How many nations can truly build and control their own artificial intelligence infrastructure from the ground up? The answer reveals a stark reality: only 32 countries host AI-specific data centers, leaving 160 nations entirely dependent on foreign infrastructure. The United States and China control over 90% of global AI data-center capacity, creating a significant power imbalance that shapes geopolitical relationships and national security strategies.
Only 32 nations host AI-specific data centers while 160 countries remain entirely dependent on foreign infrastructure controlled by the US and China.
Training frontier AI models demands billions in compute resources, engineering talent, hyperscale data centers, and cutting-edge semiconductors. No country achieves complete self-sufficiency in AI development and deployment. This reality drives nations toward what experts call “sovereign AI”—prioritizing domestic development using nationally governed data for business and strategic advantages. Many governments are exploring cloud-native architectures to reduce reliance on foreign providers and simplify integration across services.
The power requirements tell a compelling story. AI training workloads consumed 5 gigawatts of data center capacity last year, while inference used 2 gigawatts. By 2030, training will require 23 gigawatts and inference 54 gigawatts. AI will account for 50-70% of total data center computing by decade’s end, straining electrical grids worldwide.
US-China tensions accelerate this competition through tariffs and export controls on advanced chips and AI technologies. These restrictions aim to maintain development leads but spur innovation in response. China has developed efficient models like DeepSeek and hardware alternatives like Huawei Ascend. Europe reduces exposure to Chinese hardware over surveillance and disruption concerns.
European responses include significant investment. The EU allocated £11 billion in its 2026 budget for AI research and the Digital Europe programme. Proposed Cloud and AI frameworks simplify data-center construction and improve interoperability. Leaders invoke an “Airbus Moment” for European AI infrastructure to counter American dominance.
Distributed architectures offer practical alternatives. They enable fine-tuning and inference on local sensitive data, reducing unilateral dependencies while applying domestic security and privacy standards. Countries can specialize by their advantages: renewable power, manufacturing data, healthcare expertise, or regulatory frameworks. This approach fosters ecosystems for domestic contributions in energy-efficient centers, model tuning, and edge-AI deployment, balancing competitiveness with sovereignty as complementary rather than competing objectives. Organizations without AI resources risk falling behind in technology competitiveness, productivity, customer engagement, and decision-making. Effective governance requires states to cultivate institutional capability to evaluate and adapt AI systems at technological speed.
