Highlights

  • Bullish (NYSE: BLSH) is extending a $100 million debt facility to USD.AI to fund GPU-collateralized AI infrastructure loans.
  • USD.AI, built by Permian Labs, now holds more than $225 million in locked crypto assets.
  • Bullish plans to list USD.AI's yield-bearing token, sUSDai, across trading pairs with dedicated market-making support.
  • The loans are structured as non-recourse debt secured solely by GPU and server hardware, isolating risk from borrowers' balance sheets.
  • The deal follows Bullish's earlier $4 million equity stake in USD.AI, its first investment since going public.

Bullish, the crypto exchange operator behind CoinDesk, is providing a $100 million debt facility to USD.AI, a protocol that finances AI-infrastructure operators against GPUs and other high-performance computing hardware, according to a report by PANews on August 28. The capital will fund USD.AI's lending book, and Bullish intends to list the protocol's yield-bearing token, sUSDai, in trading pairs on its exchange. USD.AI, developed by Permian Labs, currently holds more than $225 million in locked crypto assets.

How the GPU-Backed Facility Works

USD.AI's model turns physical compute into on-chain collateral. AI infrastructure operators pledge GPUs and servers against loans, while depositors who fund that lending pool receive USDai, a dollar-pegged token, or its staked variant sUSDai, which accrues yield from the underlying loan book. In a statement announcing the facility, the companies said the loans are structured as non-recourse debt secured exclusively by the GPU infrastructure itself, keeping the risk isolated from borrowers' broader balance sheets. David Choi, chief executive of Permian Labs, framed the logic behind the deal directly:

"Compute is becoming a credit market in its own right," and the new facility will "help USD.AI finance more of the AI buildout."
A person holding money in front of a computer screen
Photo by Jakub Żerdzicki on Unsplash

The $100 million line adds meaningfully to a protocol that has scaled quickly: USD.AI's total value locked has grown from roughly $62.7 million a year ago to more than $225 million now, built on a $13 million Series A round led by Framework Ventures. The protocol has already underwritten facilities well beyond its own balance sheet, including financing for data-center operator QumulusAI and GPU deployments managed by Hydra Host, evidence that demand for non-dilutive, asset-backed compute loans is outpacing what many banks are willing to underwrite.

Why an Exchange Is Underwriting AI Compute

Related: Stablecoin Supply Snaps Three-Month Slide With $1.7B August Jump

For Bullish, the facility is a bet that GPU-backed credit becomes a durable, sizable asset class rather than a niche crypto experiment. Thomas Cowan, Bullish's head of tokenization, argued that "credible, well-structured real-world assets belong onchain," and the companies pitched compute financing as a capital-intensive category that could eventually rival auto loans and home-equity lines of credit in scale within private credit markets. The arrangement also plugs into a broader trend of stablecoin-denominated capital chasing real-world yield, arriving as overall stablecoin supply has been climbing again after months of stagnation. By committing exchange liquidity to originate loans and then creating a secondary market for the resulting yield token, Bullish is positioning itself as both financier and marketplace for AI-infrastructure debt, a dual role that could deepen liquidity but also concentrates exposure to a single counterparty relationship.

What to Watch Next

The most concrete near-term signals are the rollout of sUSDai trading pairs on Bullish's exchange, backed by a dedicated market-making program, and how much of the $100 million facility USD.AI actually draws down as it originates new GPU-backed loans. If early tranches perform without defaults, the structure could become a template other exchanges and asset managers look to replicate, further blurring the line between crypto-native balance sheets and traditional AI-infrastructure project finance.