Gates Foundation commits $1 billion to put AI tools in classrooms, clinics and fields

The Gates Foundation said on 15 September that it will spend at least $1 billion over two years to expand access to artificial intelligence in education, health and farming, arguing that the market will otherwise build the best tools first for people who can already pay.

The pledge sits inside the foundation’s annual Goalkeepers report, which tracks the United Nations sustainable development goals due in 2030. Chair Bill Gates and chief executive Mark Suzman both signed the warning. “Left to the market alone, the most capable tools will be built first for the people and institutions most able to pay for them, not necessarily for those who could benefit most,” Gates wrote.

The money is split on a published grid. About 40 percent, or $400 million, goes to education: AI tutors that adapt to a student, and classroom tools that help a teacher rewrite a lesson around the points a class missed. Another 40 percent goes to health: diagnostic support for frontline workers, checks on clinical notes for missed symptoms, maternal and newborn tools, and work on drug and vaccine discovery. About 10 percent, or $100 million, goes to smallholder farmers, including advice matched to soil, weather, pests and fertilizer. The last 10 percent funds datasets in languages that current models barely speak.

Suzman said models that work well in English, and to some extent in Mandarin, still fail in languages spoken by the people the foundation is supposed to serve. The language pot is meant to close that gap. The report cites Maharashtra’s MahaVISTAA service, a free advisory line for farmers, as a working example of advice delivered in a local setting rather than as a generic chatbot.

The timing is political as well as philanthropic. This week the UN Security Council, under France’s September presidency, is expected to hold a session on AI and security during the General Assembly. UN human rights chief Volker Türk has asked states and companies for “urgent action” on frontier models. President Donald Trump on Monday called talk of an existential AI threat a “hoax” and described regulation campaigns as a “SICK conspiracy.” The Gates package does not answer that fight. It spends on clinics and fields instead of on a treaty.

A billion dollars in two years is large for a foundation programme and small beside what Microsoft, Google, Amazon and the Gulf funds are putting into data centres. The point of the split is to buy specific tasks: a tutor that can mark a homework script in a language a commercial model ignores, a checklist that catches a missed fever in a clinic without a specialist, a pest alert that knows the crop on a one-hectare plot. Those products do not attract the same capital as a general assistant sold to offices in California.

Education money will land with organisations the foundation already funds. The report does not name a new global platform. It talks about expanding work that exists. That is slower than a splashy app and harder to measure in a press cycle. It is also how the foundation has moved on vaccines: pay for delivery in places companies skip.

Health money is the riskiest line. Drug and vaccine discovery with AI is fashionable. Diagnostic support for a nurse in a district hospital is less so, and more likely to hit a real patient this year. The report lists both. Readers should watch which invoices are paid first.

Agriculture is the smallest slice and the one with the clearest existing product. MahaVISTAA and similar services already send weather and pest notes to phones. The question is whether a $100 million top-up buys better local data or only a new wrapper. Suzman has also praised agricultural work in the United Arab Emirates as a partner model. That praise is diplomatic. The farming line will be judged in Maharashtra and in similar states, not in Dubai.

The language datasets are the part that other funders often skip because they do not produce a demo. If the $100 million actually builds speech and text corpora that labs can use, commercial models will absorb them. That is the intended leak. The foundation is paying to make the public layer that companies will then train on.

Goalkeepers is a scorecard, not a budget law. The $1 billion is a floor. How much of it is new money, and how much is a relabel of programmes already planned, will show up in the next two annual reports. For now the public claim is simple. The foundation will spend at least a billion dollars by 2028 so that AI tools reach students, clinicians and farmers who are not the first customers the market would choose.