Artificial intelligence (AI) is a significant advance that has the potential to reshape businesses, economies, and societies. However, a major problem is emerging as AI technology progresses – the escalation of global disparities. The gap between nations with strong digital infrastructures and those without is widening, raising concerns about the equitable distribution of AI’s advantages.
Internet connectivity and affordability
The disparity between AI readiness and adoption is stark. Research by the International Monetary Fund indicates that high-income countries have the necessary infrastructure, talent, and resources to develop and implement AI technology efficiently. In contrast, existing digital divides, particularly in terms of internet access and broadband affordability, put low- and lower-middle-income countries at a disadvantage.
High-income countries have widespread, affordable internet access, with fixed broadband costs of just 1% of monthly income which is below the UN’s 2% target. In upper-middle-income countries, the cost climbs to 3%, reaches 8% in lower-middle-income countries, and a staggering 31% in low-income countries with limited connectivity.
These differences mean that wealthy nations are better positioned to use AI for productivity and innovation in vital areas such as banking, medicine, manufacturing, and defence.
See also: Should Artificial Intelligence be regulated? | Experts’ Opinions
On the other hand, AI threatens to undermine one of primary economic advantages of developing nations – cheap labor costs. As wealthy nations automate manufacturing and logistics, they lessen their reliance on outsourced labor, making it more difficult for poorer countries to compete.
Infrastructure and compute capacity
Access to computational resources is a crucial aspect of AI development, but the distribution of these resources is fairly uneven. Countries such as Indonesia and Malaysia are investing in data centers, with Malaysia’s Johor province becoming Southeast Asia’s fastest-growing data center market, according to Asia Pacific Data Centre H1 2024 Update.
Africa, on the other hand, has just over 100 data centers, less than 1% of global data centre capacity, despite housing 18% of the world’s population. This lack of infrastructure reduces the continent’s ability to process and store data locally which results in increased slowness and reliance on foreign servers. This scarcity also hampers the ability of low human development index countries to develop and deploy AI technologies effectively.
Another infrastructural problem is access to a reliable power supply. Power accounts for a significant percentage of data center expenses and generative AI systems use approximately 33 times more energy to finish a task than task-specific software. However, many African countries face chronic energy shortages and unreliable grids. For instance, Nigeria’s national grid collapsed 46 times between 2017 and 2023, leading to frequent blackouts that disrupted data centre operations.
These power issues force data centers to rely heavily on diesel generators, which are expensive to operate and maintain. This reliance on costly back-up power sources increases operational expenses and hinders the scalability of AI infrastructure.
The high cost of graphic processing units (GPUs), essential for AI computations, is also a major hurdle. In Kenya and Senegal, the price of a GPU represents 75% and 69% of GDP per capita, respectively. Such financial barriers further entrench technological disparities.
Talent and data disparities
The development of AI technologies requires a skilled workforce, including data scientists, machine learning engineers, and AI researchers. However, the distribution of such talent is heavily skewed towards High-Income Countries (HICs).
Nations such as the United States, Israel, and Japan have a large concentration of AI professionals. For example, Israel has over 140 scientists and technicians per 10,000 employees, which is one of the highest ratios in the world. The U.S. dominates in AI research output and created 61 significant AI models in 2023, leaving the EU with 21 and China with 15 far behind.
Lower-middle-income and low-income countries face substantial obstacles in developing AI talent. Less than 2% of LIC graduates specialize in computer science or related subjects with the reasons for this being underfunded universities, the lack of computer labs, outdated curricula, and minimal internet access. This dramatically limits the local capacity to develop AI tools, train models, or contribute to the global digital economy whereas over 10% of graduates in HICs specialize in these areas because they have well-resourced universities, grants, and private-sector pipelines that feed into AI innovation.
Meanwhile, just 5% of Africa’s AI talent has access to the computing resources required to complete complicated jobs.
Data availability and quality are also vital for AI development. However, many low-income countries lack the infrastructure to collect, store, and process large datasets. This data divide results in AI models frequently being trained on data that fails to represent the diversity of global populations which creates biases and less effective solutions for underrepresented communities.
The concentration of AI capabilities in developed nations has far-reaching economic implications. While AI is expected to contribute around US$13 trillion to global GDP by 2030, much of this growth is likely to be captured by wealthier economies thanks to their advanced infrastructure and greater capacity to invest in innovation.
To counter this imbalance, Yanis Ben Amor, Executive Director at Columbia University’s Center for Sustainable Development, stresses the importance of equipping low- and middle-income countries with the human capital needed to remain competitive. He advocates for universities in these regions to begin training a new generation of AI-enhanced workers who will be able to adapt to the rapidly changing job market over the next two decades.
Building on this, Sarah Choudhary, CEO of Ice Innovations, highlights the critical role of academia-industry partnerships in cultivating such talent. She points to internships, apprenticeships, and hackathons as practical pathways to bridge the gap between classroom learning and real-world application to ensure that students are prepared for the demands of an AI-driven economy.
But developing talent alone will not be enough. As Brian Reiff of the World Economic Forum explains, the broader challenge lies in making AI itself accessible and equitable. To truly bridge economic divides, he argues, AI must be affordable and guided by inclusive principles. This will necessitate cooperation between governments, tech companies, and non-profits to ensure global, not just Northern, distribution of its benefits.
The United Nations’ Mind the AI Divide report echoes this sentiment, warning that without intentional, inclusive strategies, the AI revolution could widen the gap between high- and low-income countries. The report demands coordinated international efforts for investment in digital infrastructure, policy reform, and human capacity to prevent any nation from being left behind in the algorithmic age.
