Beyond the US and China: What does it take to win the AI race?

The US and China are the world’s leading AI superpowers. But which countries come next?

First, some perspective: the gap between them and everyone else remains substantial. They have far more investment, computing power, leading companies and research capacity, and are competing at the frontier.

Even they are not evenly matched. Stanford’s 2026 AI Index records 59 notable AI models from US institutions in 2025, compared with 35 from China. The US leads in frontier models and higher-impact patents, while China leads in research publications, citations and total AI patents. Yet the gap between their leading models has largely disappeared.

Below them, rankings become much harder. The UK, Singapore, France, South Korea, Israel, Canada and others feature highly, but there is no single measure of AI strength. A country may have exceptional researchers but insufficient computing power; manufacture critical chips without producing leading models; attract investment but struggle to build global companies; or have limited AI research but become a major market for adoption.

So rather than asking who is number three after the US and China, it’s more useful to ask what different countries bring to the AI economy: talent, capital, chips, energy, infrastructure, companies, or the ability to deploy AI at scale.

AI is becoming an infrastructure race

The first AI boom was largely about people: researchers, engineers and investors. Now another constraint is becoming critical: AI requires enormous computing power and therefore chips, data centres, electricity and networks.

Stanford estimates that global AI computing capacity has grown threefold annually since 2022, reaching the equivalent of 17.1 million H100 GPUs (a widely used benchmark for computing power). The US has 5,427 data centres, more than ten times as many as any other country, while Taiwan sits at the heart of the chip supply chain: TSMC (Taiwan Semiconductor Manufacturing Company) manufactures almost all of the world’s leading AI chips.

Investment is following. The OECD estimates that AI attracted $258.7bn in VC in 2025, 61% of global VC investment. More than $109bn went into IT infrastructure and hosting – over twice the amount invested in generative-AI companies.

An AI economy therefore depends on electricity, chips, data centres and networks, capital, technical talent, and businesses and consumers ready to use AI.

This broadens the picture. France has nuclear power; South Korea is a semiconductor powerhouse; the UAE combines capital, energy and political will to build data centres rapidly; Britain has strong AI talent and investment but faces greater constraints on power and infrastructure. The AI race is increasingly a contest over energy, infrastructure, capital and the ability to build at speed.

Britain has the talent but needs the infrastructure

Britain has a strong claim to being the leading AI ecosystem outside the US and China. Its universities have produced generations of leading researchers, DeepMind put it at the centre of frontier AI, and London combines major technology companies with deep pools of international finance.

The investment numbers are significant. According to the OECD, British AI companies attracted $13.8bn in VC investment in 2025, around 5% of the global total, while British investors accounted for 9% of global outgoing AI VC, second only to the US.

The challenge is physical scale. The UK government estimates it will need at least 6GW of AI-capable data-centre capacity by 2030, three times today’s level, including several sites of at least 500MW and an AI Growth Zone exceeding 1GW.

That exposes a structural weakness: Britain’s technology economy has grown around mobile assets, such as talent, IP, capital and services, whereas AI increasingly requires land, power, grid infrastructure and planning capacity. Its intellectual capacity can therefore grow faster than its physical infrastructure.

There is a similar challenge with commercial scale. Britain has created highly valuable technology companies but has struggled to turn them into global giants while retaining ownership and control. DeepMind’s acquisition by Google enabled enormous growth but also illustrates the gap between Britain’s ecosystem and America’s vastly deeper pools of capital and corporate scale.

Singapore is showing that you don’t need to build the next OpenAI

Singapore is an intriguing counterpoint, having achieved an extraordinary AI position without the population, industrial base or research capacity of larger contenders.

Some indices rank it third globally, ahead of the UK, with particular strengths in talent, research and investment. Its AI adoption is even more striking: Stanford estimates that 61% of Singapore’s population has adopted generative AI, compared with 28.3% in the US and 64% in the UAE.

Singapore therefore offers a different model of AI competitiveness. It doesn’t need to produce the next OpenAI; its advantage is providing an environment where international companies can locate, invest, experiment and deploy across Asian markets.

This builds on Singapore’s established role as a global financial and commercial hub. AI fits that model particularly well as adoption spreads across financial services, logistics, manufacturing and professional services. A country can become important to the AI economy without becoming an AI giant itself.

France has something AI increasingly needs: Power

France’s proposition is more industrial. At the 2025 Paris AI summit, President Emmanuel Macron announced around €109bn in French AI investment commitments. At the 2026 Choose France summit, SoftBank announced plans to invest €45bn in three data centres with combined capacity of 3.1GW, potentially rising to €75bn. Macron has also positioned France’s nuclear power as an advantage in the competition for AI infrastructure.

