The Global Artificial Intelligence Index 2024

Two years after the launch of ChatGPT, generative AI continues to dominate the AI landscape, driving investments, policies and business operations.

But it’s been a busy year all round. In the last twelve months, the UK gathered international governments for the first AI safety summit and co-hosted a second one in South Korea; Europe passed the first comprehensive AI law; industry players released high-performing models, and global private investment in generative AI almost doubled.

While some influential figures have questioned the sustainability of the AI investment wave, AI technologies are continuing to attract attention, and governments remain interested in their transformative potential.

The Global AI Index by Tortoise, now in its fifth iteration, is an attempt to make sense of this complex landscape.

It reveals which countries are leading the global AI race and why, and uses a combination of absolute and relative indicators to measure countries’ total AI capacity (scale) and their AI capacity relative to their population and economy size (intensity).

This year the index expands the list of examined countries to 83, with new indicators and improved data organised across three pillars: Implementation, Innovation and Investment. Together, the pillars cover talent, infrastructure, operating environment, research, development, commercial ecosystem and government strategy.

Key findings

The United States continues to lead the Global AI Index rankings. This year it extends its lead over China, which retains second place. The two superpowers are ahead of all other countries in the index by a significant margin.

Singapore maintains third place, confirming its position as Asia’s most dynamic AI hub after China. It scores highly on most relative indicators (e.g. number of AI scientists per million people). But the country has also made big advancements on absolute metrics, especially on AI research and investment.

The UK narrowly holds on to fourth. But France is snapping at its heels. The country has jumped to 5th place, entering the top five AI nations for the first time. While the UK is strong on commercial AI, France now outperforms it on open-source LLM development and in other key areas including public spending and computing.

South Korea is still sixth in the rankings, and it is particularly strong in applying AI in key industrial sectors. Other countries in the top ten include Israel, the third most popular destination for private AI funding, and Canada, which has the third most comprehensive government strategy for AI.

India enters the top ten in the index for the first time, with a particularly strong and diverse AI workforce, driven by demographics and academic excellence. But much of India’s AI talent moves overseas, and the country’s success in AI has not yet translated into high private investment levels and computing capacity.

Governments are stepping up AI funding across the board, especially in Saudi Arabia whose public AI spending commitments significantly outrank those of the US and China.

Overall AI private investment remains relatively flat, but generative AI funding has increased. The global market for generative AI is largely dominated by the US, specifically by a handful of tech giants.

Scale and intensity

Singapore, Israel and Switzerland lead the ranking on ‘intensity’ – the analysis tool of the Global AI Index that measures AI capacity in relation to a country’s population and economy size.

From computing power to capital investment, cutting-edge AI development and deployment is increasingly dictated by scale and concentrated in large technical institutions. This shift is cementing the dominance of major players like the US. But in isolating relative indicators (e.g. the number of AI scientists per million people) we can also identify the small yet dynamic AI players which are playing an outsized role in this space*.

Like last year, Singapore tops the ranking for intensity, performing well by all relative indicators. But its performance by absolute metrics is also notable. For example, it gets close to challenging China and the US on AI infrastructure and patents. Similarly, Israel scores highly on absolute investment, mainly due to large amounts of private funding going to Israeli AI companies.

* This is reflected in the index weighting, with 75/25 per cent ratio between absolute and relative indicators. 

Talent

The geographical concentration of AI specialists, their movements and degree of expertise remains key to AI capacity. The Global AI Index measures AI talent by looking at:

• AI-related activity on online software development platforms;
• AI scientists and professionals.

The US hosts the vast majority of AI talent, which includes AI scientists, engineers and researchers in the field, and is a global magnet attracting the best and brightest from other countries. For example, it hosts 27 per cent of the global pool of advanced career AI scientists.

India comes second in talent*. This is partly due to sheer demographic weight, but it is also thanks to the country’s strong network of higher education institutions, and its global position as an IT outsourcing powerhouse.

Within talent, India is particularly strong when it comes to developers or professionals working with AI. But when looking at the narrower pool of highly specialised AI research scientists, India ranks behind many smaller nations (24th).

