Fed Chair Kevin Warsh, Welcome to Your No-Win Scenario, Courtesy of President Donald Trump

The market’s key mistake is treating this as a pure policy story when the bigger transmission channel is the cost of capital for duration assets. A hiking bias or even a hawkish hold would hit the most levered beneficiaries of the AI capex cycle first: not just chip demand, but every supplier financing rack build-outs, data-center REITs, and software names priced on distant cash flows. NVDA is still a structural winner on unit demand, but its multiple is the most vulnerable to a 25-50 bps re-pricing in real yields; the first-order earnings hit is small, the second-order valuation hit can be large.

The relative winners are balance-sheeted cash generators and financials that can reprice assets faster than liabilities. BRK.B should screen better than the index in a tighter-rate regime because float and cash earn more without relying on cheap leverage or multiple expansion. By contrast, consumer-facing names like TGT face a margin squeeze from sticky input costs and slower discretionary demand, while bank beta like OZK is ambiguous: modest NII support near term, but worsening credit quality if inflation and rates stay high for another 1-3 quarters.

Contrarian risk: the consensus may be overestimating how much the Fed can or will do in one or two meetings. If labor softens, markets may quickly fade hike odds and re-expand growth multiples before any actual policy move. Falsifier for the hawkish-duration bearish view: core inflation decelerating for 2 consecutive prints and a dovish Fed hold; that would squeeze shorts in NVDA and other long-duration names within days, while preserving the structural AI spend story over 6-18 months.