AI’s new battleground: Infrastructure, not algorithms

AI sovereignty is unrealistic; alone, nations must balance domestic control with trusted global partnerships.

A new World Economic Forum white paper, produced with Bain & Company, argues that AI infrastructure, not algorithms, has become the real battleground for national competitiveness. Titled “AI Infrastructure in the Age of Sovereignty,” the report lays out how nations should think about building, sharing, or borrowing the physical backbone that powers AI.

The three pillars of AI infrastructure

The report breaks AI infrastructure into three building blocks: compute (processing power), connectivity (networks linking data), and data storage. Each faces hard physical limits energy, water, land, hardware, and cybersecurity that determine what’s actually buildable. Global data center electricity use alone could nearly triple by 2035, and a single AI training facility can require 200 acres of land and millions of gallons of water daily.

No country can go it alone

Perhaps the report’s central message is that true self-sufficiency in AI infrastructure is unrealistic for almost everyone. Even China and the US, the only economies with near-full domestic control, still rely on concentrated global supply chains for chips. Most nations, the report says, must choose a strategy somewhere between two extremes:

  • Trusted international partnerships — leaning on foreign partners for compute and storage while keeping legal control (think Estonia storing government data in Luxembourg).
  • Extensive domestic ownership — building and controlling everything at home, best suited for well-resourced economies or sensitive workloads like defense.

In practice, most countries will blend both, as Singapore does by combining a domestic supercomputing center with heavy use of international cloud providers.

Enter “Digital Embassies”

The report’s most novel contribution is a Global Framework for Innovative and Trusted Digital Embassies: secure, extraterritorial digital spaces in which a country’s data and AI workloads are hosted abroad yet remain governed by its own laws. Pioneered by Estonia’s 2017 data-hosting deal with Luxembourg, this model is now gaining traction, with Saudi Arabia and Bahrain drafting similar legal structures.

For this to work, the Forum argues, five things need to be trusted: political commitment, legal clarity, data governance rules, technical safeguards, and transparent operations. Without them, digital embassies risk becoming legal grey zones rather than genuine extensions of sovereignty.

The bottom line

The report’s call to action is blunt: governments, investors, and infrastructure operators must treat energy and land constraints as design realities, not afterthoughts, and build partnerships with enforceable safeguards rather than blind trust. As AI infrastructure decisions today will shape economic resilience for decades, the paper positions digital embassies and disciplined, hybrid strategy-making generally as key tools for countries seeking sovereignty without isolation.

Author