Over the past five years, the number of scientists switching from academia to industry has markedly increased. In most countries, over half of AI scientists are in industry positions. This is also true for India, although, in comparison to countries like France and Germany, a relatively high proportion of its AI scientists stay in academia. India’s AI commercial ecosystem still remains relatively small.

*Although it is likely to be third behind China, which is under-measured by the index due to data limitations regarding AI talent.

Infrastructure

AI is embedded in software, but hardware is becoming key for cutting-edge AI training and deployment. This year, therefore, the index focuses more on the computing infrastructure that underlies modern artificial intelligence, which includes measuring:

• National supercomputing capacity;
• Involvement in the highly complex global semiconductor manufacturing process;
• Access to and usage of high-end GPU chips for large-scale AI training.

The US maintains an edge on the market for AI computer chips through Nvidia – the hottest stock in the world – while its rivals scramble to produce competitor chips. Tortoise’s global analysis of the cited use of Nvidia GPUs in AI research publications reveals that US export bans on GPUs to China have only been partly effective.

Chinese researcher use of Nvidia’s powerful H100 and A100 chips lags behind the US, but continues to increase year on year with Chinese institutions circumventing export bans to acquire the high-end chips. Chinese researchers are also turning to Nvidia’s various less powerful, general purpose RTX-series GPUs, the use of which has exploded in China in recent years, overtaking cited use in US research papers.

Nvidia GPUs are the end product of a highly complex semiconductor supply chain, where the manufacture of various components are monopolised by key national players, most famously ASML and TMSC. As reflected in the index, semiconductor manufacturing is an area dominated by East Asia: after the US, the next five highest ranking countries in the Infrastructure sub-pillar are China, Singapore, Japan, Taiwan and South Korea.

However, success in infrastructure is not necessarily a recipe for AI success elsewhere. South Korea and Singapore are strong overall, but other leaders in semiconductor infrastructure (e.g. Japan, Taiwan and the Netherlands) are all AI underperformers relative to their size and wealth.

Operating environment stands for the social, legislative, economic and cultural factors that affect the implementation of AI technologies. The Global AI Index measures countries’ operating environment by analysing:

• AI in legislation;
• Public trust in AI;
• AI labour mobility.

Overall the public perception of AI has improved in the past year, especially in high-ranking countries such as the UK, South Korea, Canada, France, Italy and Japan. But a greater percentage of people trust AI in lower ranking countries such as Thailand, Malaysia and Indonesia.

Global mentions of AI in legislative proceedings have never been higher. The UK tops the ranking, followed by the US, Australia, Ireland and Spain. In terms of the number of AI bills passed into law, the US comes first with 23 AI bills passed in 2023.

Research

The Global AI Index measures AI research by analysing the volume, quality and novelty of specialist AI research activity, which includes:

• Activity and impact in AI research publications and academic conferences;
• The quality of educational institutions;
• The development of novel AI architectures and systems through building large-scale AI models.

The US and China, Singapore, the UK, Israel and Switzerland all have strong research ecosystems. The index differentiates between ‘foundational’ research – the development of fundamental algorithms and models within computer science and AI – and ‘applied’ AI research, which addresses the application of existing AI techniques to other fields. India, for example, has a relatively greater share of applied research, while France, Israel and Singapore have a relatively greater share of foundational research.

In terms of research on ground-breaking new large-scale AI models, the US and China have dominated for the past five years, thanks to easier access to capital and computing power as well as research excellence. The UK used to be a greater contributor (mostly through DeepMind) but is now lagging behind.

Development

AI development is measured by the Global AI Index through:
The training, development and publication of open-source large-scale AI models;
The application of existing AI technology across industries through patents.

The most powerful large-scale foundational models are still proprietary and entirely dominated by the US tech giants. However, open-source models have rapidly caught up in capabilities in the last few years, opening up the competition to non-American players.

France has rapidly established itself as the third global player in AI model development after the US and China, exemplified by its jump to 5th place in this year’s overall rankings. Mistral is the country’s national champion, having recently released models that compete with the most capable ones coming from the US and China.

This is the result of France’s extensive efforts towards developing non-english LLMs, with a strong ecosystem of generative AI startups. The UK’s performance in this particular area is comparatively poor, although the country still has an edge in ‘traditional AI’.

