Green AI takes centre stage at MWC Barcelona - UNESCO

A multistakeholder moment

The session was opened by Lacina Koné, CEO and Director-General of Smart Africa, who framed green AI as a foundational governance priority requiring coordinated action across governments, international organizations and industry. Moderated by Thelma Quaye, Director of Infrastructure, Skills and Empowerment at Smart Africa, the session's diverse panel was itself a statement: building resilient and efficient AI cannot be achieved by any single actor.

Two sides of the Green AI Challenge

The policy toolkit

Digital policies and sector policies are developed in silos," he observed. "We need to work across those silos to manage AI sustainably.

Spain leads by example

David de Francisco Marcos, Deputy Director-General for the Promotion and Regulation of Artificial Intelligence at Spain's Ministry of Digital Transformation, presented Spain's national Green Algorithm Programme as proof that governments can and must lead by example on AI efficiency. One of the most concrete government-led initiatives globally, the programme uses public procurement standards and regulatory incentives to systematically reduce the computational and environmental footprint of public sector AI. For policymakers across Africa and beyond, the message was clear: embedding green requirements into government AI systems is both feasible and replicable, at any stage of a country's digital transition.

The private sector as partner and driver

Jeanine Vos, Head of the SDG Accelerator at GSMA, illustrated how industry is already driving both sides of the green AI agenda at scale. On greening with AI, FarmerChat reduced the cost of reaching farmers from 35 dollars to just 35 cents through a smaller, locally-adapted model, while Husk Power Systems achieved a 40% reduction in diesel use through predictive AI for solar mini-grids. On greening AI itself, telecoms are reducing the energy consumption of their own networks and shifting toward leaner, more resilient models designed for low-resource environments. 

These solutions work precisely because of those constraints. When AI is designed for the hardest environments first, it scales faster, costs less, and reaches the people who need it most.

From conversation to commitment

Quaye closed with a clear call: the frameworks, tools and examples for resilient and efficient AI now exist. What is needed is political commitment to embed both dimensions of the green AI agenda into national AI strategies, data centre licensing, energy planning and public procurement, before infrastructure choices lock in unsustainable patterns. The alignment visible at MWC between governments, international organizations and the private sector was an encouraging signal that this shift is firmly underway.