The Economics of Digital Intelligence Capital: Endogenous Depreciation and the Structural Jevons Paradox

TL;DR

A micro-founded economic theory of the AI industry is developed by modeling large language models as a distinct asset class-Digital Intelligence Capital-characterized by data-compute complementarities, increasing returns to scale, and relative (rather than absolute) valuation.

Abstract

This paper develops a micro-founded economic theory of the AI industry by modeling large language models as a distinct asset class-Digital Intelligence Capital-characterized by data-compute complementarities, increasing returns to scale, and relative (rather than absolute) valuation. We show that these features fundamentally reshape industry dynamics along three dimensions. First, because downstream demand depends on relative capability, innovation by one firm endogenously depreciates the economic value of rivals' existing capital, generating a persistent innovation pressure we term the Red Queen Effect. Second, falling inference prices induce downstream firms to adopt more compute-intensive agent architectures, rendering aggregate demand for compute super-elastic and producing a structural Jevons paradox. Third, learning from user feedback creates a data flywheel that can destabilize symmetric competition: when data accumulation outpaces data decay, the market bifurcates endogenously toward a winner-takes-all equilibrium. We further characterize conditions under which expanding upstream capabilities erode downstream application value (the Wrapper Trap). A calibrated agent-based model confirms these mechanisms and their quantitative implications. Together, the results provide a unified framework linking intelligence production upstream with agentic demand downstream, offering new insights into competition, scalability, and regulation in the AI economy.

Figures (18)

• Figure 1: Industry structure and feedback mechanisms of digital intelligence capital. The figure illustrates the vertical structure of the AI production stack and the three core mechanisms analyzed in the paper. Upstream, foundation model providers invest in digital intelligence capital $K_{AI}$, characterized by data--compute complementarity, increasing returns to scale, and endogenous economic depreciation. AI services are intermediated through an API market with capacity constraints and oligopolistic markups. Downstream, agent developers optimally choose architecture complexity (e.g., reasoning depth, tool integration), which amplifies token usage via a token multiplier. Falling API prices induce more compute-intensive architectures, generating super-elastic demand (Jevons paradox). Aggregate usage produces proprietary feedback data, which feeds back into upstream capability accumulation. When data accumulation dominates data decay, the feedback loop destabilizes symmetric competition and leads to winner-takes-all outcomes.
• Figure 2: The structural Jevons paradox: mechanism decomposition. Panel A shows the downstream architectural response to declining API prices: as inference prices fall, firms optimally increase architectural complexity (e.g., reasoning depth and agentic loops), with a regime shift from simple prompting to agentic workflows. Panel B maps architectural complexity to token intensity, illustrating how more complex architectures amplify token usage through a token multiplier. Panel C combines these effects to show the resulting aggregate outcome: despite declining unit prices, total expenditure on AI services increases because the architecture-induced increase in token intensity dominates the standard scale effect. Together, the three panels demonstrate that the Jevons paradox in the AI industry is structural, arising from endogenous architectural choice rather than exogenous demand elasticity.
• Figure 3: Endogenous economic depreciation and the relative valuation of digital intelligence capital. The figure illustrates how the shadow value of digital intelligence capital $q_i$ depends on a firm’s capability relative to the market leader. In the absence of competition, the value of capital remains constant, as in traditional physical capital. With competition, however, valuation becomes inherently relative: when models are weakly substitutable (low $\theta$), the shadow price declines gradually with relative inferiority, whereas under strong substitutability (high $\theta$), small capability gaps lead to sharp value collapse. The shaded region highlights an economic obsolescence zone in which capital loses value despite unchanged physical capacity. The inset panel shows the implied endogenous depreciation rate $\delta_i$, which increases as relative capability deteriorates, capturing the Red Queen effect driven by competitive pressure rather than physical decay.
• Figure 4: Panel A: AI Capital Stock ($K_{AI}$)
• Figure 5: Panel B: Market Share ($s_i$)

Theorems & Definitions (12)

• Definition 4.1: Digital Intelligence Capital
• Definition 5.1: Static API Market Equilibrium
• Proposition 6.1: Negative Pecuniary Externality
• proof : Proof Sketch
• Theorem 6.1: The Red Queen Effect
• proof
• Proposition 8.1: Jevons Paradox in AI
• proof
• Proposition 9.1: Flywheel Stability Threshold
• proof