That matters because data centres consume enormous amounts of electricity, making reliable, relatively low-carbon power increasingly valuable. France also has Mistral, giving it something few European countries possess: a prominent domestic frontier-AI company.

The challenge is turning these advantages into a durable ecosystem. France has technical talent, industrial capability, nuclear power and an assertive government strategy, but a smaller venture capital market than the US and a more complex regulatory environment. Data centre expansion also faces constraints around land, infrastructure and public opposition.

France may therefore have one of Europe’s strongest combinations of energy, industrial policy and AI capability. The question is whether it can build quickly enough and capture enough of the resulting economic value.

Israel has the talent. Can it keep the value?

Israel’s strength is more concentrated. Its technology ecosystem brings together universities, defence research, cybersecurity expertise, venture capital and an unusually strong entrepreneurial culture. It is also one of the world’s most AI-intensive investment markets.

The OECD says more than half of Israeli VC investment in 2025 went to AI firms, while AI investment relative to GDP has historically been among the world’s highest. Much of this capital flows into cybersecurity and core infrastructure, but also into defence technology, battlefield testing and dual-use technologies with both civilian and military applications. The structural question is what happens when Israeli companies outgrow the country’s relatively small domestic market.

The biggest pools of global late-stage capital and the largest customer base remain disproportionately American. That creates a distinction between producing AI companies and building an AI industry. Israel is exceptionally good at the first. Whether it can capture a correspondingly large share of the value created when those companies scale is less certain.

India’s advantage is scale

India presents a different proposition. It is far behind the US in frontier models and computing infrastructure but has a vast technical workforce and a domestic market few countries can match. Stanford’s 2026 research found that more than 80% of surveyed Indian workers use AI regularly or semi-regularly at work.

That gives India a potentially important role downstream from the frontier. If AI becomes a general purpose technology across software, finance, manufacturing, healthcare and government, India’s scale could make it one of the world’s most important markets for AI adoption, particularly where technology can deliver services more cheaply and at greater scale.

The risk is that India becomes a vast market for technology, IP and infrastructure owned elsewhere. Its long-term position will depend on turning its human capital and domestic demand into AI companies, intellectual property and infrastructure of its own.

South Korea owns part of the machine

South Korea shows why looking only at AI models can distort the picture. Stanford identifies it as the global leader in AI patents per capita. It is also home to Samsung and SK Hynix, two dominant memory-chip manufacturers, and a major producer of high-bandwidth memory for advanced AI systems.

This matters because the AI economy has a long supply chain. A country doesn’t need to build the most capable model to occupy a critical position within it. South Korea’s position is therefore closer to Taiwan’s than Britain’s or France’s.

AI leadership increasingly means manufacturing the components on which AI leadership elsewhere depends.

The Gulf is testing how quickly an AI ecosystem can be built

The UAE is conducting one of the world’s most ambitious AI experiments. It has capital, energy, land and a government able to make major infrastructure decisions quickly. Its generative-AI adoption rate is among the world’s highest, while its strategy centres on attracting international AI companies and building domestic computing capacity.

Stargate UAE is the clearest example.  The Stargate Project is a massive, multi-billion-dollar US initiative launched to build the physical data centres, supercomputers, and energy networks required to power next-generation artificial intelligence. OpenAI announced a 1GW Stargate UAE cluster in Abu Dhabi, with the first 200MW targeted for this year. Construction is under way, but the first phase has yet to come online. Saudi Arabia is pursuing a similar strategy on a larger scale, including its $100bn Project Transcendence initiative, which aims to build a global AI powerhouse and tech hub.  

The question is whether capital and infrastructure can compensate for the slower development of technical institutions, experienced founders and research communities. The Gulf is effectively testing how much of an AI ecosystem money and political will can build and whether that’s enough to make the region a major AI hub.

There’s more than one way to win the AI race

The AI map is more complicated than a simple league table suggests. The US and China dominate the frontier, but other countries are building different advantages: Britain has talent and finance; Singapore, connectivity and adoption; France, energy and industrial capacity; Israel, technical entrepreneurship; India, scale; South Korea and Taiwan, critical semiconductor expertise; and the Gulf, capital and infrastructure.

No single advantage is enough. The countries best placed to benefit will be those that can combine several or find an essential role in the wider AI ecosystem.

What does AI mean for your international growth strategy? Contact Oban to continue the conversation.