AUK-based company has launched a solar-powered streetlight fitted with embedded AI processors, offering a distributed and off-grid solution to the rising power demands of global AI data centres.
AI energy consumption
Artificial intelligence workloads are driving up electricity use worldwide. According to the International Energy Agency, data centre consumption attributable to AI is set to more than double from 415 terawatt-hours (TWh) to about 945 TWh by 2030. This growth has spurred several countries to announce plans for expanding nuclear power generation.
Research also indicates the scale of environmental impact created by the current approach. A recent Cornell University study found that, by 2030, the increase in US demand for AI could emit as much as 44 million tonnes of carbon dioxide annually. Separately, University of California research suggests AI-related water consumption for cooling could reach 1.7 trillion gallons by 2027.
Streetlight alternative
The iLamp, developed by Conflow Power Group, provides its own energy using cylindrical solar panels housed in borosilicate glass. Each lamp generates up to 200-600W of solar power while consuming only 80W, according to the manufacturer. The surplus supplies embedded Nvidia Jetson Edge AI modules. These processors reportedly achieve up to 67 trillion operations per second with a 15W power demand.
High-performance versions are built to handle even more demanding AI models, accommodating hundreds of billions of parameters for on-device inference tasks. This distributed approach enables computation to take place locally, bypassing the need for data centre facilities and water cooling infrastructure.
Business model
The initiative allows owners of the streetlights to profit by selling compute time to external AI firms. Companies such as OpenAI and Google Gemini would pay to use remote edge processing resources, rather than relying solely on centralised data centres. iLamp units come with a 20-year warranty and a starting price of GBP £7,500.
The company operates via territorial licence agreements, enabling licensees to sell and deploy the streetlights locally. Customers-in some cases governments or infrastructure operators-may access local and federal tax incentives, as well as rebates. The distributed compute model may also lower ongoing energy costs for operators.
Safety technology
A licensing deal last year saw US$45 million paid for exclusive rights to the iLamp in Florida. A subsequent agreement, valued at US$80 million, enabled deployment of safety infrastructure in 4,400 schools. This market has been valued at US$777 million by the participating parties.
The school safety version of the lamp integrates AI for threat identification and response. Capabilities include weapon detection, gunshot recognition, advanced number plate identification, facial recognition, and fire and smoke warnings. Connectivity features, such as private wireless networks and vehicle speed detection, are also available.
Sports applications
A further collaboration with a technology company is developing an iLamp for rugby clubs in the UK and France. The unit will support night training, AI-driven match analysis, and tactical video review using on-board analytics.
“While tech giants scramble to build nuclear power stations to feed their AI addiction, we’ve built something smarter. Right now, in order to power AI, the likes of OpenAI or Google Gemini need a huge building full of GPUs and they have to pump massive amounts of electricity into it, along with a huge water supply for the cooling system. This is inefficient and we need a smart solution,” said Edward Fitzpatrick, Director, Conflow Power Group.
“There are streetlights everywhere in our towns and cities. By replacing them with iLamps fitted with Nvidia Jetson processors, you create a huge distributed data centre which is clean, non-water-hungry and low latency because the servers are near to the users. We are already in advanced negotiations with several large companies and world governments to make this a reality,” said Fitzpatrick.
Source: itbrief
