The mostly debated issue about AI is that this new technology can displace workers doing cognitive tasks, exacerbating in some extent income disparities hitting the individuals living of their wage income, and favoring capital owners with respect to wage earners. A less debated issue, but which is extremely relevant, is whether AI promote or mitigate income disparities between and within countries.
A recent global study titled Artificial Intelligence and income growth divide: Evidence from global regional data takes a close look at these questions. Rather than focusing on countries as a whole, it zooms in on regions within countries. By examining more than 600 regions across 43 countries between 2005 and 2020, the study offers one of the first most comprehensive pictures to date of how AI innovation relates to income growth around the world.
The first striking fact is how concentrated AI innovation really is. While Artificial Intelligence is often described as a global revolution, its creation is heavily clustered. A handful of countries dominate AI patenting activity. Among them are the United States, China, Japan, Germany, and South Korea. Together, they account for roughly three quarters of global AI patent output in the period studied.
Even within those countries, AI activity is far from evenly spread. In the United States, AI patenting is concentrated in a few states such as California and Washington. In China, regions like Shenzhen, Beijing, and Shanghai stand out. In Europe, parts of Germany, France, the United Kingdom, Sweden, and the Netherlands host much of the continent’s AI innovation. The study shows that in some countries, the top 5 percent of regions account for more than half of all AI patenting. In China and South Korea, the concentration is even more extreme.
Innovation has always tended to agglomerate in urban hubs where universities, skilled workers, venture capital, and large firms come together. What is new is the scale and speed at which AI has emerged as a technological frontier. The question is whether this new wave of digital innovation reinforces old patterns of inequality or offers new opportunities for catch-up.
To explore this, the work builds a novel dataset tracking AI-related patents at the regional level. Patents are not a perfect measure of innovation, but they provide a useful and internationally comparable way to track technological activity. By combining this with data on regional GDP per capita, the study estimates how changes in AI patenting are associated with income growth over time.
The headline result is nuanced. AI innovation does have a significant positive effect on regional income growth. Regions that increase their AI patent stock tend to grow slightly faster than others. However, the size of the effect is modest. AI appears to contribute to growth, but it is not a magic wand that instantly boosts prosperity. This finding helps temper some of the more exaggerated claims about AI delivering explosive economic gains across the board. It also echoes earlier debates about the so-called productivity paradox, in which rapid technological progress does not immediately translate into dramatic gains in measured output.
One of the most interesting findings concerns the distribution of these gains. The positive impact of AI on income growth appears to be stronger in regions that are at the lower end of the income growth distribution. Put simply, slower-growing regions may benefit relatively more from AI innovation than fast-growing ones.
This suggests that AI has the potential to support convergence rather than divergence, at least under certain conditions. Regions that have been lagging behind might use AI as a lever to accelerate growth. However, this effect is not automatic. It is also slightly stronger in regions that are already specialized in AI-related technologies. That means having a base of digital capabilities and related knowledge can enhance the benefits.
The global context matters as well. Over recent decades, patterns of income convergence have shifted. Rapidly growing Asian economies have narrowed the gap with advanced Western countries. At the same time, Asia has become a powerhouse in digital technologies and AI research. According to international data, China and India now account for a substantial share of AI-related scientific publications, and China leads the world in certain categories of AI patents.
This raises an important possibility: the geography of AI may be reshaping the global economic landscape. Regions that successfully integrate into AI-driven innovation systems could leapfrog older industrial structures. But those that lack the necessary skills, institutions, or complementary technologies may struggle to keep pace.
The study carefully considers whether the observed relationship between AI and growth might be driven by other factors. Regions that are good at AI may simply be good at innovation in general. They may have better universities, more educated workers, stronger infrastructure, or better access to global markets. To address this, the analysis controls for general patenting activity, demographic characteristics, labor force participation, and country-level variables such as human capital and trade openness.
The analysis reveals that the effect of AI on regional income growth is still small. One explanation is that technological revolutions often take time to diffuse. In the early stages, benefits may be confined to a narrow set of firms or sectors. Complementary investments in skills, organizational change, and infrastructure are often required before productivity gains become widespread. History offers many examples, from electricity to information technology, where the full economic impact only materialized after years of adaptation.
Another explanation is that AI’s benefits may be unevenly distributed within regions. Firm-level studies often find substantial productivity gains for companies that adopt AI technologies. But these gains do not necessarily translate into broad-based regional income growth if they are concentrated among a small group of firms or workers. In some cases, AI could even increase inequality by favoring high-skilled workers and capital owners over others.
The regional lens of this study is especially valuable because it highlights subnational disparities. National averages can mask enormous internal variation. A country may appear to be benefiting from AI overall, while large swaths of its territory see little direct impact. Policymakers concerned with regional cohesion need to understand these dynamics.
The evidence suggests that simply waiting for AI to spread organically may not be enough to ensure balanced development. Regions that lack pre-existing digital capabilities may need targeted support. Investments in education, digital infrastructure, research institutions, and innovation ecosystems could help create the conditions under which AI contributes to local growth.
At the same time, the finding that slower-growing regions may benefit relatively more from AI offers a note of optimism. Emerging technologies can open windows of opportunity. Regions that were not leaders in older industrial paradigms might still carve out niches in new technological domains. The key lies in building related capabilities and connecting to broader knowledge networks.
The global nature of AI also raises questions about international competition and cooperation. As leading economies race to dominate AI research and applications, concerns about technological sovereignty and strategic autonomy are growing. Differences in intellectual property regimes, data governance, and regulatory approaches may influence where AI innovation flourishes. The study’s use of variation in intellectual property rights as part of its identification strategy underscores how institutional frameworks shape technological outcomes.
For a general audience, the main takeaway is both simple and subtle. Artificial intelligence does matter for economic growth. Regions that generate more AI-related innovation tend to grow slightly faster. But AI is not a silver bullet. Its impact is incremental rather than explosive, and it interacts with a broader ecosystem of skills, institutions, and complementary technologies.
The future trajectory remains uncertain. The data in the study cover the period up to 2020, just as generative AI tools were beginning to enter mainstream awareness. Since then, advances in large language models and other AI systems have accelerated. It is possible that the economic effects of AI will intensify in the coming years. It is also possible that measurement challenges will continue to obscure the full picture.
What is clear is that the geography of AI will play a central role in shaping the economic map of the twenty-first century. Whether AI becomes a force for convergence or divergence will depend not only on algorithms and patents, but on policy choices, institutional quality, and the ability of regions to adapt and innovate.
The study provides a careful, data-driven contribution to this debate. By grounding the discussion in detailed regional evidence, it moves beyond hype and fear. It shows that AI is already influencing growth patterns, albeit modestly. It also suggests that the door remains open for lagging regions to harness AI as part of their development strategy.
As societies grapple with the promises and risks of artificial intelligence, understanding its economic geography is essential. Growth is not just about how much technology is invented, but where it is invented and how widely its benefits spread. The story of AI and income growth is still unfolding, but the early evidence points to a world in which technology reinforces some existing hubs while offering new opportunities for others willing and able to seize them.
References
- Jabłońska, J., Parteka, A., Venturini, F. (2025). Artificial Intelligence and income growth divide: Evidence from global regional data. Gdansk University of Technology, mimeo
Francesco Venturini
