We introduce Inference Economics as a new subfield of economics organized around the production, pricing, and consumption of AI inference tokens — the fundamental digital commodity that provides access to artificial intelligence capabilities. As frontier AI models from oligopolistic providers (OpenAI, Anthropic, Google, xAI) become as foundational to production as electricity was in the early twentieth century, a new economics is required to understand how tokens are produced, priced, consumed, and allocated across a diverse population of developers and organizations. Drawing on a nine-paper foundational series (Brass 2026a–i), we develop the conceptual architecture of Inference Economics around four intellectual pillars (production economics, platform economics, digital infrastructure markets, and AI innovation systems) and five core theoretical constructs (the Token Production Function, the Token Kuznets Curve, the Jevons Paradox of AI Tokens, the General Equilibrium of the Token Economy, and the Copula Two-Part Demand Model). We locate these constructs in the broader economics literature, derive their key empirical implications, and set out a seven-direction research agenda. Inference Economics is not a relabeling of existing subfields; it is a synthesis with distinctive theoretical primitives — cognitive capital, agentic multipliers, token intensity curves, and inference gap inequality — that do not exist in any single prior tradition. The empirical calibration from the foundational series yields tractable, testable predictions about AI market structure, token demand dynamics, welfare distribution, and long-run growth that constitute the positive theory of this new subfield.