Chan Zuckerberg Initiative
The Chan Zuckerberg Initiative (CZI) is the philanthropic organization founded in December 2015 by Mark Zuckerberg, the founder and chief executive of Meta, and his wife Priscilla Chan, a pediatrician. It is organized as a limited liability company (LLC) rather than a charitable foundation, a choice that drew criticism at launch and in later legal analysis.[1][2] CZI began with broad goals in education, social policy, and science. Over the following decade it narrowed toward biomedical research, and since 2023 it has concentrated on using artificial intelligence to model cells. Its science teams now operate under the name Biohub, with CZI providing funding and operational support.[11][12][13] CZI-backed projects include a 1,000-plus GPU computing cluster for biology, the CZ CELLxGENE single-cell data platform, cell models such as TranscriptFormer and rBio, and the Virtual Biology Initiative, a data-generation program that partners valued at $1.8 billion after an October 2026 expansion.[8][14][25]
History
Founding and the LLC structure (2015)
Zuckerberg and Chan announced the initiative in a Facebook post on December 1, 2015. They pledged to give 99 percent of their Facebook shares to the effort, a stake then valued at more than $45 billion, with the stated aims of advancing human potential and promoting equality for children.[1]
The couple chose a for-profit LLC instead of a private foundation. In a 2018 analysis in the Stanford Social Innovation Review, Brooklyn Law School professor Dana Brakman Reiser noted that an LLC is not bound by the private foundation rules on annual payouts, limits on business holdings, lobbying, and political activity, and that it does not have to file the public Form 990 disclosures that foundations file. Zuckerberg would also have had to give up control of his Facebook shares had they gone into a private foundation.[1] Critics argued at the time that the structure could serve to avoid tax. Zuckerberg replied that the couple received no tax deduction for moving shares into the LLC, as they would have with a traditional foundation, and that they would pay capital gains tax when the LLC sold shares.[2] Reiser noted that owners can still deduct gifts the LLC later makes to qualifying charities, so the deduction is deferred rather than forgone.[1]
Science program and the Biohubs (2016-2023)
On September 21, 2016, Chan and Zuckerberg committed $3 billion over ten years to science, with the goal of helping to cure, prevent, or manage all diseases by the end of the century. The plan included $600 million for the Chan Zuckerberg Biohub in San Francisco, a research center built with the University of California, San Francisco, the University of California, Berkeley, and Stanford University. Rockefeller University scientist Cori Bargmann was chosen to lead the science effort.[3] Bargmann stepped down as head of science in 2022 to return to her lab, and Stephen Quake, a Stanford bioengineer who had been co-president of the San Francisco Biohub since 2016, took over in July 2022.[5]
In December 2021 CZI said it would invest up to $3.4 billion over 10 to 15 years in biomedical technology. The Associated Press reported that the plan included $500 million over 15 years for an AI institute at Harvard University named for Zuckerberg's mother, Karen Kempner Zuckerberg; $600 million to $900 million for a biomedical imaging institute; $1 billion for a new Biohub Network; and $800 million to $1 billion for the San Francisco Biohub.[4] In 2023 CZI chose a Chicago team from 58 applicants to form CZ Biohub Chicago, co-led by Northwestern University, the University of Chicago, and the University of Illinois Urbana-Champaign, which develops sensors that measure inflammation inside human tissue.[6] In October 2023 CZI committed $250 million over ten years to CZ Biohub New York, a partnership with Columbia University, Rockefeller University, and Yale University to build tools for detecting disease early inside cells.[7]
Reorientation toward AI and biology (2023-2026)
In September 2023 CZI announced a computing cluster of more than 1,000 GPUs for training biological models (see Computing).[8] Over the same period the organization pulled back from its non-science work. The San Francisco Standard reported that CZI laid off dozens of education staff and stopped funding education policy in August 2023, and that it described itself as a "science-first philanthropy" in 2024.[28] In March 2025 CZI announced that the San Francisco Biohub and its imaging institute would merge into a single imaging-focused Biohub on a new Redwood City campus planned for 2027.[10]
On November 6, 2025, Chan and Zuckerberg announced that CZI's scientific teams would unite as one organization called Biohub, which would receive the large majority of the couple's philanthropic resources, while CZI would serve as infrastructure and support. Zuckerberg said Biohub would be "the main focus of our philanthropy going forward." The Associated Press reported that the couple had given $4 billion to basic science research since 2016, not counting the cost of operating the computing cluster, and that the organization expected to double that amount over the next decade with an operating budget of about $1 billion a year.[11][12] In the same announcement Biohub said it would expand its computing capacity tenfold to 10,000 GPUs by 2028 and that the team of EvolutionaryScale, the protein-model company behind ESM3, would join it. EvolutionaryScale co-founder Alex Rives became Biohub's head of science.[11]
