HBR Staff; oxygen/Getty Images; Unsplash
Summary. As the AI power centers emerge and shift around the world, they will shape which AI applications are prioritized, which societies and sectors of the economy get the most benefits, what data are used to train algorithms, and which biases get included and which get...
This year has been the breakout one for artificial intelligence. AI talk is now everywhere, from businesses and schools to Hollywood and election campaigns. Even as some investors complain of AI fatigue, a riveting drama explodes over the leadership of OpenAI — arguably, the most prominent of players at the technology’s frontiers — bringing our attention back to the many unresolved issues around AI. Meanwhile, there is no escaping its enormous potential. Generative AI alone could affect 300 million jobs and create as much as $4.4 trillion annually in new economic value worldwide, according to some analysts. It is not just the tech powers in competition to capture the value; there is a global race among nations for AI leadership — an emerging geography of AI.
This race will determine which applications get priority, where innovative capacity and investments can be focused, what AI regulations emerge, what risks might arise, which biases and data deficiencies get heightened or mitigated, and whether competitive innovation gets prioritized over safety and public oversight. The geography of AI is key to the technology’s future.
Already, there are a few clear axes. The U.S. and China are both vying to be the world’s top AI economy, locked in a “digital cold war,” but the larger cast of nations is evolving. The EU had led the effort to regulate AI in democratic societies, and now the U.S. is playing catch-up, while Canada is the first country with a national AI strategy. As some countries tighten AI regulation, others might lure cutting-edge companies by promising unfettered “pro-innovation” environments. Alternatively, others might attract those that prefer safety or openness. Populous developing countries, such as India, are aiming to be the leading data-rich nations, with fast-growing pools of data. Despite operating under sanctions, Iran has declared its aspirations to be among the world’s top 10 in AI. This is likely to ratchet up worries about the national security risks of AI and put pressure on other actors in the region to aim for a similar goal.
Given the high stakes of this race, which countries are in the lead? Which are gaining on the leaders? How might this hierarchy shape the future of AI? Identifying AI-leading countries is not straightforward, as data, knowledge, algorithms, and models can, in principle, cross borders. Even the U.S.–China rivalry is complicated by the fact that AI researchers from the two countries cooperate — and more so than researchers from any other pair of countries. Open-source models are out there for everyone to use, with licensing accessible even for cutting-edge models. Nonetheless, AI development benefits from scale economies and, as a result, is geographically clustered as many significant inputs are concentrated and don’t cross borders that easily.
Using data from over 20 different institutional sources — including public databases such as the ITU and the World Bank and proprietary data partnerships such as SeekOut and George Washington University’s Data Governance Hub — as well as our own databases and models from Digital Planet, we’ve mapped that emerging geography of AI leadership, identifying where — and how — the forces that drive AI development are lining up.
Drivers and an Index of AI Leadership
Rapidly accumulating pools of data in digital economies around the world are clearly one of the critical drivers of AI development. In 2019, we introduced the idea of “gross data product” of countries determined by the volume, complexity, and accessibility of data consumed alongside the number of active internet users in the country. For this analysis, we recognized that gross data product is an essential asset for AI development — especially for generative AI, which requires massive, diverse datasets — and updated the 2019 analyses as a foundation, adding drivers that are critical for AI development overall. That essential data layer makes the index introduced here distinct from other indicators of AI “vibrancy” or measures of global investments, innovations, and implementation of AI.
To compare AI across countries, we considered four drivers:
Data — the volume and complexity of the core resource used to train and improve algorithms.
- Broadband consumption, aggregate (fixed and mobile): The overall data consumption in a country.
- Broadband consumption, per capita (fixed and mobile): Data usage per internet user in a country, which serves as a proxy for how complex the data is representing different kinds of uses.
Rules — how data can be accessed.
- Open data participation: The degree to which an economy promotes the use of and access to public data sources.
- Data governance policies: The country’s regulatory approach to data — personal, non-personal, open, proprietary, public, and private — particularly with respect to privacy protections.
- Cross-border data flows: The degree to which an economy promotes and engages in data flows with other economies, as well as the degree to which an economy actively localizes data within its borders.
Capital — the human, financial, diversity, and digital foundations for building AI.
- Talent: The quality and quantity of AI talent available.
- Investment: Investment flows into AI and emerging technologies.
- Diversity: Diversity of AI talent.
- Evolution of the digital economy: The evolution of a country’s digital foundations, including computational capabilities.
Innovation — advances in AI models, techniques, creative sourcing of data, and new applications.
- Number of patent applications: The number of patent applications from each country in AI-related technologies.
- Number of citations for top 10 AI papers: The total number of citations accrued by authors from each country.
- Aggregate AI publications: The total number of publications in the field of AI in each country.
