The man who invented the Fed’s magic trick just died. His successor is about to try it again

The article links the current AI-driven market rally to Greenspan’s late-1990s 'irrational exuberance' era, framing the key policy question as whether the Fed should keep rates elevated or allow the boom to run. It highlights the risk that AI valuations could prove 'wild' if productivity gains arrive too late, with the Fed potentially facing a repeat of the dot-com-era mispricing dynamic. The piece is mostly historical and analytical, but it underscores a market-wide policy debate that could influence rates, valuations, and risk appetite.

Analysis

The market’s biggest vulnerability here is not “AI overvaluation” in the abstract; it is policy asymmetry. If the Fed tolerates asset inflation under the banner of future productivity, multiples on the most narrative-sensitive AI beneficiaries can stay detached from fundamentals much longer than bears expect, but the same tolerance raises the probability of a later, sharper policy reset once labor or goods inflation stops cooperating. That creates a barbell regime: quality mega-cap AI platforms can keep compounding on real cash flow, while the thinly capitalized AI-adjacent cohort is exposed to a sudden de-rating when rates or risk appetite change.

The second-order effect is that this is less about semiconductors than about duration exposure across the equity market. The longer AI remains a “rates don’t matter” story, the more capital gets pulled into long-duration software, hyperscalers, and power/infra beneficiaries, while defensives and cyclicals lag on relative flows. But if the productivity dividend fails to show up in the hard data over the next 2-4 quarters, the unwind should hit the most levered beneficiary chains first: compute, networking, and unprofitable application names that rely on multiple expansion rather than earnings.

The contrarian view is that the market may be underpricing how hard it is for the Fed to actually repeat the late-1990s playbook. Today’s disinflation tailwinds are weaker, fiscal deficits are larger, and the labor market is less forgiving, so the tolerance for a speculative overshoot may be much lower than investors assume. In that scenario, the right trade is not a broad short on AI, but a dispersion trade against the most crowded, least profitable names while staying long the structural winners with hard free cash flow and pricing power.

Key Decisions for Investors

• Long MSFT / short an AI-adjacent basket of unprofitable software and infrastructure proxies for 3-6 months; use this as a dispersion trade that benefits if AI leadership persists but breadth narrows.
• Buy QQQ put spreads 4-6 months out financed by selling downside in XLU or XLP; this hedges the risk that the Fed eventually leans hawkish and duration equities re-rate lower.
• Long NVDA only on post-event pullbacks, but pair it against a basket of high-multiple, negative-FCF AI app names; the market is likely to continue rewarding scarce earnings quality over pure story exposure.
• If 2Q/3Q macro data fail to confirm an AI productivity step-up, add to short ARKK or similar high-duration growth exposure for a 2-3 month window; these names should underperform first if policy expectations tighten.