Chan Zuckerberg Initiative

In a Redwood City office, the most well-funded philanthropy in modern history is doing something strange for a charity. It is training neural networks. Not to recommend a video or sell an ad, but to simulate a living human cell - how it works, how it breaks, and how you might fix it. This is the Chan Zuckerberg Initiative now. The grant checks still go out, but the center of gravity has moved to a screen full of biology that does not exist yet.

The problem they saw

Biology has no working simulator

Here is the uncomfortable thing about modern medicine: we still cannot reliably predict what a cell will do. Engineers model bridges before they pour concrete. Pilots train in simulators before they fly. Biologists, for the most part, run the experiment and wait. The cell - the basic unit of every disease - remains stubbornly opaque, studied one slow assay at a time.

Priscilla Chan saw this from the inside. A pediatrician and former teacher, she watched families wait years for a diagnosis that a faster science might have delivered in weeks. Mark Zuckerberg saw it from the other side - as a builder who knew what happens when you give a hard problem enough compute and the right data. The two views met at an obvious, awkward question: why does biology not have software?

The founders' bet

A charity that behaves like a company

CZI was announced in December 2015, wrapped in a letter to the founders' newborn daughter and a pledge to give away 99% of their Meta shares. The headline was the money. The quieter, sharper decision was the structure. CZI is not a foundation. It is a limited liability company - a “for-profit charity” that can make grants, fund startups, build software, and advocate for policy, all from the same chassis.

Critics called the LLC a tax-flavored loophole. Supporters called it the only structure flexible enough to do science at the speed software moves. Both were a little bit right, which is usually the sign that something interesting is happening.

For its first years, CZI sprawled - education platforms, criminal justice reform, housing affordability, a COVID-19 response, a school in East Palo Alto. Then, in 2025 and 2026, came the recalibration. CZI wound down or spun off much of that work and trimmed roughly 8% of its staff, about 70 jobs. The message was blunt: everything points at AI-powered biology now.

The org chart got smaller and the dream got bigger in the same fiscal year. Make of that what you will.

The product

Virtual cells, and the data to feed them

The flagship is the CZ Biohub network - sites in San Francisco, Chicago and New York where scientists, engineers and AI researchers sit in the same building. Their headline product is not a drug. It is a virtual cell: an AI system trained to simulate how a cell behaves, how it responds to its environment, how it malfunctions under disease, and how it might be reprogrammed.

To train those models you need biological data at a scale that does not yet exist. So in April 2026, Biohub committed $500 million to the Virtual Biology Initiative - $100 million to coordinate a worldwide data-generation effort, and $400 million to build the instruments that measure, image and engineer biology. A month later it shipped a “world model of protein biology,” including a structure-prediction model, a protein language model, and ESM Atlas: a map of 6.8 billion proteins and over a billion predicted structures.

The proof

Money out, tools out, data out

Skepticism is fair - moonshots are cheap to announce. So here are numbers that are hard to wave away. Since 2015, CZI has awarded roughly $7.22 billion in grants and invested nearly $300 million in mission-aligned ventures. Its open tools are used by tens of thousands of researchers who never signed a contract or saw an invoice.

A bar chart is a blunt instrument for a billion proteins, but it does make one thing obvious: this is a lot of resources aimed at one cell.

Then there is the soft proof, which may matter more. CZ Biohub got Stanford, UC Berkeley and UCSF - three institutions not famous for sharing - to pursue a single mission together. In academic terms, that is closer to a miracle than a partnership. Academics guard their labs, their grants and their authorship lines the way nations guard borders. Getting them to pool data and compute toward one goal is the kind of thing that does not show up on a balance sheet but quietly changes what is possible.

It helps that CZI did not arrive empty-handed. Co-CEO Priscilla Chan still talks like the pediatrician she trained to be - careful about overpromising, insistent that solutions answer to real practitioners rather than press cycles. Mark Zuckerberg brings the opposite reflex: the engineer's faith that a hard problem yields to enough compute and the right data. The friction between those two instincts - one cautious, one impatient - is arguably the most productive thing the organization owns.

The mission

Why “all disease” is the point

“Cure all disease” sounds like a slogan, and it is. But it is also a deliberate framing. CZI is not trying to beat one illness. It is trying to build the shared infrastructure - the data, the models, the tools - that lets every disease researcher move faster. Cure a single disease and you save a population. Build a working model of the cell and you change the rate at which every cure arrives.

That is the difference between a foundation and what CZI is trying to be. One funds outcomes. The other is trying to fund the floor everyone else stands on.

Why it matters tomorrow

The skeptic's case, and the answer to it

The honest critique writes itself. Tech founders have promised to fix big problems before and produced press releases. The LLC structure invites suspicion. “Predictive models of life” is the kind of phrase that ages badly if the science stalls. CZI knows this; the 2026 layoffs were partly an admission that doing everything meant doing nothing especially well.

But the counter-argument is sitting in the open. The grants are real and audited. The tools are downloadable today. The protein atlas exists whether or not you find the mission grandiose. CZI made a bet that the bottleneck in medicine is not ambition or money but a missing software layer for biology - and it is spending billions and shipping code on exactly that thesis.

Back in that Redwood City office, the neural network is still training on a cell that does not exist yet. If it works, the strangest thing about the Chan Zuckerberg Initiative will not be that a charity learned to write biology software. It will be that, for once, somebody with this much money pointed it at the unglamorous floor of a problem instead of its ceiling - and then gave the floor away for free.