In January 2026 CZI notified 71 employees, most of them in technical roles such as software engineering and data science, that their jobs would end on March 30, 2026. A CZI spokesperson said the organization's work "requires new kinds of expertise and capabilities."[30]
Structure and leadership
CZI, LLC sits within a group of related entities. Biohub describes its own structure as follows:[13]
| Entity | Role described by Biohub |
|---|---|
| Chan Zuckerberg Biohub, Inc. | 501(c)(3) medical research organization through which Biohub carries out its work |
| Chan Zuckerberg Initiative, LLC | Provides strategic, funding, and operational support |
| Chan Zuckerberg Initiative Foundation | 501(c)(3) private foundation |
| Biohub Donor Advised Fund | Donor-advised fund within the group |
Chan and Zuckerberg are co-founders and co-CEOs of CZI.[13] The San Francisco Standard reported in 2025 that CZI had hired its first president, Lori Goler, who previously spent 16 years leading human resources at Meta.[28] CZI's heads of science have been Cori Bargmann (2016-2022) and Stephen Quake (2022 onward); since November 2025 Alex Rives has led science as Biohub's head of science.[3][5][11]
Computing
On September 19, 2023, CZI announced that it would build "one of the largest" computing systems dedicated to nonprofit life-science research, with more than 1,000 GPUs. It said the cluster would train AI models, including large language models, to predict the behavior of healthy and diseased cells, and that it would be one of the first systems used to build openly available models of human cells.[8] The Register reported that the cluster would use NVIDIA H100 GPUs and that a CZI spokesperson expected it to be running in 2024; CZI declined to say what it cost.[9] The training data CZI named at the time included CELLxGENE, which then held more than 50 million cells, along with imaging data and the Tabula Sapiens and OpenCell atlases from the San Francisco Biohub.[8]
In October 2025 CZI and NVIDIA expanded their collaboration. The partners said they would use GPU-accelerated tools to process petabyte-scale biological datasets covering billions of cell-level observations, would publish NVIDIA models such as the CodonFM RNA model on CZI's Virtual Cells Platform, and had co-developed cz-benchmarks, an open Python package for evaluating biological models on common tasks.[22] Biohub's November 2025 target of 10,000 GPUs by 2028 represents a tenfold increase over the original cluster.[11]
AI models
CZI and Biohub distribute their models through the Virtual Cells Platform, a public hub for models, datasets, and benchmarks.[22] Most of the models are trained on single-cell RNA sequencing data or other cellular measurements.
| Model | Released | Description |
|---|---|---|
| TranscriptFormer | April 2025 | A family of generative transformer models for single-cell transcriptomics, trained on up to 112 million cells from 12 species spanning about 1.5 billion years of evolution. It can classify cell types in species absent from training and identify disease states such as SARS-CoV-2 infection without fine-tuning.[19][20] |
| GREmLN | 2025 | A graph-aware model of gene regulatory networks.[11] |
| rBio | August 2025 | A conversational reasoning model for cell biology. It was post-trained with reinforcement learning using rewards derived from TranscriptFormer and other sources of biological knowledge, with answers scored by how likely they are to be correct rather than right or wrong. Its expertise was limited to gene perturbation questions.[21] |
| VariantFormer | November 2025 | Predicts tissue-specific gene activity from a person's genetic variants.[11] |
| CryoLens | November 2025 | A pretrained model for cryo-electron tomography that measures structural similarity without labels.[11] |
| scLDM | November 2025 | Generates synthetic single-cell data.[11] |
| ESMC and ESMFold2 | May 2026 | Biohub's "world model of protein biology." ESMC is a protein language model trained on about 2.8 billion sequences, and ESMFold2 predicts structures of proteins and complexes and can help design binders. They were released with the ESM Atlas of 6.8 billion sequences and 1.1 billion predicted structures, under an MIT license.[23][31] |
The reported benchmark results for TranscriptFormer and rBio come from the developers' own preprints.[19][21] Independent benchmarks of other single-cell foundation models have found that they do not reliably beat deliberately simple baselines at predicting the effects of genetic perturbations.[32]
Data platforms and atlases
CZI's science strategy treats data as the main constraint on AI models of biology, and much of its funding has gone into shared datasets.
