The Compute Bloc: How Semiconductors Redrew the Map of Alliances
Access to advanced compute has displaced access to energy as the organising principle of twenty-first century coalitions — with consequences for states that produce neither.
A familiar structure with new inputs
For most of the twentieth century, the map of alliances tracked the map of energy. Who held reserves, who controlled transit routes, and who could be cut off shaped alignment more reliably than ideology did.
The same structural logic now applies to compute. Advanced semiconductors are produced in a small number of places, depend on an even smaller number of equipment suppliers, and can be denied. That combination — concentration, chokepoints and deniability — is what turns an input into an instrument of statecraft.
Three tiers, not two
The resulting structure is often described as bipolar. It is more accurately a three-tier arrangement.
The first tier holds design, manufacturing or equipment capability at the frontier. The second holds meaningful capacity in adjacent layers — packaging, materials, specialised chemicals, or the electrical and construction capacity to host large facilities. The third holds none of these and imports both the hardware and the models that run on it.
The middle tier is where the interesting statecraft is happening. Its members have genuine leverage but not independence, which gives them room to negotiate rather than merely to align. Several have used it well.
What the third tier can actually do
States without semiconductor capability are not without options, but the options are different in kind. Three are available to almost any state with functioning institutions.
The first is to become indispensable in an adjacent input — power generation, cooling, land, or a specific mineral. The second is to build regulatory and data-governance capability strong enough that market access becomes a negotiating asset. The third, and most neglected, is to develop the human capability to adapt and evaluate models, which determines whether imported technology produces domestic value or merely domestic dependency.
None of these produce sovereignty in the frontier sense. All of them produce standing, and standing is what determines the terms on which a state participates.
The risk of over-reading
One caution is worth stating. Compute concentration is real, but the assumption that it maps cleanly onto durable geopolitical alignment is doing a lot of unexamined work. Supply chains reroute. Second-source capacity appears where margins justify it. Denial regimes tend to accelerate the substitution they were designed to prevent.
The prudent posture is to treat compute access as a genuine constraint over a five-year horizon and a much softer one over a fifteen-year horizon — and to build accordingly.