Putting these variables together, we derived a new measure — the Top Ranked AI Nations (TRAIN) index — to evaluate where 25 leading AI creator countries stand in the race for leadership, as displayed in Figure 1. It is not meant to be the complete list of countries that are playing critical roles in shaping the global AI industry. Countries such as Israel or the UAE, for example, are key players that were not part of our evaluation, since they are still small and there were data limitations across the drivers we considered.
See more HBR charts in Data & Visuals
The Emerging Geography of AI: The U.S. vs. China
The fact that the U.S. and China occupy the top two spots on the TRAIN index is not a surprise, as governments in both countries have committed to being the global AI leader. U.S. National Security Advisor, Jake Sullivan, declared the goal was to ensure “as large of a lead as possible” in frontier AI technology. Meanwhile, China aims to become the world’s “primary” AI innovation center, with AI-related gross output exceeding RMB 10 trillion ($1.5 trillion) by 2030. The competition has intensified further by the White House banning U.S. companies from exporting chip-making equipment in October 2022, a requirement that will now be extended to include AI chips.
The U.S. has historically been ahead on most of the key drivers, and American AI firms have models that significantly outperform their Chinese rivals. On the capital dimension alone, the top four global cities for AI talent, investments, diversity of talent, and the evolution of the digital economy are in the U.S. The top Chinese city, Beijing, comes in at number eight. Moreover, venture investments funded 524 AI startups in the U.S. in 2022, placing it far ahead of all other countries. U.S. AI companies attracted over two-and-a-half times more private investment compared to China’s in the last 10 years.
The private sector is the essential driving force behind AI in the U.S.; its share in the biggest AI models has spiked from 11% in 2010 to 96% in 2021, and 70% of PhDs in AI-related fields are employed by the private sector. The intense competition among U.S. AI players will likely continue to power the U.S.’s lead on the innovation dimension.
In China, the government plays a bigger role in AI development. It has used significant subsidies, support, and policy guidance to direct it toward applications such as drug development, gene research, and biology. And because of government protection, China’s domestic AI industry operates in a market with high barriers to entry for international competitors. China has the world’s largest internet-enabled population and, consequently, adoption can happen at remarkable rates. For example, China’s generative AI Ernie Bot reached 1 million users in 19 hours, while ChatGPT took five days to get to the same threshold.
China has a few important advantages that could allow it to challenge the U.S. in the future. For one, it generates and consumes a massive amount of data. As a result, China will increasingly have faster-growing pools of data, which are among the least-accessible to AI developers outside China — a factor that could both hamper and enhance China’s AI leadership capabilities.
Second, China is home to the world’s fastest-growing AI research community, with Chinese authors contributing to top AI journals at a rate roughly 2.5 times higher than U.S. contributors.
Third, China is a first-mover on AI regulations, and even with the recent ambitious Executive Order from the White House on regulating AI, China will have an experiential lead in this area. Thus far, AI regulation is still in a nascent phase, and it hasn’t factored as significantly in the TRAIN index, but over time that is likely to change. On the other hand, China must contend with many challenges. Regulation and state restrictions could diminish its innovative capacity, and chip scarcity is a critical constraint in the near term. Restrictions on data flowing into and out of China could limit its ability to develop cutting-edge AI models.
The U.S.–China competition has implications for AI development globally. A decoupling of China from the U.S. and other major centers of AI development could slow down AI advancement overall. Also, China seeking out its own chips and independent standards could lead to a “forking” of AI development, and fragmentation of the initiatives and datasets needed to train robust algorithms.
How Other Countries Stack Up
Of course, the geography of AI extends beyond these two countries, and the positions on the TRAIN are not stationary. There are several countries to watch as the AI landscape evolves, such as India, Indonesia, the U.K., the major E.U. countries, as well as Japan and South Korea. Some countries have faster-growing pools of data, while others make the data more accessible. Yet others have demographic factors that will affect their positions on the TRAIN index.
Of all the drivers that are crucial for AI leadership, the changes in accessible pools of data are likely to have the greatest impact on the positioning across the 25 countries in the near to medium term.
To get a sense of some key changes to watch, consider Figure 2, where we juxtaposed the size and momentum of aggregate data pools on the X-axis against the current TRAIN score of countries on the Y-axis; the sizes of aggregate data pools are represented by the size of the bubble and colored to indicate the accessibility of data.
See more HBR charts in Data & Visuals
While the capital dimension provides the strongest boost to a country’s AI leadership rankings, the change in its data pool is critical to its ability to move up on the TRAIN table. As Figure 2 shows, however, countries with the fastest-growing pools of data are also ones with greater restrictions on access to data. This means that companies should watch for changes in regulations and policies relating to data in different parts of the world — they could determine whether AI development activity ought to be shifted toward a particular location. Moreover, policymakers in each of the countries affected by the shifts must reconsider their own regulations and investment priorities to retain and enhance their AI leadership positions.