CZ CELLxGENE. CELLxGENE Discover is an open platform of curated single-cell datasets that follow a common metadata schema, so data from different studies can be searched and combined. As of October 1, 2024, it hosted more than 1,550 datasets with 169.3 million cells, of which 93.6 million were unique.[14] Its Census, a programmatic interface for querying the collection, listed 1,845 datasets in its November 2025 long-term release, including more than 162 million human cells (about 99.6 million unique) and 46 million mouse cells.[15] In 2023 CZI said about 75 percent of CELLxGENE data came from researchers outside CZI.[8]
Tabula Sapiens. Tabula Sapiens is a human reference atlas associated with the San Francisco Biohub.[8] Its first release, published in Science in 2022, covered nearly 500,000 cells from 24 tissues and organs and more than 400 cell types.[16]
Human Cell Atlas. CZI has funded the international Human Cell Atlas consortium. In June 2019 it awarded $68 million to 38 collaborative "seed network" teams mapping healthy human tissues.[17]
Billion Cells Project. Launched on February 6, 2025, with 10x Genomics and Ultima Genomics, the project aims to build a single-cell dataset of one billion cells for training AI models, using 10x's Chromium GEM-X platform and Ultima's UG 100 sequencer. CZI said the industry partners would lower costs and that results would be released openly.[18]
Virtual Biology Initiative
On April 29, 2026, Biohub launched the five-year Virtual Biology Initiative with a $500 million commitment: $400 million for internal data generation and measurement technology, and $100 million for research at other institutions. Founding partners included the Allen Institute, the Arc Institute, the Broad Institute, the Wellcome Sanger Institute, the Human Cell Atlas, the Human Protein Atlas, NVIDIA as technology partner, and Renaissance Philanthropy. Rives said that building AI that can represent biology requires "orders of magnitude more data than exists today."[24]
On October 7, 2026, Biohub, the U.S. Department of Energy, and the National Institutes of Health announced an expansion that the partners valued at $1.8 billion in funding, data, computing, and measurement technology. The figure combines Biohub's $500 million, more than $500 million in planned DOE spending over five years through the Genesis Mission, NIH datasets built with more than $500 million in earlier federal funding, and $300 million from Google DeepMind, Isomorphic Labs, and Meta.[25] According to reporting by The Next Web, the commercial funders get one year of exclusive access to data they help produce before it becomes public.[26]
Criticism
Legal structure. The LLC form has been criticized for reducing transparency and public accountability compared with a private foundation. Reiser wrote that philanthropy LLCs keep compensation and self-dealing decisions private, allow political spending and lobbying, and leave founders legally free to reclaim assets, although Zuckerberg's public statements did not suggest any intent to do so.[1]
End of DEI and social advocacy funding. In February 2025 CZI chief operating officer Marc Malandro told staff that CZI would no longer have a diversity, equity, inclusion, and accessibility team and would end its "social advocacy funding," which had supported racial equity and immigration reform grants. CZI also ended its Science Diversity Leadership Awards and a hiring practice that required diverse candidate slates. TechCrunch, following The Guardian's reporting, noted that the decision came weeks after CZI's head of human resources had told employees that Meta's own DEI cuts would not affect CZI.[27] The Associated Press reported that CZI faced criticism for scaling back this work and had removed DEI references from its website; Chan and Zuckerberg said their shift toward science began years before the 2024 election.[12]
