Assessing preparedness for AI.

 

With regard to national preparedness for AI, countries may be considered under the following four categories according to adoption and development capacities, as shown in figure III.7: a) Leaders – High capacities for both AI adoption and development. b) Creators – High capacity for AI development, but relatively low capacity for adoption. c) Practitioners – Low capacity for AI development but high capacity for adoption. d) Laggards – Low capacities for both AI adoption and development. The four categories of AI preparedness help assess a country’s current position, illustrating its relative strengths and weaknesses as well as its potential catchup trajectories (e.g. from laggards to practitioners, then to leaders). The following overview of country preparedness uses proxy indicators that have wide country coverage for infrastructure, data and skills. These can be complemented by insights from the frontier technologies readiness index and refined through detailed reviews of STI ecosystems. The analysis uses indicators for intensity and level, to capture different mechanisms influencing AI adoption and development. 

For instance, the proportion of the population with Internet access reflects the potential extent of AI adoption within an economy. Higher levels of data creation and transmission proxy instead a country’s potential for AI development. In assessing national preparedness, comparisons of intensity and level illustrate how the strategic options for AI can be determined by country size. 

On average, developed countries have the highest incidence of Internet penetration and LDCs have less than half of the incidence in developing countries (figure III.8). Similarly, investments in telecommunications services are the lowest among LDCs. Both developing countries and LDCs show high variability in the two indicators. In the top right quadrant, the leaders are largely developed countries in Europe and North America, but also some middle and high-income economies in Asia. In the bottom right, the creators include India and Nigeria, which have high levels of investments in telecommunications services, although less than half their populations have stable Internet access.  

In the top left quadrant, the practitioners have a high capacity for adoption but low capacity for development, and include small upper middle-income and high-income countries such as Seychelles. In the bottom left quadrant, the laggards include several countries in Africa, such as Burundi and Chad, which have low levels of Internet penetration and investments in telecommunications services, and risk being excluded from the development opportunities offered by AI. Some middle-income developing countries show high capacities for both AI adoption and development. In Africa, for instance, Egypt and Morocco exceed the global averages in both indicators. This is partly due to the submarine cables under the Mediterranean that connect them to the European continent and beyond. Egypt, for example, due to its geographical position, and links to more than 160 global submarine cable operators, can become a hub connecting three continents. Between 2009 and 2020, the number of submarine cables to Egypt increased from 6 to 13 and after 2025, is expected to exceed 18. In Asia, the better performers include Malaysia, Singapore and Viet Nam, which have been improving their digital infrastructure. In Malaysia, for example, the Ministry of Digital created the Malaysia Digital Economy Corporation in 1996, aiming to establish the country as a digital hub in the Association of Southeast Asian Nations. In 2023, the Government introduced the digital ecosystem acceleration scheme, to further strengthen digital infrastructure through a series of incentives, such as investment tax credits on capital expenditure. Countries in South-East Asia have generally attracted significant investment from major technology companies. In 2024, to advance new cloud and AI infrastructure, Microsoft announced an investment of $1.7 billion in Indonesia and $2.2 billion in Malaysia. In 2024, Google planned to invest $2 billion in Malaysia, to develop a data centre and cloud hub. In 2025, Amazon Web Services aims to launch a new hub in Thailand and invest $5 billion by 2037. A core element of such investment is cloud infrastructure, which offers computing capabilities and storage with flexible access and at a relatively low cost, thereby supporting AI diffusion among SMEs. Cross-country comparisons are hindered by a lack of internationally comparable statistics, yet it may be noted that cloud computing is strongly concentrated among a few large providers; an indicator of availability is therefore the number of services. With regard to the top 10 economies in terms of cloud infrastructure services from major providers, China and the United States have more services than the rest of the world combined; India and Brazil are two developing countries on the list along with Singapore, and four of the top 10 countries in terms of cloud infrastructure are thus from the Global South (figure III.9). With regard to cloud services by region, it may be noted that even if China is not included, Asia stands out. In addition to China, Japan, the Republic of Korea and Singapore, there are several cloud infrastructure services in South-East Asia. Africa is some way behind. At the end of 2023, eight companies controlled about 80 per cent of the worldwide market share, led by Amazon, Microsoft and Google. These companies may have limited interest in countries that do not generate enough data traffic and profits, which could contribute to deepening digital and AI divides between countries.

ITU has set the affordability target for fixed broadband at 2 per cent of gross national income per capita. On average, developed countries score better in data affordability, with many developing countries and LDCs still far from the ITU target (figure III.10). 

The gap between developed and developing countries for data traffic is narrower, with LDCs lagging behind. Among the leaders, China performs well in both affordability and data quantity. A number of high-income economies, such as Hong Kong (China), Germany, the Russian Federation, the United Kingdom and the United States, also have a wealth of data that can be used to train and develop AI systems. Creators include Pakistan and the Bolivarian Republic of Venezuela, which have low levels of adoption but a high development potential.

