CEA Chair Phelan says the Fed's rate hike 'was a mistake

September payrolls increased by 29,000, but Council of Economic Advisers Chair Christopher Phelan argued the report was stronger than the headline because unemployment barely changed as labor-force participation rose. Phelan said AI has not yet materially displaced workers and that AI-related investment is not inflationary. He also called the Federal Reserve's September rate hike a mistake, disputing the view that elevated rates will not curb AI demand.

Analysis

The investable question is whether labor-force re-entry is genuinely expanding productive capacity or merely masking weaker labor demand. If participation rises while wage growth, weekly hours, and temporary-help employment soften, the market can price a faster easing path without requiring a recession; that is supportive over the next 1-3 months for duration-sensitive software and unprofitable growth, but less supportive for cyclicals whose earnings require accelerating hiring and end-demand. A single payroll print is insufficient to establish this regime shift, particularly given recurring revisions and survey volatility.

The AI-capex complex is more rate-sensitive than the demand narrative implies. Hyperscalers can fund infrastructure internally, but downstream buyers of AI hardware, power capacity, and data-center real estate depend materially on financing conditions; higher real yields compress the present value of long-dated returns and can delay enterprise deployments even if GPU demand remains robust. The second-order beneficiary of disinflation is therefore not necessarily NVDA alone, but application software and cloud names where lower discount rates can permit multiple expansion before AI revenue is fully visible.

Consensus may be too binary on the inflation implication of AI spending. Near-term investment is disinflationary only if productivity gains arrive before power, grid, construction, and skilled-labor bottlenecks raise input costs; the latter risk is more likely over 6-18 months and favors grid equipment and power-exposure names over broad technology. This thesis is falsified by reacceleration in average hourly earnings/core services inflation, or by a sustained rise in 10-year real yields despite labor-market cooling.

Key Decisions for Investors

• Use a 1-3 month relative-value expression: long IGV versus short IWM in equal dollar beta-adjusted sizes. Softer labor data and lower yields should reward recurring-revenue, duration-heavy software more than small-cap companies carrying refinancing and labor-cost sensitivity; exit if the 10-year real yield rises above its pre-report high or next payrolls rebound materially alongside firmer wages.
• Do not chase broad AI hardware on this signal. Maintain NVDA/SMH exposure only against a hedge in rate-sensitive infrastructure beneficiaries such as VRT or ETN, and add only after hyperscaler capex guidance confirms demand rather than relying on policy commentary; a cut to 2026 capex plans or weakening data-center backlog would invalidate the trade.
• Add a tactical long in TLT or receive 2-year rates only after confirmation from the next CPI and wage data, rather than on the payroll headline alone. Risk/reward is favorable if cooling labor translates into 25-50 bps of easing-path repricing over 1-3 months; stop out if core services inflation reaccelerates or unemployment falls back without participation improvement.
• Monitor power and grid bottlenecks as the 6-18 month contrarian AI expression: build a watchlist for ETN, PWR, GEV and CEG on valuation pullbacks. The structural thesis requires sustained utility interconnection queues and rising contracted data-center load; regulatory rate disallowance, falling power prices, or materially slower hyperscaler buildouts are the key falsifiers.