Education and community programs. CZI had funded the Summit Learning personalized-learning platform, which drew parent protests in 2018 over concerns including student data privacy, and The Primary School, a free school in East Palo Alto that had operated since 2016 and that Chan had once led as chief executive.[28][29] In April 2025 the school announced that it would close at the end of the 2025-26 school year; parents told the San Francisco Standard that CZI was ending its funding. CZI said it would invest $50 million over the following years in the East Palo Alto, Belle Haven, and East Bay communities, including transition support for families.[29] A June 2025 San Francisco Standard investigation, based on interviews with a dozen current and former staff, described an organization whose priorities shifted with political and public relations pressure; a CZI spokesperson said the organization had "always been nonpartisan."[28]
Open data with an embargo. Biohub presents the Virtual Biology Initiative's datasets as an open resource, but data paid for by the commercial partners is held back for a year. Rives defended the delay as the incentive needed to bring companies into the project.[26]
Timeline
| Date | Event |
|---|---|
| December 1, 2015 | Zuckerberg and Chan announce CZI and pledge 99% of their Facebook shares[1] |
| September 21, 2016 | $3 billion science commitment; $600 million for the San Francisco Biohub[3] |
| June 21, 2019 | $68 million for Human Cell Atlas seed networks[17] |
| December 7, 2021 | Up to $3.4 billion for biomedical technology, including a $500 million AI institute at Harvard[4] |
| July 2022 | Stephen Quake succeeds Cori Bargmann as head of science[5] |
| 2023 | Chicago team selected for CZ Biohub Chicago; $250 million for CZ Biohub New York (October)[6][7] |
| September 19, 2023 | 1,000-plus GPU cluster announced[8] |
| February 6, 2025 | Billion Cells Project launched[18] |
| February 2025 | DEI team and social advocacy grantmaking ended[27] |
| April 2025 | TranscriptFormer released[20] |
| August 21, 2025 | rBio released[21] |
| October 28, 2025 | Expanded collaboration with NVIDIA[22] |
| November 6, 2025 | Science teams unite as Biohub; EvolutionaryScale joins; 10,000-GPU target set[11] |
| January 2026 | 71 layoffs notified, mostly technical staff[30] |
| April 29, 2026 | Virtual Biology Initiative launched with $500 million[24] |
| May 27, 2026 | ESMC, ESMFold2, and ESM Atlas released[31] |
| October 7, 2026 | Virtual Biology Initiative expansion valued at $1.8 billion[25] |
See also
References
- ^1 ^2 ^3 ^4 ^5 ^6Dana Brakman Reiser. The Rise of Philanthropy LLCs. Stanford Social Innovation Review, Summer 2018.
- ^1 ^2Kia Kokalitcheva. Mark Zuckerberg Takes on Critics of His $45 Billion Giveaway. Fortune, December 3, 2015.
- ^1 ^2 ^3Shayla Love. Mark Zuckerberg and Priscilla Chan announce $3 billion effort aimed at curing disease. STAT, September 21, 2016.
- ^1 ^2Haleluya Hadero. Zuckerberg, Chan to invest up to $3.4B for science advances. Associated Press via TechXplore, December 8, 2021.
- ^1 ^2 ^3Chan Zuckerberg Initiative. Stephen Quake Named New Head of Science at the Chan Zuckerberg Initiative. June 23, 2022.
- ^1 ^2Northwestern University McCormick School of Engineering. Transforming Our Understanding of Human Biology. McCormick Magazine, Fall 2023.
- ^1 ^2M.J. Prest. Chan Zuckerberg Initiative Awards $250 Million for Cell-Health Research Hub in New York. The Chronicle of Philanthropy, October 25, 2023.
- ^1 ^2 ^3 ^4 ^5 ^6 ^7Chan Zuckerberg Initiative. Chan Zuckerberg Science To Build AI GPU Cluster To Model Cell Systems. September 19, 2023.
- ^Chan Zuckerberg Initiative to build giant AI H100 cluster. The Register, September 20, 2023.
- ^Biohub. CZI announces Biohub to develop breakthrough imaging technologies to observe cells in action. March 26, 2025.
- ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10Biohub. Biohub launches frontier AI x frontier biology initiative to cure disease. November 6, 2025.
