TS Lombard urges diversification, not an early exit, in its $20 trillion AI-crash scenario

TS Lombard’s answer to a hypothetical AI crash is not to flee stocks. A November 1 account in The Motley Fool describes its recommendation to retain normal risk through diversified global shares, despite a scenario erasing roughly $20 trillion in equity value. Government bonds enter the playbook only when financial strain appears.

The underlying paper, “Wargaming the AI Crash,” comes from TS Lombard strategist Dario Perkins. Yahoo Finance covered it on October 23, quoting his estimate that U.S. stocks could lose at least 30%. The November account details the firm’s portfolio response: avoid trying to call the peak, because an early defensive move could prove costly if investment keeps accelerating.

How a compute promise finances construction

Perkins’s concern starts with the funding chain behind the buildout, rather than share prices alone. In the report’s description, future purchases of computing power help secure financing for infrastructure today. The arrangement depends on AI labs continuing to raise money to meet those commitments.

TS Lombard compares that dependence to subprime lending, while acknowledging that the analogy is stretched. It does not expect a repeat of 2008. Its assessment also separates businesses that could face acute distress from the largest cloud companies, which it expects to suffer revenue downgrades without becoming serious credit risks.

Where Perkins sees acute vulnerability

• AI-focused cloud providers and Oracle.
• Standalone model developers.
• Construction companies and equipment suppliers serving the buildout.

Large market losses, a smaller economic shock

The November account breaks the equity scenario into a 40% fall in technology stocks and a 20% decline elsewhere, producing an S&P 500 loss of at least 30%. Those are modeled losses, not a description of a crash that has already happened.

The two accounts give slightly different debt-loss estimates. Yahoo Finance quotes Perkins at $125 billion to $200 billion across debt markets and private credit, alongside tech defaults and sharply wider credit spreads—the extra borrowing cost over safer debt. The Motley Fool gives a narrower $125 billion to $190 billion range for AI-related debt, including private credit.

Yet TS Lombard estimates a relatively mild U.S. recession, with output about two percentage points below the level it otherwise would have reached. Its reasoning rests on healthy private-sector balance sheets and room for Federal Reserve rate cuts. Lower borrowing costs could revive demand in areas squeezed by tight monetary policy, including housing.

That does not mean painless. Perkins expects capital spending to plunge and substantial construction job losses. TS Lombard estimates the data-center construction boom has created roughly 500,000 jobs, potentially reaching 800,000 by the end of 2027. Those jobs represent one direct route from an investment reversal to household incomes.

The unresolved decision is when to turn defensive

The firm’s working assumption puts a reversal in early 2028, but it says the timing could slip. Perkins does not claim to know whether recent market weakness marks the beginning of the end or will pass. The scenario therefore supplies a possible path through a bust, not a reliable countdown to one.

That uncertainty drives its recommendation. TS Lombard argues that this boom is especially concentrated in the United States and the AI supply chain, leaving room to diversify elsewhere. Betting against U.S. technology or buying long-dated bonds too soon risks underperformance if stocks keep rising and bond yields move higher. Its defensive shift waits for financial fragility, rather than a date on the calendar.