Fed Uncertain as AI Capex Jitters Shakeout Speculation

Markets traded with an unusual mix of strong micro data and fragile macro sentiment last week and nowhere was that clearer than the reaction to Nvidia’s excellent earnings. The fundamentals showed strong demand, a robust product cycle, and clean forward guidance—yet the stock slumped after an early surge. That tells me the market is wrestling less with Nvidia’s numbers and more with crosscurrents and lingering anxiety about whether the extraordinary AI capex cycle ultimately pays off. This is reminiscent of the late-1990s fiber-optic buildout—demand was real, but there turned out to be massive excess capacity. The good news is that this reset is washing out speculative excess rather than undermining the leaders of the cycle. That’s healthy. There is also an evident Bitcoin–NASDAQ correlation that is striking again with the Bitcoin selloff cascading into some of leading tech stocks as positioning unwinds.

The labor market continues to show the same “no-fire, no-hire” pattern I’ve emphasized for months. Weekly jobless claims are sitting right at the 220,000 sweet spot, squarely in the range that signals stability rather than stress. But continuing claims creeping to post-pandemic highs tell us workers who were laid off are struggling more to re-enter. Firms are holding tight to labor but not expanding headcount—a setup that often precedes a productivity-led growth spurt if demand stays firm.

This is why the holiday spending data in the coming two weeks is crucial. Real-time credit-card reads and retail commentary will reveal far more about underlying consumer momentum than backward-looking payroll reports that remain distorted by the shutdown. Strong spending will tilt the Fed toward a December pause; soft spending makes the December meeting genuinely live.

This is the most uncertain FOMC meeting in years because the Committee itself doesn’t yet know the answer. Powell prefers to signal decisions well in advance, but the data simply is not speaking loudly enough. Unless claims rise meaningfully over the next two prints or holiday spending surprises sharply lower, the base case is that the Fed waits—but sends strong guidance that cuts are coming early next year. Williams’ comments about “room for a cut” show that the dovish wing is already laying groundwork. Meanwhile, hawks will point to unemployment barely budging and demand still holding. It sets up a rare, genuinely suspenseful meeting—one where investors should expect volatility around both the statement and the new dot plot.

Private credit is seeing its first real pause after hundreds of billions in inflows. Widening CDS spreads on large investment-grade issuers like Oracle highlight that markets are reassessing risk rather than ignoring it. This is not crisis-level widening—far from it—but it signals that the era of “priced for perfection” financing is over.

Crucially, loan demand remains muted across the banking sector, with year-over-year deposit growth slipping under 3%. Banks aren’t funding broad credit creation, and the money supply has drifted lower again in recent weeks. I don’t want to overinterpret M2 because shutdown-related adjustments can distort the weekly flow, but the underlying liquidity environment is not inflationary. The Fed has the room to cut.

Long-term yields holding above 4.1% are entirely consistent with my view that the long bond is unlikely to revisit the sub-3% world of the 2010s even if the Fed cuts to the 3.25–3.50% range next year. The term premium has normalized, mortgage rates will stay near 6%, and exotic proposals like 50-year mortgages offer little relief—only about a 12% payment reduction compared to a 30-year, while leaving three-quarters of the principal unpaid after three decades. Housing relief will not come from financial engineering; it requires more land, more building, and fewer restrictions.

The broader macro hinge remains AI adoption. Consumers love AI as a personal assistant, but the trillion-dollar question is enterprise integration: Are firms deploying these tools, and are they realizing productivity gains large enough to justify the capex surge? Full-self-driving, robotics, and automation could free 6–8 million transportation jobs over time, unleashing major efficiency gains—but the diffusion timeline matters. Markets will continue oscillating between enthusiasm for the long-term payoff and fear of short-term overbuilds. I still believe this volatility is a healthy occurrence.