Commercial ecosystem

Commercial ventures are responsible for most implementations of artificial intelligence around the world. The Global AI Index analyses AI commercial activity by measuring:
- AI companies;
- AI private investment;
- Acquisitions of AI startups.

The data shows generative AI is making up an increasingly large share of AI investment globally, driving overall AI private investment growth over the past few years.

The US remains the undisputed leader in private AI investment – making up over 60 per cent of the total. Its success is driven on the one hand by the ‘magnificent 7’ tech giants that are heavily funding both AI startups (e.g. OpenAI and Anthropic) and their internal AI, and by technology VC funds (e.g. Andreessen Horowitz). However, there is not a clear business model for generative AI and experts are now raising concerns around investment returns and long-term sustainability.

The UK’s commercial ecosystem on generative AI is now being outfunded by the ecosystems of France and Germany.

Government strategy

Over the past year, governments’ commitment to artificial intelligence capacity has increased significantly, while global private funding has slowed down. Tortoise’s own analysis shows how this is reflected in national public funding allocated to AI. The Global AI Index measures countries’ government strategies for AI by analysing:

• National AI strategy documents;
• Government spending commitments.

Saudi Arabia tops the ranks on government strategy, mainly due to announced large spending commitments to be rolled out over the next decade. The US is second for publicly- declared AI government spending, followed by:

• South Korea, which has shown a clear strategy for AI funding;
• France, which currently outspends the UK by 60 per cent;
• Germany and Spain, which have both attached hefty budgets to their newly-updated AI strategies;
• and Canada, which tops the ranking on public investments for high-end AI computing infrastructure.

More countries have published a national AI strategy or are drafting one. Most recent AI strategies are considering AI ethics within it. And a quarter of the countries included in the index are backing the development of a foundational model.

1. Yasir Al-Rumayyan, governor of Saudi Arabia’s Public Investment Fund. PIF announced a new $100 million fund for semiconductors and related industries, and plans to create a $40 billion fund for investing in AI. 
2. Fei-Fei Li, professor and co-director of Stanford’s Human-Centered AI Institute. Known as ‘the godmother of AI’ who invented computer vision, this year she launched World Labs, an AI startup built in just four months and the latest to secure billions of investments following OpenAI’s release of ChatGPT.
3. Yann LeCun, head AI scientist at Meta. He is pushing for open-source AI development against the grain of most other US tech giants, which are working on proprietary models. 
4. Bhavish Aggarwal, founder of India’s first AI unicorn, OlaKrutrim. He started his career at Microsoft.
5. Arthur Mensch, CEO and co-founder of Mistral AI, France’s national AI champion and the only European generative AI startup to challenge its American and Chinese counterparts.
6. Faisal Al Bannai, secretary general of the UAE government’s Advanced Technology Research Council. Last year, the ATRC spent millions of dollars building Falcon, a series of open-source large language models which at the time beat offerings by Meta and Google.
7. Elizabeth Kelly, director at the United States Artificial Intelligence Safety Institute, which was unveiled last year at the world’s first AI Safety Summit in the UK. Kelly was one of the lead drafters of Biden’s AI executive order.
8. Dario Amodei, CEO of Anthropic. The company’s latest model – Claude 3.5 Sonnet – was by some measures the most powerful publicly accessible AI upon its launch.
9. Zhang Linghan, professor at the Institute of Data Rule of Law at China’s University of Political Science and Law. She is the first author listed on the Chinese draft preliminary AI law and sits on the UN High-Level Advisory Body on AI. 
10. Thierry Breton, former European commissioner for the internal market, tasked with facilitating the recent passage of the European Union’s Artificial Intelligence Act – the world’s first and most comprehensive AI law.
11. Geoffrey Irving, research director at the UK AI Safety Institute. He previously worked at Google DeepMind and OpenAI. 
12. Sasha Luccioni, AI and climate lead at Hugging Face, co-authored one of the first studies to calculate the carbon footprint of large language models.

Authors: Serena Cesareo, Joe White
Graphics: Joe White, Bex Sander, Katie Riley
Additional research: Conor O’Brien, Arabella Mensah
Edited by Alexi Mostrous.