Multicoin Proposes AI Chip Collateral for Crypto Lending Expansion

Multicoin Capital partner Shayon Sengupta proposed allowing AI companies to secure stablecoin borrowing with GPUs and future computing revenue. Sengupta's Oct. 8 research note outlined blockchain markets for trading or pledging ownership claims on hardware, future data center capacity and customer payments. Accelerating AI infrastructure borrowing creates an opening for decentralized finance (DeFi) protocols to fund companies facing expensive equipment purchases.

The Kobeissi Letter reported record AI-linked US dollar corporate bond issuance of $419 billion in 2026 through Oct. 21. That amount exceeded the entire 2025 total by 90% and the 2024 total by 437%. Alphabet, Amazon, Meta, Microsoft and Oracle together accounted for $157 billion of the borrowing. Investment-grade and high-yield data center and computing companies borrowed another $103 billion. However, many smaller AI infrastructure operators have customer contracts and equipment orders without the financial strength or lender relationships available to technology giants.

USD.AI co-founder David Choi said the GPU financing platform had received more than 700 funding requests worth $26.5 billion since launch, including about $8.7 billion under evaluation. Most applicants seek $10 million to $300 million for computing capacity expansion. Choi argued that traditional lending struggles with this transaction volume because lenders often negotiate each agreement separately. USD.AI uses standardized GPU-backed loans to connect borrowers with stablecoin capital and institutional lenders. USD.AI reported cumulative deployed capital exceeding $420 million and total value locked of $659 million as of Oct. 2.

Sengupta proposed that data center operators issue transferable digital claims granting access to GPU clusters during specified future periods. Operators could finance undelivered computing capacity by selling those claims to customers or pledging the claims for stablecoin loans. Companies could also pledge anticipated payments under existing customer contracts, potentially limiting reliance on traditional lenders. Sengupta believes blockchain networks could let lenders check collateral, track existing borrowing and automate payments during service delivery. Sengupta also proposed specialist intermediaries connecting computing buyers and sellers while handling collateral, financing and price risks across operators. Such markets could give smaller providers financing against verified computing capacity instead of relying exclusively on long-term customer contracts.

USD.AI lets investors stake USDai stablecoins in exchange for sUSDai, a yield-bearing token exposed to financing activities that include GPU infrastructure loans. PayPal's PYUSD stablecoin provides the backing for USDai. Returns for sUSDai holders depend on the underlying loan portfolio. USD.AI's September update reported a net annualized sUSDai yield of 6.9%. Physical computing equipment secures the loans and provides assets that lenders could recover or sell after nonpayment.

However, individual sUSDai investors neither own the GPUs directly nor decide how the equipment is handled after default. Dedicated legal entities own the equipment under USD.AI's financing arrangements, while lenders and designated agents hold collateral recovery rights. Following default, agents can control relevant accounts, redirect customer payments, substitute operators or sell equipment to recover unpaid debt. Recovery proceeds pass through the lending structure, leaving sUSDai investors dependent on that process rather than individual enforcement rights.

Unlike USDai, sUSDai offers no immediate redemption at a fixed dollar amount. Withdrawals follow 30-day cycles, and payments depend on the protocol's available liquid capital. Requests exceeding available funds can result in partial payments and further withdrawal cycles before investors recover the balance. The protocol does not automatically sell performing GPU loans to fund withdrawals, potentially creating a mismatch between infrastructure loan durations and investor liquidity needs.

GPU values can fall when newer equipment enters the market. Computing operators need continuing customer demand to produce loan repayment revenue. Equipment sales after default may leave unpaid balances even when lenders hold enforceable collateral rights. USD.AI provides residual-value warranties on certain loans to cover some losses when equipment sells below agreed values, subject to contractual conditions.

Conventional AI lending faces related risks: CoreWeave obtained a $2.6 billion credit facility maturing in roughly five years against customer contracts averaging about three years. CoreWeave's lenders therefore face exposure to whether CoreWeave renews contracts or secures replacement customers after those agreements end. Sengupta acknowledged that enforceable legal rights must support digital claims because blockchain records cannot by themselves ensure ownership or recovery of physical assets.