Four Forces of AI Power

The Four Forces of AI Power is a framework identifying the four contested layers that determine durable advantage in the AI era. Each force operates as a separate axis of scarcity, control, and economic capture. An entity can win one force, lose another, and still be misread by markets that treat AI as a single category rather than a layered system.

The four forces are Compute, Interface, Alignment, and Energy. They are not stages of an AI value chain. They are parallel contested layers, each with its own incumbents, chokepoints, and substitution dynamics.

Compute is the layer of model training, inference capacity, and the physical infrastructure that produces cognition. Power in this layer accrues to entities that own scarce silicon, foundation model access, or fine-tuning capability at scale. Compute is the most visible force because it is the most easily measured — but visibility and durability are not the same. Compute scarcity moves quickly when supply expands, when efficiency improves, or when substitution emerges from a competing architecture.

Interface is the layer of user-facing surfaces, agent endpoints, and the access points through which AI reaches economic activity. Power in this layer accrues to entities that own the surfaces where AI gets used — browsers, app stores, operating systems, search bars, agent frameworks, and emerging native AI interfaces. Interface power compounds because changing where an interaction begins is harder than changing what runs underneath it.

Alignment is the layer of safety, policy, regulatory compliance, and the gatekeeping authority over what models are permitted to do. Power in this layer accrues to entities that set, enforce, or pass through alignment requirements — frontier labs setting internal policy, regulators setting external policy, and standards bodies setting interoperability rules. Alignment increasingly functions as a permissioning layer: capability without alignment access is not deployable at scale.

Energy is the layer of power supply, grid capacity, cooling, and the physical resources that make Compute possible. Power in this layer accrues to entities that own generation, transmission rights, or long-dated power purchase agreements. Energy is the slowest-moving force because the infrastructure cycles are decadal rather than quarterly, which makes energy positioning harder to reverse once established.

Three principles govern how the forces interact:

The first is that the forces are independently contested. Winning Compute does not imply winning Interface. Winning Interface does not imply winning Alignment. Each force has its own incumbents, its own challengers, and its own substitution risk. Strategic clarity requires reading each force on its own axis.

The second is that the binding force shifts. At any given moment, one of the four is the constraint that determines the pace of AI progress. Compute was binding through the early scaling era. Energy is increasingly binding now. Interface will become binding as agentic deployment matures. Alignment binds intermittently, particularly during regulatory inflection moments. The framework's analytical value comes from identifying which force is currently constraining and which entities are positioned across the constraint.

The third is that the forces compose into entity-level positioning. An entity's strategic posture is the vector across all four forces, not a single-force position. Reading any AI participant requires reading their Compute exposure, their Interface position, their Alignment relationship, and their Energy footing simultaneously.

The Four Forces sits underneath the doctrine layer that connects to sPEG (Scarcity-Adjusted PEG) at the valuation layer, ADS (AI Deployment Signal) at the operating-state layer, and TCM (Tokenized Cognition Model) at the unit-economics layer. Together they form a stack: the Four Forces describes what is contested, sPEG prices the scarcity, ADS reads who is actually deploying, and TCM measures whether the deployment is economically viable per token.

The framework is not predictive about which entities will win each force. It is structural about which entities are even playing on each axis — and which ones are exposed to a force they have not built against.