Cramer Backs ‘Compute Bonds’ as AI Hardware Becomes Collateral
While Wall Street leaders including Goldman Sachs’ David Solomon and NVIDIA’s Jensen Huang champion the instrument, analysts warn of obsolescence risks despite rising hedge fund interest in the chipmaker.
On the 11 August episode of CNBC’s Mad Money, host Jim Cramer endorsed the emerging concept of “compute bonds,” a proposed financial instrument that utilises artificial intelligence hardware as collateral for securitised credit. Cramer detailed how major investment institutions are preparing to issue these bonds backed by data centre equipment, drawing direct parallels to established securitised credit markets such as mortgage-backed or auto loan securities.
Cramer highlighted support from Goldman Sachs CEO David Solomon and NVIDIA CEO Jensen Huang, arguing that data centre assets retain value better than traditional collateral like vehicles. He noted that NVIDIA graphics processing units have shown little depreciation, with older models holding up better than cars which lose value immediately upon purchase. Cramer suggested that NVIDIA itself might issue bonds by the tens of millions to augment its cash program, describing the strategy as a brilliant new asset class comparable to gold or real estate.
The endorsement comes amid a structural shift in financial markets where AI hardware is increasingly viewed as long-lasting collateral. Cramer argued that the notes would entice yield-seeking money managers, stating that if losses occur, NVIDIA would likely absorb them. He expressed strong personal confidence in the asset class, believing that Jensen Huang and his associates have invented a new category of investment that appeals to investors chasing extra yield.
Despite the bullish commentary, analysts have warned of risks related to technological obsolescence and potential valuation write-downs. Unlike real estate or physical gold, semiconductor technology faces rapid obsolescence, meaning existing server clusters could suffer sudden devaluation if next-generation architecture delivers significant efficiency leaps. Fixed-income skeptics also highlight that if end-user demand for AI inference or model training cools, secondary market rental rates for compute power could decline.
Despite these concerns, hedge fund holdings in NVIDIA increased in the first quarter of 2026. The stock was held by 275 hedge funds in Q1 2026, up from 264 in the prior quarter, indicating rising high-conviction backing across active money managers. At the same time, the short percentage of float for NVIDIA sits at a minimal 1.26%, suggesting that few investors are betting against the chip titan’s central role in the artificial intelligence trade.