Practitioners include smaller economies such as Eswatini, Kuwait and Monaco that have high levels of AI adoption but a relatively low development potential; their small populations limit the data available for local AI models. Laggards, which show low potential in both AI adoption and development, are mostly developing countries in Africa and Latin America and the Caribbean. China has the world’s greatest fixed-broadband traffic, due to its large population and because it has significantly reduced fixed-broadband prices, from around 5 per cent of gross national income a decade prior to 0.5 per cent at present, which is about one sixth of the global median. The Government has put in place regulatory reforms to increase competition among Internet service providers while encouraging new market entrants. The fibre-optic network has been upgraded and expanded to enhance connectivity in rural and underserved areas.

Financial incentives to Internet service providers have lowered costs for consumers, and fair pricing has been promoted by consumer protection measures and price caps. Additional information on data preparedness is available by analysing the number of Internet exchange points. These are physical locations where Internet service providers connect and exchange traffic between their networks and are a crucial element of middle-mile digital connectivity. Traffic per Internet exchange point is highest in high-income countries, although the average number of members per point is highest in upper middle-income economies, partly because they host some of the world’s largest Internet exchange points, such as Ponto de Troca de Tráfego Metro São Paulo in Brazil, Qianhai New-Type Internet Exchange in China and Moscow Internet Exchange in the Russian Federation. Low middle income and low-income economies show low values for both Internet exchange point traffic and membership (figure III.11).

European Internet exchange points are well-established with many years of experience; they generate the highest traffic volume and have the highest number of members per Internet exchange points. In contrast, Africa is far behind, with limited participation and data flows

GitHub is a major platform through which developers can collaborate, and hosts a large number of open-source projects. Country groupings illustrate the differences in AI skills preparedness, with LDCs scoring rather low in both GitHub developers as a share of the working-age population and the proportion of the working-age population with tertiary education. With some noticeable exceptions, developed countries rank better than developing countries in both indicators (figure III.12).

The leaders in the top-right quadrant are mainly developed economies, such as Canada, Ireland, the Republic of Korea, and the United States. Hong Kong (China) and Singapore have particularly high numbers of GitHub developers. Countries in the bottom-right quadrant have low AI adoption but high development potential and include developed economies in Europe, such as Romania, and some island countries such as Maldives and Seychelles. There are relatively few economies with high potential in AI adoption but low development capacity.

 In fact, most developing economies display relatively low skills capacity for both adoption and development. The proportion of developers in the population does not tell the whole story. Large countries may have a low proportion of developers, but this could still represent a substantial body of developers on which to build AI development advantages. 

The United States has the most GitHub developers, followed by India and China (figure III.13). 

China and India have the world’s largest populations and, despite relatively low shares, can leverage a significant mass of AI developers, which puts them in favourable positions with respect to AI development and the production of AI-related scientific knowledge. Many developing countries have achieved rapid growth in the number of developers (figure III.14). 

The fastest increase, at 40 per cent, was in Nigeria, Ghana and Kenya, which have become promising hubs for technology companies. The growth in developer numbers is also notable in Latin America and the Caribbean, for example in Argentina, the Plurinational State of Bolivia, Colombia and Brazil. In Asia and the Pacific, India, Viet Nam, Indonesia and the Philippines already had a significant number of developers but had increases of more than 30 per cent.  Many students in Asia perform well in the Programme for International Student Assessment, particularly in science and mathematics, signifying a strong potential for both AI adoption and development. There are large talent pools in India, with around 13 million developers, and in Brazil, with 4 million. These two countries are also among the leading countries in creating GenAI projects on GitHub, and are significant contributors to advances in AI. The lead of India partly reflects government policy. The Government has closely collaborated over the years with the private sector and academia to build centres of excellence, such as the Indian Institute of Technology Hyderabad and the Indian Institute of Technology Kharagpur in AI, the Kotak Indian Institute of Science Artificial Intelligence–Machine Learning Centre and the National Association of Software and Service Companies centre of excellence in data science and AI. In 2024, the Cabinet approved the India AI mission to strengthen the AI innovation ecosystem, aimed at, for example, reducing barriers to entry into AI programmes and increasing the number of AI courses in tertiary education, focusing on small and medium-sized cities. Brazil has also been cultivating AI talent, at both the federal and state levels. For example, through strategic partnerships between public and private institutions, the Research Foundation of the state of São Paulo has created a network of applied research centres. The initiative is also aimed at creating scholarships to attract researchers and further boost performance in terms of AI publications. These approaches highlight the importance of training AI specialists to sustain the development of a strong and diffused AI ecosystem and attract and cultivate AI talent.