Building sustainable AI infrastructure requires a coordinated action plan involving many stakeholders. A preliminary set of five capabilities frames how this strategic objective can be delivered: Sustainable and responsible green energy To minimize the environmental impact of AI, it is crucial that the rapid expansion in data centres is powered by sustainable and responsible energy sources. The substantial financial investment in this infrastructure in some regions will not be affordable for most countries. Therefore, viable alternatives will be necessary, alongside strong collaboration among energy providers, environmental organizations, technology firms and governments. Illustrating the substantial scale of the commitments that global tech firms have begun to make in the most mature AI markets, Microsoft recently signed a power purchasing agreement in the US to buy carbon-free energy using only nuclear power.1 This agreement will reopen the Three Mile Island plant, shut down since 2019, solely to supply the Microsoft data centre with green energy. For the benefit of emerging and developing economies, the World Bank has created a $2 billion, 10-year, multi-phased initiative to add 15 gigawatts (GW) of renewable energy capacity (enough to charge 650 million electric cars every year). The plan aims to eliminate around 240 million tonnes of carbon emissions (equivalent to avoiding the combustion of 100 billion litres of gasoline). The first initiative of the programme involves a $657 million financing facility for Turkey and an enabling framework to attract private capital to scale up renewable energy.2 As AI use cases expand, extensive decarbonization opportunities are emerging to support global climate and energy conservation goals. AI is already being embedded into building and network management, predictive maintenance, grid optimization and fleet management. Some use cases demonstrate conservation rates of up to 60%, with potential for further optimization.3 Secure networks and resilient AI supply chains Resilience means that national critical infrastructure and enterprise AI systems are protected against disruption and can withstand cyberattacks and other risks. Mitigating these risks necessitates cooperation among cybersecurity firms, infrastructure providers and governments to develop robust AI infrastructure and associated regulatory guardrails.Foreign investment plays a significant role here. It can accelerate development and provide essential resources but may also create geopolitical complexities. Without a strategy to ensure the diversity of AI hardware, countries may find themselves grappling with strategic and political considerations, particularly related to national security and economic sovereignty. Governments are thus compelled to assess the implications of being reliant on other countries or tech firms for critical infrastructure. Governments can guide industry to deploy resilient, scalable and secure foundations for AI by establishing a national AI security framework, facilitating public-public or PPPs, or directly leading AI infrastructure development. These strategies offer various balances between government control, private sector involvement and international cooperation. The development of new international trade corridors presents another viable route to building resilient AI supply chains, offering flexibility, reduced risk, and improved resilience through diversified sourcing and distribution networks. Trade corridors can alleviate supply chain issues related to AI hardware (such as microchips and critical resources like cobalt used in semiconductors) by improving material access and streamlining transport. The NY SMART I-Corridor illustrates the potential of trade corridors; it aims to establish a world-leading semiconductor cluster by joining more than 100 regional semiconductor suppliers together to provide local industries with expansive growth opportunities.4 Access to high-speed connectivity Access to high-speed networks is crucial for the effective functioning of AI systems and to encourage inclusive digital participation. Expanding this infrastructure, especially in underserved areas, can represent a significant opportunity for telecommunication providers and technology firms. Currently, approximately 2.6 billion people (one-third of the global population) remain offline.5 Digital public infrastructure (DPI) is a set of secure and interoperable digital systems built on open standards. They enable a community of competitive market players to provide innovative digital services to deliver public service objectives. DPI is another way to bridge the digital divide and aid inclusion by enabling businesses and citizens to gain access to online resources like digital payments, healthcare services and technology.6 AI is being added to DPI to personalize user interactions and translate languages in real time. Additionally, DPI can be used as a catalyst for AI To minimize the environmental impact of AI, it is crucial that the rapid expansion in data centres is powered by sustainable and responsible energy sources. Blueprint for Intelligent Economies 9

A Blueprint for Intelligent Economies 2025