Fed Chair Kevin Warsh Has an Inflation Conundrum, and Trumpflation Is Only Part of the Problem

The FOMC unanimously raised rates 25bps to 3.75%-4.00% on Sept. 16 as inflation remained above its 2% target for 66 consecutive months. The article attributes persistent price pressures to renewed 10%-12.5% tariffs on more than 80 countries, energy and supply-chain disruption from the Iran war and Strait of Hormuz closure, and AI infrastructure shortages and pricing power. Further tightening could curb debt-funded AI data-center investment, pressure elevated technology valuations, lift recession risk, and potentially end the market's AI-led rally.

Analysis

The investable issue is not a single 25 bp move but a higher-for-longer term-premium regime: AI capex competes with Treasury issuance and pushes the discount rate applied to long-duration equities higher even before policy rates materially constrain demand. That creates a nonlinear valuation risk for NVDA and the broader AI complex, where earnings may remain strong but multiples compress if the 10-year Treasury yield rises another 40-60 bp. The first-order beneficiaries are cash-rich, near-term-FCF compounders; the vulnerable cohort is leveraged data-center infrastructure, power developers, and smaller AI vendors reliant on frequent external financing.

The market may be overstating the direct CPI transmission from AI hardware. Accelerator, memory and networking bottlenecks are primarily an enterprise capital-goods issue, while consumer-price pass-through depends on cloud vendors successfully raising prices to businesses and businesses passing those increases through. The more immediate inflation channel is energy, shipping and tariff-related input costs; if these normalize, the Fed can pause without an AI-capex collapse. Conversely, a sustained rise in hyperscaler debt issuance and long-end yields would matter more than another isolated inflation print because it would raise project hurdle rates and force capex prioritization.

Over the next 1-3 months, the key catalyst is whether hyperscaler capex guidance remains intact alongside widening corporate bond spreads and a rising 10-year yield. A weak point in the bearish AI thesis is that NVDA's supply-constrained revenue and customer prepayments can cushion near-term demand; the larger fundamental risk emerges over 6-18 months when incremental capacity arrives and customers assess AI monetization against a higher cost of capital. Falsification for the cautious duration view: a retreat in the 10-year yield below its pre-hike level, stable IG spreads, and upward revisions to hyperscaler capex plans would support renewed AI multiple expansion.

Key Decisions for Investors

• Reduce net long exposure to high-duration AI beneficiaries into strength; retain NVDA only as a quality-weighted core versus less-profitable AI infrastructure names. Use a 10-year yield move of +50 bp from current levels or a 1-turn forward-P/E de-rating as the trigger to add hedges, rather than reacting solely to the next policy meeting.
• Establish a 1-3 month relative-value hedge: long XLF versus short QQQ, sized beta-neutral. Banks and insurers benefit from a moderately steeper curve and reinvestment yields, while QQQ carries concentrated duration and AI-capex sensitivity; exit if credit spreads widen sharply, which would turn a yield increase from constructive reflation to recession risk.
• Watch for a hyperscaler funding signal before shorting the AI supply chain: accelerate a cautious short basket in data-center equipment and lower-quality semiconductor names only if investment-grade spreads widen by 25 bp or more and capex guidance is cut. Missing inputs are customer concentration, backlog cancellation terms and debt maturity schedules; without them, a broad AI short is premature.
• Prefer energy and defense exposure as an inflation/geopolitical hedge over a blanket short of equities for the next 1-3 months, using XLE and ITA as liquid proxies. Take profits if shipping and crude benchmarks retrace materially or if evidence of a durable supply-route normalization emerges, since that would remove the most policy-relevant inflation impulse.