Across the United States, towns are pushing back against data centers, the giant buildings full of computers that power the internet and artificial intelligence. Neighbors worry they strain the grid and drive up electricity bills.

In rural Africa, the same kind of machines may do the opposite. They could help bring electricity to villages that have never had it.

That’s the idea behind the Emergent Grid , a project launched this week by an arm of The Rockefeller Foundation. It will test whether selling computing power can help pay for small, local power systems in Africa, and make electricity cheaper and more reliable.

Here’s how it works. Every time you ask an AI chatbot a question, a computer somewhere does the work to answer it. The tech industry calls that work “compute.” Much of it can happen anywhere with electricity and an internet connection, including a shipping container of computers, about the size of a one-car garage, parked next to a solar farm in rural Kenya.

The electricity never leaves the village. The answer goes out over the internet, and money comes back. The customer pays the company that owns the computers, and that company pays the local power supplier for the electricity it uses. That gives the supplier steady income from power that would otherwise go to waste. With it, the supplier can keep its equipment running, string wires to more homes, and charge families less.

When villagers need that power, the computers pause.

The catch: in America, data centers are resented because they can outbid everyone else for power. The project has to show that won’t happen in African villages. And so far, the idea has been proven not with AI, but with bitcoin.

“Compute can be a geographically agnostic buyer of last resort,” Erik Hersman, CEO of Gridless , told me. In plain terms: the computers can go anywhere, and they’ll buy whatever power no one else wants. Gridless has spent four years doing just that at remote African power sites.

More than 560 million people in sub-Saharan Africa have no electricity. A World Bank-led effort called Mission 300 aims to connect 300 million of them by 2030. About 64 million have been connected so far.

For those beyond the reach of power lines, the answer is often a mini-grid: a small local power system, usually solar panels and batteries or a small hydro plant.

Mini-grids are built bigger and supply more energy than villages need at first. A newly connected family might start with a few light bulbs and a phone charger. The average mini-grid customer uses about as much electricity in a month as American homes do in 12 hours. The rest often goes unsold. But the panels, batteries, and wires cost the same either way. So most African minigrids depend on grants to cover the gap, and that scares off private investors.

“They’ll say, wait a second, more than half of what you earn comes from grant money,” Andrew Herscowitz, who leads the Rockefeller-backed Mission 300 Accelerator, told me. “That’s not sustainable. I’m not going to invest in that.”

How Bitcoin Paid For A Village

Hersman’s solution started with bitcoin mining. Bitcoin is a digital currency, and “mining” is how new coins are made. Nobody hires the miners. Their computers race to solve math puzzles, and the winners are rewarded with new bitcoin, which they sell for dollars, much as gold prospectors sell what they dig up. More electricity means more puzzles solved.

That makes a bitcoin machine an ideal minigrid customer. It will use every bit of spare power and shut off instantly when villagers need the electricity.

Hersman’s best example is Zengamina, a small hydro plant in rural Zambia. The village used barely a third of the plant’s power, and the operator kept asking donors for money. Gridless brought in a shipping container of bitcoin-mining machines to use the leftover power. Unlike AI computers, these machines answer no one’s questions. They only solve bitcoin puzzles.

Within a year, the new income let the operator connect about 2,000 more people and about 40 businesses, including corn mills, barbershops, and clinics.

At a Gridless site in Kenya, the price villagers paid for electricity dropped by more than a quarter once the operator could sell all the power it made.

“When the energy company is making enough money, they don’t want to overcharge consumers who don’t have a lot of disposable income,” Hersman says.

Bitcoin “mining” earns only a few cents for each unit of electricity it uses, according to the project team, so “miners” can pay the power supplier very little. AI work earns far more from the same electricity. The question at the heart of the project: can AI computers do what the bitcoin machines did, paying the power supplier more while still switching off whenever the locals need the power?

There are hopeful signs that customers exist. Liquidstar, one of the project’s partners, says it has lined up American companies to rent more than 1,500 AI chips at a data center in South Africa. Those customers send their work over the internet and pay for the results. That African site runs on the regular grid, though, not a minigrid. For the Emergent Grid, Liquidstar will test AI computers on solar power while supplying extra electricity to a nearby community.

But AI is harder. It needs far pricier machines, and the project is betting on work that can be scheduled around a village’s needs.

The test covers up to 10 sites in Kenya, the Democratic Republic of Congo, Sierra Leone, and a fourth country still being finalized. Rockefeller’s affiliate will help pay for equipment and expects to be repaid from computing income, so the money can fund more sites. It isn’t paying for bitcoin “mining,” though some sites will run both. Arizona State University will independently judge the results.

There are no AI results yet, and the bigger worry is the one Americans are already protesting. Bitcoin can’t crowd out villagers, because it pays less for power than they do. AI can pay much more. That’s the fear behind the backlash against data centers in the United States.

The project’s answer is a firm rule. “You always have to make sure that the community is prioritized,” Herscowitz says. The computers switch off when villagers need the power, and the university will track whether prices and service hold up.

And who will rent computing from a container in rural Africa instead of from Amazon or Google?

“We don’t know 100% if this is going to work,” Herscowitz says. “We don’t want to oversell it.”

He argues that the tech giants have reserved much of the world’s computing capacity for years, pushing smaller companies to look elsewhere. And a container of computers can be set up in weeks or months, while a giant data center takes years.

“What we want to see is that electrification isn’t the biggest problem anymore,” Hersman says. “It’s more that we have affordable electrification.”

In America, the fight over data centers is about who gets the power. In rural Africa , the Emergent Grid is testing whether computers can help pay for it, without ever taking power away from the people it’s meant to serve.