The Fed Has AI’s Back | Dan Krausz on “Liquidity Trap Door” In AI Infrastructure and Why Policymakers Are Inclined to Support Financial Conditions

Dan Krausz attributes the AI-hardware selloff chiefly to tightening liquidity, not a sudden collapse in AI fundamentals; the fundamental concerns, he says, were already present four or five months earlier. He sees three main risks to the AI trade: weaker demand, policy blocking data-center construction, and inadequate financing. Demand still appears strong. AI capex may be near its peak rate of increase, but not its peak level, and can continue growing if financing remains available. He distinguishes real AI adoption from the assumption that every potential use is already funded: the immediate vulnerability is the capital-intensive infrastructure buildout.

Krausz defines liquidity as the global availability of capital, shaped by fiscal deficits and their financing, central-bank rates and balance sheets, lending, and the dollar. Treasury and AI issuance have risen as liquidity has weakened, while Treasuries and hyperscalers tend to secure funding first. Hyperscalers have used accumulated cash for recent investment but are increasingly turning to markets, competing for capital and putting pressure on weaker borrowers. Rising rates can rapidly reduce excess liquidity, especially with non-residential investment near its highest share of GDP. Financing spreads illustrate the strain: CoreWeave’s unsecured bonds reportedly yielded as much as 15%, versus roughly 9–10% at issuance. Bank standards had mostly held up, though JPMorgan reported incremental tightening and private-credit lenders had also tightened.

Krausz thinks policymakers are inclined to keep liquidity stress from derailing AI investment, given officials’ statements that winning the AI race is crucial and AI will be disinflationary. Greater Treasury funding needs, he argues, leave them less able to tolerate prolonged stress than in past cycles. Treasury tools such as front-end issuance and buybacks can add liquidity; front-end financing is more liquidity-positive than long-end issuance and could accommodate long-end borrowing for AI infrastructure. A government-backed, Fannie Mae-like AI financing model has been discussed by investment figures, but Krausz says it is neither necessarily needed nor imminent. Policy support is no guarantee: persistent inflation could force policymakers to withdraw liquidity.

The Fed’s October commentary struck Krausz as mixed, not clearly dovish or hawkish: it wants to keep tightening because inflation remains too high, but at a measured pace to avoid tightening financial conditions so severely that it must reverse course. Warsh acknowledged tighter financial conditions and said this was partly intended; the Fed also described the economy as strong and discussed broadening capex. Krausz says asset prices respond to liquidity changes before the wider economy, potentially alerting policymakers early. In a hypothetical example, a modest liquidity decline could send the S&P down 10% before officials respond and markets rebound. A brief tightening may unsettle markets without materially damaging activity, while six to nine months of persistent tightening can affect the economy.

His equity outlook depends chiefly on liquidity. Without further tightening, he expects the economy and earnings to remain “okay”; additional tightening could trigger a 10% or greater “trapdoor” correction. He is watching for a spike in term premium, weakening U.S. capital inflows, substantial dollar weakness, or interest rates above nominal GDP as signs deficits are turning negative for equities; he says none is present now. He expects material slowing over the coming year, including a weaker ISM and weaker non-AI cyclicals, but not a recession. A 6% fiscal deficit and strong capex can offset weaker consumption. He also expects the quality factor to do well and says the consumer’s discretionary-spending capacity has fallen materially over four years, though consumers have not yet weakened.

Meta, Google, and Microsoft did not raise capex guidance in their latest reports, the interviewer noted, unlike the repeated increases of the prior two years. Krausz considers holding guidance steady an appropriate pause that may ease financing pressure; further increases could prove unsustainable. He sees no material slowdown in AI spending yet, but says it could slow substantially by the second half of next year if liquidity remains tight. Returns are currently clearer for AI providers and investors than for businesses using AI to grow revenue. He questions whether early efficiency gains will become lasting margin advantages as competitors adopt the technology, and says cost savings will take time to assess. If liquidity rebounds in three or four months, he says the infrastructure selloff could recover substantially, potentially about 70% of the way back; if it does not, company balance sheets supporting the buildout may not be enough to sustain valuations.