TS Lombard has modeled a potential $20 trillion decline in equity markets if the AI investment boom unravels alongside a mild recession.
The scenario lands at an awkward moment. Credit markets are already flashing warnings about some of the most indebted players in the AI buildout, and stocks tied to the trade just had a rough week.
The skeptics are in the trade too
TS Lombard economist Dario Perkins argues the debate on Wall Street has moved past whether AI is a bubble. In his telling, the conversation now centers on timing and severity: when the unwind arrives and how much it costs.
“Even the AI sceptics are fully invested.”
Last week offered a rough preview of what that could look like. Neoclouds, the specialist companies that rent out GPU computing power for AI workloads, shed a fifth of their value over five trading sessions.
The Nasdaq, meanwhile, posted its worst week since Liberation Day.
CoreWeave becomes the canary
The sharpest stress showed up at CoreWeave, one of the most prominent neocloud operators. Its credit default swaps, which work like insurance policies that pay out if a borrower fails to repay its debt, priced roughly 63% odds of default.
In mid-October 2026, CoreWeave’s five-year CDS reached highs between 1,100 and 1,194 basis points. A basis point is one-hundredth of a percentage point, so buyers were paying more than 11% of the insured amount each year for protection. Those levels implied a 60% to 63% likelihood of default over five years.
The bond market told a similar story. CoreWeave’s unsecured bonds yielded near 15%, and their prices slipped below 80 cents on the dollar as of October 15, 2026. The company carries approximately $50 billion in debt.
Equity holders have felt it too. CoreWeave’s shares have recently dropped around 15% to 25%.
A revenue loop that feeds itself
The deeper worry in TS Lombard’s analysis is structural. The firm notes that about 85% of revenues within the AI ecosystem come from recycled capital expenditure among large tech companies and related firms.
The vulnerability appears if financing tightens or if businesses and consumers outside the circle fail to pay for AI at scale. In that case, the spending could slow at once, and revenue across the chain would fall together.
Why Perkins does not expect a full-blown meltdown
For all the alarming numbers, Perkins stops short of predicting economic catastrophe. He argues the likely outcome is not a balance-sheet recession across the wider economy, and that any downturn tied to an AI unwind would likely remain contained rather than triggering significant macroeconomic fallout.
It is worth noting the scenario TS Lombard models pairs the AI crash with a mild recession, not a severe one. The $20 trillion figure, in other words, is not built on an apocalyptic macro backdrop.
What investors should watch
The first thing to track is credit. CDS spreads and bond prices for leveraged AI infrastructure firms like CoreWeave have proven to be faster warning signals than stock prices. If those spreads keep widening, refinancing gets harder and more expensive. For companies carrying debt loads in the tens of billions, that can turn a sentiment problem into a solvency problem.
The second thing to watch is whether outside demand for AI shows up. TS Lombard’s 85% recycled-capex estimate implies the industry needs external customers to pay meaningfully for AI products before the internal spending loop runs out of steam.
The third factor is positioning. Perkins’ observation that even skeptics are fully invested suggests crowded trades. Last week’s five-session drop in neocloud valuations showed how fast that can move.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.