KEYTAKEAWAYS
-
USD.ai introduces a dual-token model where USDai provides dollar stability while sUSDai captures yield from AI hardware loans and U.S. Treasuries.
-
The project turns GPUs into on-chain collateral through the CALIBER framework, offering AI firms new financing channels while embedding into DeFi ecosystems like Curve and Pendle.
-
Rapid growth has pushed TVL above $250M, but long-term success depends on borrower adoption, regulatory clarity, and the protocol’s ability to manage collateral and liquidity risks.
CONTENT
THE AMBITION OF A NEW PLAYER
In the crypto market, stablecoins have long been dominated by giants. USDT and USDC hold overwhelming market share, while older decentralized options like DAI are slowly losing momentum. The space seems set, leaving little room for newcomers. Yet in 2024, a New York–based team called Permian Labs entered with an unusual idea: combining blockchain with artificial intelligence hardware. Their goal was to transform GPUs — the core resource of the AI industry — into on-chain collateral. That is how USD.ai was born.
The project’s leading figure, David Choi, came from traditional finance, having worked at Deutsche Bank’s investment banking division before founding the NFT lending platform MetaStreet. His thinking was simple: if blockchain can tokenize digital collateral, why not bring real-world hardware into the system? GPUs are the most critical resource for AI firms, but financing channels are limited. Stablecoins, on the other hand, are about turning reserves and credit into circulation. Combining the two felt like the perfect way to fill a gap in the market.
Investors quickly took notice. In August 2025, USD.ai announced a $13.4 million funding round led by Framework Ventures, with participation from Dragonfly, the Arbitrum Foundation, Bullish, and later YZi Labs (formerly Binance Labs). Within a month, the protocol’s total value locked surged past $250 million, far beyond expectations. In a space that looked saturated, USD.ai carved out a new niche with a fresh story.
TECHNOLOGY AND DESIGN
At first glance, USD.ai may look like just another collateral-backed stablecoin: users deposit USDC or USDT to mint USDai at a 1:1 ratio. But the structure is designed as a two-layer system that separates stability from yield. USDai itself is pegged to the dollar and generates no return, serving as a safe asset. When users stake USDai, they receive sUSDai, which carries more risk but accrues the interest generated by the protocol. Current yields hover around 6–8% annually. This split between stable principal and yield exposure allows the protocol to appeal to both conservative holders and yield chasers.
The real innovation lies in the collateral. Through Permian Labs’ CALIBER framework, AI companies can tokenize GPU hardware as NFTs. These tokens then serve as loan collateral, with loan-to-value ratios, insurance, and liquidation mechanisms in place. In essence, GPUs are treated like on-chain mortgages, giving borrowers liquidity while lenders collect interest. Meanwhile, idle funds are parked in U.S. Treasuries and other low-risk instruments to ensure liquidity and stable returns.
USD.ai is also linking itself to the broader DeFi ecosystem. USDai is already part of Curve pools on Arbitrum, providing stablecoin trading liquidity. sUSDai can be deposited into Pendle to split and lock in yields. These integrations help USD.ai embed itself into existing infrastructure rather than remain a closed system.
Still, the design is not flawless. GPUs depreciate quickly as new chips hit the market, and liquidations are more complex than selling tokens. Legal contracts and centralized processes are still needed to enforce claims. Even with tight on-chain logic, the off-chain layer must hold. USD.ai opens a new possibility, but it remains tethered to real-world execution.
RISKS AND CHALLENGES
Behind its rapid growth, USD.ai faces serious tests. The first is utilization. Early on, most deposits were not loaned to AI companies but held in Treasuries or money markets. During that phase, the protocol acted more like a high-yield savings tool than an AI financing platform. If this imbalance continues, its core narrative may weaken.
Liquidity is another concern. The QEV redemption queue was designed to manage large withdrawals by delaying or discounting redemptions. But stablecoins depend on trust. In a crisis, will users still get their money back quickly enough? So far, that remains largely untested. And if GPU values plunge, liquidation may fail to cover loans.
Regulation also looms large. USD.ai effectively takes deposits and makes loans — a function that normally requires a banking license. Running on-chain does not guarantee exemption from oversight. By channeling part of its reserves into Treasuries, USD.ai may trigger stricter disclosure and audit requirements. Cross-border hardware collateral only adds complexity, since enforcement varies across jurisdictions.
Market sentiment cuts both ways. The Allo Game reward program drew thousands of users by gamifying ICO allocations and airdrops, fueling FOMO and driving liquidity. But reliance on such incentives may prove unstable. Once the excitement fades, demand could fall sharply. Stablecoins run on confidence, and when it wavers, rebuilding it is far harder than sparking it.
FUTURE OUTLOOK AND USER ACCESS
Despite the risks, USD.ai is an experiment worth watching. In the broader RWA narrative, it represents a new path. Until now, RWA in crypto mostly meant bonds or real estate. By turning AI hardware into collateral, USD.ai expands the scope. For AI companies, it creates flexible financing for expensive GPUs. For crypto users, it makes stablecoin holding a way to tap into AI’s growth.
Long-term success depends on two conditions. First, the borrower side must grow, with more AI firms actually using USD.ai loans. Second, compliance and transparency must improve to satisfy regulators and reassure the market. Only then can USD.ai secure a lasting role as a distinct type of stablecoin.
For ordinary users, the entry point is straightforward. By depositing USDC or USDT, anyone can mint USDai. This stablecoin behaves like a dollar-pegged asset but offers no yield. To earn returns, users can stake USDai and receive sUSDai, which currently yields 6–8% annually. Those looking for more can deploy sUSDai into Pendle or other DeFi protocols to split or optimize yields. The distinction is simple: USDai offers stability, while sUSDai offers yield with risk.
This design gives users flexibility. Conservative holders can treat USDai as a stable store of value, while risk-tolerant users can pursue yield with sUSDai. Risks remain — GPUs can lose value, redemptions may be delayed under stress, and yields are not guaranteed. Still, USD.ai gives everyday users a new way to take part in AI infrastructure financing through just a few DeFi steps.
In a market where stablecoins seemed frozen, USD.ai has injected fresh imagination. Whether it becomes a durable bridge between crypto and AI, or just another fleeting experiment, it has already expanded the boundaries of what a stablecoin can represent.
