Bullish has provided USD.AI a $100 million stablecoin facility to fund loans backed by GPU infrastructure, according to a market update posted by CoinMarketCap on August 29, 2026. The deal points crypto-native capital at one of the tightest chokepoints in technology right now: the graphics processing units that train and run AI models.
The move lands during a soft session for crypto. Bitcoin traded at $77,723, down 3.0% on the day as of August 29, 2026, with Ethereum at $2,442, down 2.4%. The wider Fear and Greed Index still read 77, or Greed, even as spot prices slid. Against that backdrop, a nine-figure credit line routed into hardware financing reads as a bet on cash flow rather than token price.
Crypto capital meets the compute shortage
USD.AI is a lending protocol that finances physical AI infrastructure, using tokenized debt to move stablecoin liquidity toward operators who need to buy or deploy GPUs. The pitch is straightforward. Data-center operators and AI firms need expensive chips faster than traditional lenders will underwrite them, and stablecoins can settle that credit in hours instead of weeks.
Bullish supplying the facility gives that model a large, regulated counterparty. Bullish runs a crypto exchange and custody business and went public earlier in 2026, so the capital here comes from a balance sheet that answers to public-market disclosure, not an anonymous treasury. A $100 million commitment from that kind of lender is a signal that the GPU-financing thesis has moved past the pilot stage.
Loans backed by chips, not tokens
The collateral is the detail worth sitting with. Most on-chain credit is secured by crypto assets, which means a loan and its backing fall together when the market drops. Here the loans are backed by GPU hardware, an asset whose value is tied to AI demand rather than to the price of any coin.
That changes the risk shape. A drawdown like the one printing on August 29 does not directly impair the collateral behind these loans, because a rack of chips keeps earning rental income whether or not Bitcoin is red on the day. The trade-off is different: GPU values can fall if a newer chip generation lands or if compute demand cools, and hardware is harder to liquidate quickly than a token position. Underwriting standards, resale markets, and depreciation curves now matter more than a liquidation bot.
A funding gap the market keeps flagging
The financing squeeze around AI compute has surfaced repeatedly this year. Publicly traded miners have pivoted rack space toward AI hosting to chase steadier revenue, and some have leaned on share sales to close the gap between what buildouts cost and what they generate. Routing stablecoin liquidity into the same shortage gives operators an alternative to dilutive equity raises, at the price of taking on secured debt.
For the crypto side, this is another example of on-chain dollars finding a use case outside trading. The same stablecoin rails that settle exchange balances and card payments are now underwriting industrial hardware. It widens the definition of what "crypto finance" funds, from leveraged positions toward real-economy assets with cash flow.
Details beyond the headline figure remain thin. The single confirmed source so far is the CoinMarketCap update; loan terms, interest rates, the size of individual advances, and the seniority of Bullish's claim were not disclosed at the time of writing. Anyone weighing the credit risk should wait for USD.AI or Bullish to publish the structure directly. This is reporting on an announcement, not an endorsement of the underwriting, and none of it is financial advice.
Overview
Bullish has committed a $100 million stablecoin facility to USD.AI to fund loans secured by GPU hardware, connecting a regulated crypto lender to the AI compute shortage. The structure swaps token collateral for physical chips, which insulates the loans from a day like August 29's 3% Bitcoin slide but introduces hardware depreciation and liquidity risk instead. The key numbers to watch next are the loan terms and default provisions, which have not yet been published.
