Bond Traders Are Agonizing Over $70 Billion of Shadow Credit Backstops For AI Companies

August 16, 2026 - 02:45

Even before Nvidia Corp. unveiled its eye-popping $500 billion financing partnership this week, bond traders were already losing sleep over a quieter but troubling number: roughly $70 billion in shadow credit backstops tied to major artificial intelligence companies. These are liabilities that do not show up on balance sheets, yet they could suddenly become very real obligations at the worst possible moment.

The concern centers on credit support arrangements, often structured as guarantees or standby facilities, that AI firms have quietly arranged with banks and other lenders. These backstops are meant to reassure creditors and suppliers, but they are not recorded as debt. That means investors have little visibility into how much risk is actually building up in the system. When markets are calm, these arrangements are barely noticed. But if a downturn hits, or if one of these AI companies stumbles, the hidden liabilities could flood into the open, forcing banks to absorb losses or triggering a cascade of margin calls.

Traders are now trying to figure out just how much of this off-balance-sheet exposure is out there, and what it would take to make it visible. The $70 billion estimate is rough, and some think the real figure is higher. The problem is that these structures are often bespoke, with terms that are not publicly disclosed. That lack of transparency is making it harder for bond investors to price risk accurately, especially as AI companies continue to borrow heavily to fund data centers and chip purchases.

The Nvidia deal, while massive, has only added to the unease. It shows how deeply intertwined AI financing has become with the broader credit markets, but it also highlights how much of that financing relies on promises that are not fully accounted for. For now, the market is holding steady, but the anxiety is real. Bond traders are watching closely, knowing that a single shock could turn these phantom liabilities into very real pain.