Beyond the two leaders, countries such as India, the UK, France, Canada, Germany, and Australia are strong in the capital dimension. It is for this reason that, in the medium term, these are the countries that are poised to make the most significant advances. Meanwhile, since growth in data pools is also a major contributor, countries such as Indonesia, South Africa, Nigeria, and India ought to be watched closely, as they have the greatest rates of change in the aggregate volumes of data consumption. Of these, India and Indonesia have the largest absolute pools of data. Enhancing access to their data could improve these countries’ TRAIN standing in the future.
With strengths across many dimensions, India is the one with greatest upward potential. It has the largest volume of mobile data consumption and is expected to top the world in data consumption by 2028. It already processes more digital payments than any other country in the world and has the third-largest AI talent pool. While it places restrictions on access to data, its AI regulation rules are still fluid. In July, the Telecom Regulatory Authority of India issued a new consultation paper that called for a statutory authority to regulate AI in India through the lens of a “risk-based framework.” It also suggested collaborating with governments and international agencies to advance the “responsible use” of AI globally — a process that India could be quite significant in shaping.
As for the UK, watching how it competes with EU countries — specifically France and Germany — reveal pros and cons of two distinctly different approaches. The UK’s AI industry is supported by a national strategy but is committing to a light-touch approach to regulation that aims to be “pro-innovation” as the field develops. Indeed, the UK is one of the most innovative AI countries, home to companies such as DeepMind (acquired by Alphabet), whose work on protein structure could have breakthrough implications in areas from drug discovery to food security. The UK has tried to balance its light regulatory touch with public efforts to lead on AI safety, including coordinating a declaration encouraging global cooperation on the issue, and the launch of a UK AI Safety Institute to conduct safety evaluations of frontier AI systems.
In contrast, the EU AI regulations could slow down AI development in the member countries when they’re eventually applied. This could help the UK maintain its current lead over countries such as France and Germany, which have committed to a common AI ecosystem to set up new collaborative projects through a joint declaration of intent and a “Research and Innovation Network in Artificial Intelligence.” Offsetting potential regulatory friction, data pools in France and Germany have been growing faster than those in the UK and there was intense debate over reconsideration of the stringency of the EU regulation due to lobbying from industry. Ultimately, the regulations have settled on a compromise two-tier approach requiring “transparency” from all but the biggest of the foundation models. The crisis at OpenAI, where CEO Sam Altman was fired and then reinstated with an even stronger role anticipated for its biggest investor, Microsoft, could create an opportunity for emerging European players to position themselves as more responsible and open alternatives to those in the U.S. Companies such as Kyutai and Mistral AI may be preparing for such a role, potentially positioning Paris as a “trustworthy AI” hub.
Next, consider Japan and South Korea, both countries with a strong need to develop AI because of their own demographic and growth priorities that have invested heavily in the area. Both have invested in robotics and AI to supplement human work. Both face headwinds as we consider the future. Their data pools aren’t growing as fast as some of the other Asian countries with younger demographics and growing numbers of internet users. Each faces different challenges.
Japan, for example, will face a deficit of 789,000 software engineers by 2030, according to the country’s Ministry of Economy Trade and Industry, which could, in turn, limit its capacity in deep learning and software development and slow down its advances in generative AI. It also lacks access to adequate supercomputing capabilities, which constitutes yet another bottleneck. Japan’s SoftBank intends to help shift the country’s position from “defense mode” to “offense mode,” but the extent to which it can be successful remains to be seen.
South Korea has several core strengths to sustain its advantages for a while — for example, its semiconductor industry and leadership in AI patents and research. However, policymakers and experts are concerned about a talent shortage and inadequate government support. The government’s AI R&D budget was cut by 43%, according to one legislator; this will add to the challenges that South Korea might face in the future, as its cumulative investments in AI from 2013 to 2022 was only $5.5 billion, far behind the U.K.’s $18.2 billion, and its six AI-related listed firms compare poorly against Japan’s 26.
- • •
The global race for AI leadership is unsurprising, given the technology’s potential reach and impact. As former Google CEO Eric Schmidt reminds us, widely used technologies have always been critical to determining which nations become dominant: Gunpowder established the Ottoman, the Safavid, and the Mughal empires; muskets helped the conquistadors overwhelm the Incas; and the Industrial Revolution put Germany and the UK ahead of Russia. Today’s Russian president Vladimir Putin famously said that whoever leads in AI will be the “ruler of the world.”
As the AI power centers emerge and shift around the world, they will shape which AI applications are prioritized, which societies and sectors of the economy get the most benefits, what data are used to train algorithms, and which biases get included and which get neutralized — and how we balance accelerating AI innovation against building in safeguards. It is essential that business and policy leaders pay attention, as the geography of AI will determine the future of AI and its usefulness to societies everywhere.
The authors are grateful to Paul Trueman of Mastercard and Christina Filipovic, Iris Xue Niu, Stella Henderson, and Max Agigian at The Fletcher School’s Digital Planet.