- ^1 ^2 ^3Barbara Ortutay. Zuckerberg, Chan shift bulk of philanthropy to science, focusing on AI and biology to curb disease. Associated Press via Phys.org, November 6, 2025.
- ^1 ^2 ^3Biohub. Team. Organizational structure and founders. Accessed October 16, 2026.
- ^1 ^2CZI Cell Science Program, et al. CZ CELLxGENE Discover: a single-cell data platform for scalable exploration, analysis and modeling of aggregated data. Nucleic Acids Research 53(D1), D886-D900, 2025. DOI: 10.1093/nar/gkae1142.
- ^Chan Zuckerberg Initiative. Census data releases. CZ CELLxGENE Census documentation. Accessed October 16, 2026.
- ^Tabula Sapiens Consortium. The Tabula Sapiens: A multiple-organ, single-cell transcriptomic atlas of humans. Science 376(6594), eabl4896, May 13, 2022. DOI: 10.1126/science.abl4896.
- ^1 ^2Chan Zuckerberg Initiative. Chan Zuckerberg Initiative Awards $68 Million to Support the Growth of the Human Cell Atlas. June 21, 2019.
- ^1 ^2Chan Zuckerberg Initiative. CZI launches Billion Cells Project with 10x Genomics, Ultima Genomics to advance AI in biology. February 6, 2025.
- ^1 ^2James D. Pearce, et al. A Cross-Species Generative Cell Atlas Across 1.5 Billion Years of Evolution: The TranscriptFormer Single-cell Model. bioRxiv preprint, version 2 posted October 11, 2025. DOI: 10.1101/2025.04.25.650731.
- ^1 ^2Biohub. TranscriptFormer: A generative cross-species cell atlas across 1.5 billion years of evolution. April 30, 2025.
- ^1 ^2 ^3Chan Zuckerberg Initiative. CZI Launches rBio: Reasoning Model Trained on Virtual Cell Simulations. August 21, 2025.
- ^1 ^2 ^3Brian Buntz. CZI and NVIDIA expand virtual cell push with open models and benchmarks. R&D World, October 28, 2025.
- ^Biohub. A World Model of Protein Biology. Model descriptions and availability. Accessed October 16, 2026.
- ^1 ^2Biohub. Biohub Launches the Virtual Biology Initiative to Galvanize a Global Effort to Create the Open Data Foundation for AI-Accelerated Biology. April 29, 2026.
- ^1 ^2 ^3Biohub. International, cross-sector collaboration commits nearly $2 billion to build foundational data for AI models to predict and treat disease. October 7, 2026.
- ^1 ^2Ana Maria Constantin. Biohub, Meta, Google DeepMind and the US pool $1.8bn for AI biology data. The Next Web, October 7, 2026.
- ^1 ^2Dominic-Madori Davis. Mark Zuckerberg's charity U-turns, ends DEI efforts. TechCrunch, February 19, 2025.
- ^1 ^2 ^3 ^4Emily Shugerman. How politics and image control destroyed the Chan Zuckerberg Initiative. The San Francisco Standard, June 9, 2025.
- ^1 ^2Garrett Leahy and Emily Shugerman. Distraught families say Zuckerberg pulled funds from low-income school. The San Francisco Standard, April 23, 2025.
- ^1 ^2Miranda de Moraes. Chan Zuckerberg Initiative to cut dozens of its tech employees. Redwood City Pulse, February 5, 2026.
- ^1 ^2Biohub. Biohub releases a world model of protein biology. May 27, 2026.
- ^Constantin Ahlmann-Eltze, Wolfgang Huber, and Simon Anders. Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines. Nature Methods 22, 1657-1661, 2025. DOI: 10.1038/s41592-025-02772-6.
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation. Every suggestion is reviewed for sourcing before it goes live.
1 revision · v1 · 3,277 words · full history
Fact-checks are independent of edits: a reviewer re-verifies the article against its sources and stamps the date. How we verify
Research and drafting on this wiki are AI-assisted, under named human editorial standards. How AI is used here
Cite this page: AI Wiki. "Chan Zuckerberg Initiative." aiwiki.ai, updated 15 Oct 2026. CC BY 4.0. https://aiwiki.ai/wiki/chan_zuckerberg_initiative