Kuma Publishing

AI Infrastructure Valuation Research

Kuma Publishing is an independent research practice studying how AI compute assets are valued — secondary-market pricing, economic obsolescence, and lifecycle value — applied to property tax, lending, insurance, and M&A diligence.

Working Paper 2026-01

The Depreciation Gap

Economic Obsolescence of 2023–2024 Vintage AI Compute Hardware for Ad Valorem Property Tax Purposes

County assessors value business equipment on depreciation schedules built for machines that wear out slowly. AI compute doesn't wear out — it gets displaced by the next architecture, and its value reprices to whatever the market will pay for its remaining use. This paper measures the gap between schedule value and market value for a representative H100 fleet, and finds the real assessment risk lives not where most assume. It introduces the Compute Depreciation Curve framework.

Read the Paper (PDF)

Research Focus

  • Property-tax valuation & appeals evidence
  • GPU / accelerator residual value
  • Lender & insurer collateral valuation
  • Acquisition & impairment diligence

About the Author

Michael Halper is the founder of Kuma Publishing's AI Infrastructure Valuation Research practice. He applies institutional valuation methodology — discounted cash flow analysis, market-evidence reconciliation, and a published intrinsic-value framework — to AI compute assets. USPAP coursework is in progress.

Contact

Research inquiries and engagement questions welcome.

info@kumapublishing.com