Fifty thousand poles working in concert, without drawing a single watt from the grid
In the sun-drenched streets of Nigeria's Katsina State, fifty thousand lamp posts have quietly become something the world has not seen before: a distributed artificial intelligence data center woven into the fabric of daily life. A British company, Conflow Power Group, has turned modest solar-powered street lamps into a collective computing force rivaling traditional data centers — without touching the electrical grid. At a moment when AI's hunger for energy and resources strains the planet, this experiment asks whether the infrastructure of tomorrow might already be standing on the corners of today.
- Africa's first distributed AI data center has emerged not from a purpose-built facility, but from 50,000 solar street lamps quietly humming across Katsina State.
- Each iLamp carries a 15-watt NVIDIA chip, a solar panel, and its own battery — modest alone, but together generating 13.75 petaOPS of computing power with zero grid dependency.
- The lamps double as surveillance tools, catching speeding drivers and parking violations on camera, with revenue split between the state government and Conflow Power Group.
- Experts caution that inter-pole latency makes the network unsuitable for heavy AI workloads, positioning it instead as a lightweight, distributed alternative — closer in spirit to mobile antennas than supercomputers.
- If negotiations with seven additional Nigerian states succeed, the network could grow to 300,000 units, potentially reshaping how an entire continent builds its computational future.
While tech giants imagine data centers in orbit or beneath the ocean, a Warwickshire-based company chose a more grounded approach. Conflow Power Group struck a deal with Nigeria's Katsina State to install fifty thousand solar-powered intelligent street lamps — and in doing so, created something Africa had never seen: a distributed AI data center embedded in everyday urban infrastructure.
Each iLamp is self-sufficient. A low-power NVIDIA chip, a cylindrical solar panel, and an onboard battery allow it to operate entirely off-grid. Individually, the specs are unassuming. Collectively, fifty thousand units produce 13.75 petaOPS of computing power — comparable to a conventional data center that would otherwise demand 300 megawatts of grid electricity, vast cooling water, and years of construction.
The lamps earn their place in multiple ways. Cameras mounted on each pole automatically detect traffic violations, generating fine revenue that flows to the state government, with Conflow taking a cut after three years. Computing capacity is rented to AI companies, with proceeds channeled into a green bond funding the network's expansion and upkeep. Free public Wi-Fi and Bluetooth connectivity come as an added benefit to residents below.
There are honest limitations. The physical distance between poles introduces latency that rules out heavy AI workloads. For lighter tasks, however, the iLamps function much like distributed mobile antennas — bringing computation closer to the people who need it. Should negotiations with seven other Nigerian states and various institutions succeed, the network could scale to 300,000 units.
The deeper significance lies in the timing. As AI's global expansion strains energy systems and accelerates the electronic waste crisis, Katsina's experiment points toward a different model — decentralized, solar-powered, and woven into existing urban spaces. That Nigeria, so often cast as a technology consumer rather than innovator, is leading this experiment makes it all the more worth watching.
While SpaceX dreams of data centers orbiting in space and Microsoft has tested submerged server farms on the ocean floor, a British company took a different path—one that stays firmly rooted in the ground. Conflow Power Group, based in Warwickshire, signed an agreement with the government of Katsina State in Nigeria to install fifty thousand intelligent street lamps powered entirely by solar energy. The result is something Africa has never seen before: a distributed artificial intelligence data center spread across an entire region, built into the infrastructure of daily urban life.
Each lamp, called an iLamp, is a self-contained unit. Inside its housing sits a low-power NVIDIA chip that consumes just fifteen watts of electricity, a cylindrical solar panel mounted on top, and its own battery for storage. Individually, these specifications sound modest. But when fifty thousand of them work in concert, they generate 13.75 petaOPS of computing power—roughly the processing capacity of a traditional data center—without drawing a single watt from the public electrical grid. A conventional data center, by contrast, typically demands three hundred megawatts of grid electricity, millions of liters of water for cooling systems, and years of construction before it becomes operational. The iLamps require none of that.
The lamps do far more than compute. Each one can be fitted with cameras for traffic enforcement, automatically detecting speeding violations, parking infractions, and seatbelt violations. The technology exists to add facial recognition for identifying missing persons or wanted individuals, though that capability has not yet been activated. The poles also broadcast free public Wi-Fi and Bluetooth connectivity to residents in the areas where they stand. For the state government, the financial model is straightforward: Katsina keeps the revenue from traffic fines recorded by the cameras, with Conflow Power Group taking a twenty percent cut after three years. Revenue from renting out computing capacity to AI companies flows into a green bond designed to fund the network's installation and ongoing maintenance.
Yet there are real limits to what these distributed poles can do. Experts acknowledge that the iLamps are not suited for heavy artificial intelligence workloads. The physical distance between poles introduces enough latency to make certain computational tasks impractical. For lighter applications, however, they function much like mobile network antennas—distributed access points that bring computing closer to the people who need it. This distinction matters because it shapes what comes next. The Katsina agreement may be only the beginning. If ongoing negotiations with seven other Nigerian states, universities, and institutions succeed, the total network could expand to three hundred thousand iLamps, making it the largest distributed AI computing network on the African continent.
The timing of this project is significant. The global expansion of AI infrastructure is placing unprecedented strain on energy systems and natural resources. Experts warn that the current pace of growth could severely worsen the electronic waste crisis, already a grave environmental problem worldwide. Decentralized solutions powered by renewable energy and woven into existing urban spaces offer one path toward more sustainable expansion. Remarkably, Nigeria—a nation often positioned as a consumer of technology rather than an innovator—is leading this experiment. What happens in Katsina over the next few years will likely shape how other regions think about building the computational infrastructure of the future.
Citações Notáveis
A traditional data center typically demands three hundred megawatts of grid electricity, millions of liters of water for cooling, and years of construction. The iLamps require none of that.— Conflow Power Group
A Conversa do Hearth Outra perspectiva sobre a história
Why did a British company choose Nigeria specifically for this experiment?
Katsina State has both the infrastructure challenge and the political will. They needed better electricity access and traffic management. Conflow saw an opportunity to solve multiple problems at once—and to do it in a place where the government was willing to try something genuinely new.
The fifty thousand poles generate computing power without grid electricity. But doesn't that seem almost too good to be true?
It's real, but it's also not magic. Each pole is weak individually. The power comes from scale and coordination. And yes, there are real constraints—you can't run heavy AI models across them because the latency between poles makes it impractical. But for edge computing, for tasks that live closer to the user, it works.
What about the surveillance aspect? Traffic cameras are one thing, but facial recognition for missing persons—that's a different conversation.
That's why it hasn't been activated yet. The technology exists, but the ethical and legal questions don't have easy answers. Katsina is being cautious, which is probably wise. That's the conversation that needs to happen before those cameras turn on.
If this expands to three hundred thousand poles across Nigeria, what changes?
You're looking at genuine continental infrastructure. Not just computing power, but a network that could serve as the backbone for smart cities across West Africa. The question then becomes whether other countries want to replicate it, or whether they see it as a Nigerian experiment.
Does this actually solve the energy crisis in AI, or is it just a clever workaround?
It's a workaround that points toward something real. Traditional data centers are unsustainable at the scale AI is demanding. This shows that you don't need massive centralized facilities. You can distribute the load, use renewable energy, and integrate it into the places where people actually live. It's not a complete solution, but it's a direction worth following.