AI Sovereignty: A Qualitative Model of Strategic Competition as AI Becomes an Instrument of National Power
This paper addresses the lack of conceptual frameworks for AI sovereignty by introducing a novel qualitative model that analyzes micro, meso, and macro factors to identify strategic leverage points—ranging from resource control to kinetic and non-kinetic actions—through which nations can compete for AI-driven national power in the 21st century.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine that Artificial Intelligence (AI) is no longer just a cool tool on your phone; it has become the new "air force" of the 21st century. Just as nations once fought to build the best airplanes and aircraft carriers to protect their borders and project power, countries are now racing to build the best AI systems.
This paper, written by researchers from the University of Maryland and Sandia National Laboratories, proposes a new way to measure who is winning this race. They call it AI Sovereignty.
Here is the breakdown of their ideas, using simple analogies:
1. What is "AI Sovereignty"?
Think of AI Sovereignty as owning your own kitchen versus ordering takeout.
- High Sovereignty: A country that can grow its own vegetables, build its own stove, hire its own chefs, and cook its own meals. They control the whole process.
- Low Sovereignty: A country that relies on a neighbor to bring them food. If the neighbor decides to stop delivering, or if the road gets blocked, that country goes hungry.
The paper argues that nations want "AI Sovereignty" because it makes them stronger economically and militarily. If you rely on another country for your AI, you are vulnerable.
2. The "Agentic AI" Shift
The authors say we have moved past simple "Generative AI" (like a chatbot that writes a poem when you ask) to something they call "Agentic AI."
- The Analogy: Imagine the difference between a calculator and a construction crew.
- A calculator just does math when you tell it to.
- A construction crew (Agentic AI) can look at a blueprint, decide what tools to grab, hire the workers, and build a house on its own.
- The Catch: This "construction crew" is incredibly expensive to feed. It needs massive amounts of electricity, water for cooling, and huge physical spaces to operate. It's not just software anymore; it's heavy, physical infrastructure.
3. The Three Layers of the Race (Micro, Meso, Macro)
To understand how a country builds this power, the authors break it down into three sizes, like a set of nesting dolls:
- Micro (The Cabinet): This is the smallest unit. Think of it as a single server rack (a big metal box full of computer chips). These boxes get very hot and need special liquid cooling, like a car radiator. They need a lot of electricity and water.
- Meso (The Data Center): This is a warehouse full of those server racks. It's a "factory" for AI. The paper notes that old factories might not work for this new type of AI; they need to be built or upgraded specifically for high heat and high power.
- Macro (The Nation): This is the whole country's total power. It's the sum of all the factories, the total number of workers, and the total electricity available.
4. The Bottlenecks (The Levers)
The paper identifies specific "levers" that countries can pull to win or lose. These are the things that limit how fast you can grow your AI army:
- Electricity: AI eats power like a black hole.
- Water: The servers get so hot they need liquid cooling. If you run out of water, the machines overheat and stop.
- Skilled Workers: You need smart engineers to build and fix the machines.
- Data: You need massive libraries of information to teach the AI.
The "Growth Trap": The paper uses a concept called "growth and underinvestment." Imagine you buy a new car (AI capability). It goes fast, but then you realize you don't have enough gas (electricity) or a mechanic (workers) to keep it running. You stop growing until you fix the gas supply. But by the time you fix it, your rival has already passed you.
5. The New Kind of War
The most striking part of the paper is how it describes the risks.
- Old War: In the past, if you wanted to stop an enemy's air force, you bombed their airfields.
- New War: Today, an enemy's "airfield" is a data center.
- The paper cites a hypothetical scenario (based on their model) where a country uses drones to attack a data center in the Middle East.
- Because AI is so physical now, these data centers are "targetable." You can shoot at them, hack them, or cut their water supply.
- It's not just about code anymore; it's about physical infrastructure.
6. How Do We Measure the Winner?
The authors suggest a scoreboard to see who has the most "National Power" in AI. Instead of just counting how many AI models exist, they propose looking at:
- Total Power: How much raw computing power (zettaFLOPS) does the country have?
- Pacing: Can they build the newest generation of AI, or are they stuck with old models?
- Sovereignty Score: What percentage of that power is built inside their own borders vs. borrowed from others?
The Bottom Line
The paper concludes that the race for AI is becoming a race for physical resources (power, water, land, and people). Nations that can control these resources will have the most power. Nations that rely on others risk being cut off.
This isn't just about technology; it's about national security. The authors warn that we are moving toward a world where "AI Sovereignty" is as critical as having your own oil or your own army, and the competition to control it could lead to a new kind of strategic conflict.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.