OpenAI Foundation Bets 25 Billion on Data Not Models
The Ledger That Rewrites What Charity Can Buy
The OpenAI Foundation began writing cheques in 2026, and the first ones landed on laboratories rather than on language models. The foundation is on track to become one of the wealthiest non-profit organizations on the planet, with at least $25 billion in total giving. It holds a 26% stake in OpenAI, a company seeking a valuation of around $1.4 trillion. On paper, that makes this charity a candidate for the wealthiest non-profit on Earth — a strange sentence to write about an organization that is barely a year old. The pressure comes from a slice of one of the most valuable private companies in the artificial-intelligence sector.
The first grants read like a shopping list for scientific infrastructure, not for products. The foundation’s stated aim is to support the underlying scientific infrastructure so that all researchers can build on it and all patients can benefit. The bottleneck it names is the observations, not the algorithms.
Jacob Trefethen, who leads life sciences at the OpenAI Foundation in San Francisco, California, frames the logic plainly. He describes this as only the start. The stated mission is to ensure AI benefits all of humanity, and the choice of life sciences and health as the opening arena rests on a judgment about where AI can do the most good. That is a claim about comparative advantage, not about charity as such.
The Question Every Other Field Is Now Asking
Trefethen names three focus areas: AI for Alzheimer’s, public data for health, and high-burden diseases. The last category is defined with unusual precision — any disease causing more than half a million deaths worldwide each year, or a comparable burden of ill health. Tuberculosis, described as the largest infectious killer, is the explicit example. The unifying philosophy is compressed into a single sentence about intelligence plus more observations of the world. Break that sentence apart and the wager becomes clear: breakthroughs against disease come from seeing more, not from thinking harder.
The Alzheimer’s case carries the argument furthest. On this problem, Trefethen says, researchers collectively do not know enough to deliver treatments that are excellent for patients. That admission is the hinge. If the deficit were computational, a better model would close it. If the deficit is observational, no model can. The foundation’s answer is to fund the underlying scientific infrastructure so that all researchers can build on it and all patients can benefit.
> Note: Only birds’ vision was tested; other predators may perceive patterns differently.