- AI dominance depends on control over compute, chips, data and talent infrastructure
- The US leads global AI compute capacity with 75%, followed by China and the EU
- Mid-size countries like UAE, Malaysia, and India emerge as key AI infrastructure hubs
Artificial intelligence has become the central battleground of global power. Governments are now competing not just in algorithms but across four strategic assets: Compute, chips, data and talent. The result is an emerging geopolitical map often described as the "AI Cold War" with the United States, China and Europe building competing ecosystems while middle powers quietly positioning themselves as infrastructure hubs.
Yet the most important insight from recent data is this: AI dominance is not determined by models alone. It is determined by control over infrastructure.
The Four Levers of AI Power
1. Compute: The Ultimate Bottleneck
AI progress now depends on access to massive computing clusters.
The United States overwhelmingly dominates this layer, hosting about 75 per cent of global AI computing capacity, according to an analysis of GPU clusters by Epoch AI.
Next comes China with roughly 14-15 per cent of global compute power, while the EU controls under five per cent.
US data centres account for about 60-75 per cent of global computational power used for AI workloads, the report further shows.
This concentration reflects the dominance of American hyperscale cloud companies such as Amazon Web Services, Microsoft and Google.
The scale of modern AI infrastructure has exploded. A single frontier AI supercomputer can use 200,000 AI chips and cost more than $7 billion, according to a 2025 analysis of global AI systems.
The power requirements are also staggering: one cluster can require 300 megawatts of electricity, equivalent to powering 250,000 homes. That energy footprint means AI power is also becoming an energy geopolitics story.
2. Chips: The Strategic Choke Point
If compute is the engine, chips are the fuel. The semiconductor supply chain is dominated by companies aligned with the United States:
- NVIDIA
- AMD
- Intel
These companies supply the majority of data-centre GPUs used for training AI models globally.
To slow Chinese progress, Washington imposed export restrictions on advanced chips beginning in 2022 and strengthened them in 2023-2024.
But Chinese companies are already adapting.
For example, ByteDance is accessing 36,000 advanced GPUs in Malaysia through a cloud partner, allowing it to legally use top-tier hardware despite US export limits, a report in Wall Street Journal (WSJ) said.
The episode illustrates a key reality of the "AI Cold War" - that hardware restrictions are increasingly being bypassed through global cloud infrastructure.
3. Data: The Invisible Advantage
AI systems learn from enormous datasets. China arguably holds a structural advantage here due to scale: It has over 1 billion internet users, and massive digital platforms including Alibaba and Tencent.
However, Western companies dominate global datasets and enterprise data pipelines, giving US firms a strong position in training commercial AI systems.
Europe's strategy focuses on "data sovereignty", ensuring European data is processed within the region rather than on foreign clouds.
For example, European companies are now investing heavily in domestic AI data centres to reduce reliance on US cloud providers.
4. Talent: The Global AI Workforce
Talent is the one area where the map looks very different. According to global AI competitiveness rankings, here's how countries stack up:
1. United States
2. China
3. India
4. South Korea
5. United Kingdom
India's rise has been particularly striking, with explosive growth in AI hiring and open-source contributions. This makes India one of the most important talent suppliers in the AI ecosystem, even though its infrastructure capacity remains smaller than the US or China.
The Emerging AI Blocs
The competition is increasingly forming three broad geopolitical ecosystems.
1. The American AI Alliance
As the name suggests, it is led by the United States and built around cloud platforms with following key features:
- Dominance in chips and cloud computing
- Venture capital and startup ecosystem
- Leading AI labs including OpenAI and Anthropic
The US also uses export controls and security alliances to shape global AI supply chains.
2. The Chinese AI Sphere
China's model relies on state-backed technology ecosystems with particular focus on:
- AI research scale
- Massive domestic market
- State-driven infrastructure
China now produces about 36% of global AI research publications, the highest share in the world. This is a dramatic rise from less than five per cent in the year 2000.
However, its biggest vulnerability remains dependence on foreign chips.
3. Europe's "Regulatory Superpower"
Europe is attempting a third path. Instead of dominating compute or chips, the EU is focusing on:
- AI regulation
- Privacy rules (GDPR)
- Ethical frameworks
- Digital sovereignty
Europe's share of global AI patents remains smaller than that of the US or China, but their patents tend to have higher citation quality, suggesting strong research depth.
The Unexpected Winners
The AI Cold War has also created surprising winners. Several mid-size countries are becoming AI infrastructure hubs.
Gulf States: Countries like United Arab Emirates and Saudi Arabia rank among the top nations globally in AI compute capacity due to massive data-centre investments. Cheap energy and sovereign wealth funds allow them to build giant GPU clusters.
Southeast Asia: Countries like Malaysia and Indonesia are emerging as neutral AI infrastructure zones, hosting data centres used by both Western and Chinese companies.
India: India's advantage lies in talent rather than hardware. With one of the world's fastest-growing AI developer communities, it is becoming a global engineering backbone for AI development.
Why the AI Cold War Is Different
Unlike the nuclear Cold War, this technological rivalry is not purely national. Much of the infrastructure is owned by multinational corporations rather than governments.
Cloud companies, chip manufacturers, and data-centre operators now act as quasi-geopolitical actors.
The Real Question: Who Controls the Infrastructure?
The AI Cold War may ultimately hinge on a single factor: Who controls the physical infrastructure that trains and runs AI systems.
Right now, the United States leads decisively in compute and chips. China leads in scale of research and data while Europe is attempting to lead in governance.
But the next phase of the race may be decided elsewhere - in data-centre corridors in the Middle East, Southeast Asia and India because in the age of artificial intelligence, the most powerful country may not be the one with the smartest algorithm.
It may be the one that owns the machines that run them.
