FRAMEWORK

STATE EXCLUSION
AND THE INFERENCE ERA

Compute sovereignty has been mis-specified.

The dominant frame is founded in the training era and asks who can build frontier models. For almost every state outside the US and China, the answer is: you can't. Amidst an emerging geoeconomics, export controls, chip restrictions, compute thresholds all follow from that assumption.

Even as controls increasingly extend toward cloud access and API provision, the underlying question about where sovereignty is at stake remains training era. This architecture implicitly excludes most of the world - they are subjects of controls, not designers of them.

This has resulted in governance frameworks being built around the interests of the states already at the center, in a landscape defined by weaponised interdependence and fragile geopolitics of great power concentration. States that bear the highest costs and face highest barriers to entry in the current AI infrastructure architecture are systematically excluded from its design. When frameworks don't represent the full range of actors who are impacted by and depend on the infrastructure, we risk failing structurally.

MIDDLE POWER VULNERABILITY
AND LEVERAGE

The inference era governance asks: who controls deployment infrastructure?

Compute power is characterized by compounding chokepoints—API dependency, model licensing terms, deployment infrastructure, physical infrastructure, critical minerals, semiconductor production. These concentrate power and foreclose sovereign options before governance frameworks can respond.

For many states, the practical choice is not between dependency and sovereignty, it is between US or China hosted endpoints actively competing for the same middle power infrastructure contracts.

Still - middle powers have material and labour leverage that the training-era frame made analytically invisible: mineral positions, energy endowments, cable routing geography and landing station control, investment review capacity. Concentration of power is the structural condition that creates both vulnerability and leverage for actors on the edge

IIL emboldens middle powers to understand their positions and build agency within these dynamics.

By developing inference-era frameworks, IIL aims to redistribute analytical and governance capacity in compute infrastructure from Washington and Beijing, to middle powers, and ultimately a more resilient and coherent global compute economy.

P4857/Q34

USD 199.99