New stablecoin connects crypto investors to real-world Nvidia AI GPUs that earn money by renting out compute power to AI devs — USD.AI lets crypto investors make bank off AI compute rentals
Let's turn your stablecoin into a less-than-stable AI investment.
USD.AI is a new decentralized finance (DeFi) protocol that connects crypto investors to real-world Nvidia AI GPUs that earn money by renting out compute power to AI developers. According to CoinDesk, DeFi has many stablecoins backed by Treasury earnings, while many smaller AI players are struggling to raise capital to acquire GPUs for compute. USD.AI aims to bridge these two markets by allowing stablecoin holders to lend their holdings to the protocol, which then uses them to purchase AI GPUs. These are then rented to AI developers, with the proceeds from those rentals servicing the debt and providing yield to the original lenders.
This model delivers investors a much higher return than Treasury rates while giving startups easier access to AI GPUs. Although this also means a higher risk for those investing their stablecoin into the protocol, USD.AI is taking steps to reduce volatility. The protocol uses a three-tiered structure to help keep it safe: CALIBER, FiLO, and QEV. The first mechanism tokenizes each particular GPU as an NFT. These GPUs are installed at an insured data center and legally documented, ensuring enforceable claims and physical custody of the asset. Loans are then issued against the NFT to fund the equipment, turning the token into collateral.
The next mechanism, FiLo, is managed by risk curators who underwrite the loans with their own funds. These are called first-loss capital and would cover any losses if a GPU borrower defaulted, serving as a buffer for lenders. These curators would also have the power to choose who to lend out the GPUs to, so there’s no single authority that can pick and choose borrowers. But because curators only make money when the loan is repaid, it helps ensure that their values are closely aligned with both the protocol and the lender.
Finally, there is the QEV, which stands for Queue Extractable Value. As the name suggests, it’s a queue for when lenders want to extract value from their investments (i.e., withdraw capital from the protocol). Instead of letting users withdraw everything at once, they must line up according to the system, with each receiving repayments over time from loan payments. Those who want a quicker exit must pay a premium to get their investments back much earlier, which is used to compensate those waiting in line and also help ensure liquidity for the protocol.
CoinDesk reports that the current yield for staked sUSDai ranges from 13% to 17%, which is much higher than the average yield on U.S. 10-year Treasury bonds. Furthermore, this yield comes from repayments by GPU operators, not from emissions or leverage loops, making its returns real and tangible and sourced from entities that offer AI services.
This new protocol gives cryptocurrency investors the chance to get in on the AI bandwagon while giving them relatively healthy returns. However, it also opens them up to the risks of the AI bubble. Many companies, individuals, and institutions are investing billions of dollars in AI hardware and software, but the market and revenue aren’t there yet to support such massive expenditures. Experts say we aren’t there yet when it comes to the AI bubble popping, but if and when it does, we would probably see trillions wiped from the market overnight, affecting almost everything, including USD.AI.
Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.

Jowi Morales is a tech enthusiast with years of experience working in the industry. He’s been writing with several tech publications since 2021, where he’s been interested in tech hardware and consumer electronics.