AI infrastructure investment is expanding rapidly, but the benefits of that boom will remain concentrated unless countries build the wider systems needed to use artificial intelligence. The UN Trade and Development body, or UNCTAD, says investment must extend beyond data centres to electricity, connectivity, research, education, skills, finance and locally relevant applications.
Announced greenfield investment in AI infrastructure grew by an average of 47% a year between 2020 and 2025, according to the findings reported by UNCTAD. AI infrastructure now accounts for three fifths of all greenfield investment in strategic sectors worldwide. The scale of this shift places computing facilities at the centre of the current investment cycle, but the distribution of capacity and economic value remains uneven.
A handful of economies host the vast majority of AI infrastructure investment, while 100 companies account for 40% of global research and development. These figures point to a concentration problem that is larger than the question of where data centres are built. They also raise questions about whether developing economies can capture value from AI if they remain primarily locations for infrastructure rather than places where applications, skills, businesses and public services are developed.
UNCTAD’s central argument is that computing infrastructure alone does not guarantee inclusive development. Data centres require substantial and reliable electricity, but access to computing capacity is only one part of the chain through which AI becomes useful. Businesses and public institutions also need affordable connectivity, access to data, trained workers, research capabilities and financing to adapt technology to local conditions.
This has direct implications for the way governments plan infrastructure. A data centre can provide computing capacity, but it cannot by itself create a workforce capable of using that capacity, ensure that smaller firms can access tools and markets, or produce applications that address local needs. The wider infrastructure around AI therefore includes physical systems such as power and telecommunications networks, as well as institutional systems covering education, research, enterprise development and data governance.
The urban consequences are particularly significant because the sectors identified by UNCTAD as likely to generate much of AI’s economic value are closely linked to how cities operate. Acting UNCTAD Secretary-General Pedro Manuel Moreno said at the World Investment Forum that most economic value would come from use in agriculture, health systems, logistics and public services. These applications depend on reliable networks and institutions that can integrate technology into everyday delivery systems.
In health systems, for example, the relevant investment is not limited to computing hardware. Public institutions need connectivity, trained personnel, usable data systems and safeguards for handling sensitive information. In logistics, AI applications depend on reliable digital and physical networks. In public services, adoption requires institutions that can use technology effectively and standards that allow systems to work together. The supplied UNCTAD findings do not quantify the outcomes of these applications, but they establish that infrastructure deployment and practical use must develop together.
The same logic applies to local businesses. Smaller firms and innovators in developing economies need access to finance, tools and markets if they are to adapt AI to local needs and grow. Without those channels, investment may increase the availability of computing infrastructure without broadening participation in the industries and services built on top of it.
This is where the geography of AI investment becomes an urban and regional development issue. Concentrated infrastructure can attract capital and create specialised facilities, but the benefits will not automatically spread to nearby businesses, workers or public agencies. UNCTAD’s findings frame local capability as an essential part of the investment model: economies need the skills, research base, connectivity and enterprise ecosystem required to turn infrastructure into wider economic activity.
The electricity requirement is another important part of the equation. UNCTAD specifically links computing infrastructure with reliable electricity and resource-efficient infrastructure. This places AI investment within existing debates about power reliability, network capacity and the competing demands placed on urban and industrial infrastructure. The findings do not provide a country-specific assessment of electricity use or grid capacity, but they make clear that AI infrastructure cannot be treated as separate from the systems that supply and connect it.
Affordable connectivity is equally important. If access to digital networks is limited or costly, businesses, schools, researchers and public agencies may be unable to use the computing capacity being developed. Connectivity therefore functions as a bridge between infrastructure investment and actual adoption. The absence of that bridge can leave the benefits of AI concentrated among large firms and better-resourced institutions.
Skills and research determine whether economies can move beyond consumption of imported systems. UNCTAD calls for investment in education, research and skills alongside computing infrastructure. It also highlights the importance of local-language applications, which would allow AI systems to respond more effectively to the needs of different communities and public services. The findings do not identify particular languages, countries or applications, but they establish localisation as part of inclusive deployment.
The institutional framework is as important as the physical one. UNCTAD identifies transparent governance, data protection, cybersecurity and interoperable standards as conditions that can sustain investment and support wider adoption. These requirements address different points in the AI system. Data protection concerns how information is collected and used. Cybersecurity concerns the resilience of digital systems. Interoperable standards determine whether different systems can exchange information and operate together. Transparent governance affects how these systems are supervised and trusted.
Taken together, these requirements suggest that AI infrastructure policy cannot be reduced to attracting data-centre projects. A broader framework must connect investment promotion with power, connectivity, education, research, enterprise finance and public-sector capability. The supplied findings do not set out a detailed policy model or identify the responsibilities of particular governments, but they clearly place the issue across several administrative and economic domains rather than within technology policy alone.
The available numbers show both the speed and the concentration of the current investment cycle. Greenfield investment in AI infrastructure grew by 47% annually on average between 2020 and 2025. Three fifths of greenfield investment in strategic sectors is now linked to AI infrastructure, while 100 companies account for 40% of global research and development. These figures describe an investment landscape in which physical infrastructure is expanding quickly and research capacity remains concentrated.
That combination creates a gap between the location of assets and the distribution of capabilities. The construction or operation of computing facilities may be visible and measurable, while the less visible requirements—skills, local research, finance, data access and institutional capacity—determine who can use those facilities. UNCTAD’s warning is therefore not that data centres are unnecessary, but that they are insufficient as a development strategy.
The larger urban question is whether AI investment will strengthen the systems through which cities and regions deliver services, or remain concentrated in a limited number of facilities and companies. The answer depends on whether infrastructure plans connect computing capacity to reliable power, affordable networks, capable institutions and local enterprise. It also depends on whether governance arrangements can protect data, secure systems and support interoperability.
UNCTAD’s findings confirm that AI infrastructure investment is growing rapidly and that data centres occupy a dominant position in the global investment landscape. They also establish that inclusive benefits require a wider foundation of electricity, connectivity, education, research, finance, data and governance. What remains uncertain from the supplied material is how individual countries and cities will translate that framework into investment priorities. The next stage of the AI infrastructure cycle will therefore be measured not only by the number of facilities built, but by the breadth of systems and users connected to them.
