COIL: The Physical Infrastructure Credit Layer

Introduction

Collateralized Onchain Infrastructure Lending (COIL), sometimes referred to as InfraFi, lowers the capital cost and operational friction for deploying physical infrastructure. Traditional project finance gates small and mid-size operators behind credit committees, legal counsel, and underwriting timelines that routinely stretch 3-6 months, a prohibitive barrier for hardware deployments that need to respond to demand in weeks. COIL replaces that pipeline with tokenized collateral, programmable loan terms, and onchain settlement, compressing deal timelines from quarters to weeks and opening lending markets to operators that banks consider too small or too risky. By enabling rapid, accessible capital formation for physical assets, COIL shifts supply curves outward, compresses lease rates for end-users, and fundamentally changes the competitive landscape for hardware providers.

The primary use case is yield generation for stablecoin capital that has exhausted onchain alternatives. Depositors seeking returns above the Aave lending benchmark, currently the de facto floor rate for passive stablecoin yield in DeFi, can deploy into COIL vaults and access infrastructure-backed spreads that onchain money markets cannot replicate. Because COIL yield derives from offchain cash flows, GPU rental revenue, solar PPA payments, and telecom subscription receipts, rather than recursive crypto borrowing demand, it is structurally uncorrelated to DeFi rate cycles and can exceed Aave benchmark rates during periods of compressed onchain liquidity.

Unit Economics

COIL's unit economics are uniquely positioned to serve the long tail of physical infrastructure, where traditional project finance economics break down. The overhead costs of structuring a limited-recourse project finance facility (independent engineering reviews, legal counsel, insurance advisors, and financial model diligence) make facilities below ~$50M uneconomic for most institutional lenders. The result is a financing gap that locks out the long tail of GPU operators, solar installers, and battery storage developers who need $500K-5M to deploy their next tranche of hardware. COIL fills that gap, but the model is only durable if utilization stays high and defaults stay low.

The scale of the unmet financing need at the institutional level validates the macro thesis. Morgan Stanley estimates that global data center construction will require $2.9T in capital expenditure between 2025 and 2028. Of that, only $1.4T can be covered by hyperscaler cash flows, leaving a $1.5T financing gap that must be met by external capital. Small and mid-size operators below that tier, the exact borrower profile COIL underwrites, do not exist in that capital stack.

The pricing of capital in adjacent AI infrastructure markets reinforces the viability of COIL’s borrower base at current rates. OpenAI is reportedly offering private equity partners preferred equity structures with a guaranteed minimum return of ~17.5% to fund deployment through joint ventures, materially above traditional preferred return thresholds. This validates the idea that downstream buyers of compute, enterprises integrating AI into core workflows, are willing to absorb capital costs that support mid-teens financing rates upstream. For COIL, this matters directly. If end-demand for compute can sustain ~15-20% return thresholds at the application layer, then 10-15% borrower APRs for GPU-backed loans are not structurally expensive, they are consistent with the clearing price of capital in the broader AI stack. High utilization converts these return profiles into self-funding debt, while idle capacity collapses the same structure regardless of how favorable the financing terms appear.