USDai is a dollar-pegged synthetic stablecoin issued by Permian Labs, built on M0's stablecoin infrastructure and designed around a specific niche: financing the physical buildout of AI compute. Rather than backing the token with cash, treasuries, or crypto assets the way most stablecoins do, USDai is collateralised by loans made against GPU hardware and related AI infrastructure, positioning it as a credit instrument for an industry that's been starved of traditional financing.
What it does
Permian Labs issues loans to AI infrastructure operators and startups using GPUs as collateral, with a framework called CALIBER used to tokenise that hardware as verifiable, on-chain warehouse receipts governed under standard UCC commercial law rather than a purely crypto-native legal wrapper. Depositors provide the dollars that fund those loans and receive USDai; a yield-bearing variant, sUSDai, lets holders earn a share of the interest generated by the underlying GPU-backed lending book. The pitch is that approval times for AI hardware financing can be cut by more than 90 percent compared with traditional lenders, addressing a real gap — data centre and GPU financing has become a bottleneck as demand for AI compute has outpaced the capital available to build it. The protocol raised a $13 million Series A in August 2025 led by Framework Ventures, with Dragonfly and Arbitrum among the participants, and by September 2025 had filled successive $250 million deposit caps, bringing total deposits to roughly $500 million.
USDai's growth has tracked the broader AI infrastructure financing boom directly, giving it a narrative tailwind that most stablecoins don't have, but that also means its fortunes are tied tightly to the health of the AI capex cycle rather than to diversified, low-correlation collateral.
Risks
This is a genuinely novel collateral model, and novelty here means it hasn't been stress-tested through a full credit cycle. GPU hardware depreciates quickly as newer chip generations arrive, and its resale value in a downturn — when many AI infrastructure operators might be distressed simultaneously — is far less certain than the value of the treasuries or cash backing conventional stablecoins. A wave of defaults among GPU-collateralised borrowers during an AI spending slowdown could stress USDai's peg in ways that don't have a close historical precedent to model against.
Holders are also relying on CALIBER's legal and technical framework correctly translating physical hardware ownership into enforceable on-chain claims, on Permian Labs' underwriting discipline in a fast-growing, competitive lending market, and on continued strong demand for GPU compute holding up loan valuations. As with any young, narrowly focused credit protocol, concentration risk — in this case concentration in one asset class, one industry cycle, and a small set of large borrowers — is the central thing to watch.