeCun to build and lead a new lab, Facebook AI Research. LeCun, who later shared the 2018 Turing Award for his work on deep learning, served as the lab's founding director and then as Meta's chief AI scientist. The lab, later shortened to FAIR, stated a mission of advancing the state of the art in artificial intelligence through open research, and it published widely cited work in computer vision, natural language processing, and self-supervised learning, including the PyTorch framework. [9][10]
In February 2023 Zuckerberg announced a new top-level product group focused on generative AI, organized under chief product officer Chris Cox, to bring the technology into Meta's consumer apps. That unit produced features such as the Meta AI assistant built into WhatsApp, Messenger, and Instagram. The reorganization separated longer-horizon research at FAIR from product-oriented model development, a division that later became a source of internal tension over computing resources. [11][12]
Meta released the first version of its large language model family, Llama, in February 2023, followed by Llama 2 in July 2023, Llama 3 in 2024, and Llama 4 in April 2025. Rather than offering the models only through a paid interface, Meta published the trained weights for download, allowing developers to run and modify the models on their own hardware. By mid-2024 Meta reported that Llama models had been downloaded hundreds of millions of times. [13][14]
Zuckerberg has framed this approach as a deliberate strategy. On July 23, 2024, alongside the release of Llama 3.1, including a 405-billion-parameter model the company described as the first frontier-level open-source AI model, he published an essay titled "Open Source AI Is the Path Forward." In it he wrote, "I believe that open source is necessary for a positive AI future," and drew a comparison to how the open-source operating system Linux eventually overtook closed versions of Unix. [4][14]
Meta's use of the term open source has been contested. The Open Source Initiative and several researchers have noted that the Llama license places restrictions on use and does not release training data, so the models are more accurately described as open weight than open source. Meta itself referred to Llama 4 as open weight in its official announcement. [5]
Under Zuckerberg, Meta sharply increased spending on the computing hardware used to train and run AI models. In April 2025 the company raised its capital expenditure guidance for the year to between 64 billion and 72 billion dollars, citing competition with OpenAI and Google. On July 14, 2025, Zuckerberg said Meta would invest hundreds of billions of dollars in AI infrastructure over the coming years and that the company had the capital to do so from its core advertising business. [3][15]
He described several large data-center projects. The first, named Prometheus, was expected to come online in 2026 as what he called a multi-gigawatt facility, with an early site near New Albany, Ohio. A second, named Hyperion, in Richland Parish, Louisiana, was planned to scale up to about five gigawatts of power over several years. Zuckerberg said individual clusters were large enough that one would cover a significant part of the footprint of Manhattan. [3][15]
In the first half of 2025 Zuckerberg moved to reorganize Meta's AI efforts around what he termed personal superintelligence. The shift followed a period in which Meta's planned high-end Llama model, internally called Behemoth, was reported to have underperformed and been delayed. [1][16]
On June 12, 2025, Meta announced an investment of about 14.3 billion dollars in the data-labeling company Scale AI in exchange for a 49 percent stake, structured so that Meta did not hold voting control. As part of the arrangement, Scale AI co-founder and chief executive Alexandr Wang joined Meta, and Scale named strategy executive Jason Droege as its new chief executive. [2][17]
On June 30, 2025, Zuckerberg announced the creation of Meta Superintelligence Labs in an internal memo. He named Wang as Meta's first chief AI officer and said former GitHub chief executive Nat Friedman would help lead work on AI products and applied research. Investor Daniel Gross joined the leadership group the following month. The new organization brought together Meta's foundation-model teams, a new research group later referred to as TBD Lab, the FAIR research unit, and AI infrastructure teams under one structure. [1][18]
The buildout coincided with an unusually public competition for AI talent. OpenAI chief executive Sam Altman said in June 2025 that Meta had approached OpenAI employees with signing bonuses of around 100 million dollars. Meta executives disputed that characterization, with chief technology officer Andrew Bosworth saying that very large packages applied only to a small number of senior roles and were structured as multi-year compensation rather than simple signing bonuses. Reports described Zuckerberg personally recruiting candidates with offers spanning a wide range of values. [2][19]
The reorganization changed the older research culture. In October 2025 Meta laid off about 600 employees from its AI division, with cuts affecting FAIR and infrastructure teams while sparing the newly hired TBD Lab group, leaving Superintelligence Labs with just under 3,000 people. On November 19, 2025, Yann LeCun confirmed he would leave Meta after roughly twelve years to start his own company focused on world-model architectures rather than large language models. [20][21]
Zuckerberg has presented open release of AI models as both a commercial strategy and a public good. In his 2024 essay he argued that open models let a broad ecosystem of developers build compatible tools, prevent Meta from being locked into a competitor's closed system, and, in his words, "ensure that more people around the world have access to the benefits and opportunities of AI, that power isn't concentrated in the hands of a small number of companies, and that the technology can be deployed more evenly and safely across society." He framed AI safety around guarding against two categories of harm that he labeled unintentional and intentional, and argued that open models help with the first while large actors retain the resources to address the second. [4]
With the 2025 superintelligence push, Zuckerberg's public framing shifted toward the goal of building highly capable AI systems and delivering what he called personal superintelligence to individuals. In a July 2025 letter he wrote that developing superintelligence was coming into sight and described a future in which everyone would have a personal AI assistant. He also signaled that Meta would be more selective about which future models it released openly, citing safety considerations as systems grow more powerful. [16][22]
Reaction to Zuckerberg's AI strategy has been mixed. Supporters of open models, including many academic researchers and some startups, have credited Llama with broadening access to capable language models and providing an alternative to closed systems from OpenAI, Google, and Anthropic. [4][14]
Critics have raised several concerns. Some AI-safety advocates argue that openly releasing powerful model weights makes misuse harder to control, because the weights cannot be recalled once distributed. Open-source advocates, including the Open Source Initiative, have objected to Meta's use of the open-source label for the restrictively licensed Llama models. Reporting in 2025 described internal turbulence at Meta, including questions about benchmark results for Llama 4, departures of long-tenured researchers, and friction between the company's research traditions and its product-driven approach. After his departure was announced, Yann LeCun publicly characterized Meta's new AI leadership as inexperienced, a view Meta did not endorse. [5][16][21]
Meta has separately faced litigation over the data used to train its models. In the copyright case Kadrey v. Meta, a group of authors alleged that the company used their books without permission to train Llama; in 2025 a federal judge ruled in Meta's favor on the specific claims before the court while noting the ruling was narrow. [23]
| Field | Detail |
|---|---|
| Full name | Mark Elliot Zuckerberg |
| Born | May 14, 1984, White Plains, New York |
| Education | Harvard College (enrolled 2002, left during second year) |
| Known for | Co-founding Facebook (2004); leading Meta Platforms |
| Current role | Chairman, CEO, and controlling shareholder of Meta Platforms |
| Company rebrand | Facebook renamed Meta Platforms, October 28, 2021 |
| AI research lab | Facebook AI Research / FAIR, founded December 2013 |
| Open model family | Llama (first release February 2023) |
| 2025 capex guidance | 64 to 72 billion US dollars |
| Superintelligence unit | Meta Superintelligence Labs, announced June 30, 2025 |
| Chief AI officer | Alexandr Wang (from June 2025) |
| Scale AI investment | About 14.3 billion US dollars for a 49 percent stake (June 2025) |
| Net worth (Forbes, late 2025) | About 220 billion US dollars |