The S&P 500 Has Fallen in 56% of Septembers Since 1928. Here's What That Means for Artificial Intelligence (AI) Stocks.

September historically has pressured the S&P 500, with Bank of America finding it has fallen in 56% of Septembers since 1928 and averaging about -1.17% in that month. The article argues any September pullback should be an opportunity to add to AI leaders: Nvidia recently reported fiscal 2027 Q2 revenue of $96.2B (up more than 2x YoY) and adjusted EPS of $2.22 (+120%), beating estimates, while guiding fiscal 2028 sales growth of 70% vs 44% expected. For Micron, management’s fiscal 2026 Q4 midpoint calls for $50B revenue and $31 non-GAAP EPS (+342% and +923%, respectively), with the Fed meeting in September amid 3.3% July inflation to potentially influence rates.

Analysis

The seasonal tape argument is weak by itself; what matters is that September often concentrates flows from tax-loss harvesting, quarter-end de-risking, and macro-event hedging. In a crowded AI trade, that can create temporary multiple compression even when the fundamental slope is still up, so the first-order move is more about positioning than operating deterioration.

NVDA remains the higher-quality duration asset in the group: its upside is less about another earnings beat and more about whether hyperscaler capex keeps extending into 2026. The bigger near-term risk is not demand collapse but a rates-led de-rating if inflation stays sticky after the Fed meeting; that would hit the entire AI basket before it changes any revenue trajectory.

MU is the more tactical name because memory is a cleaner read-through on whether AI infrastructure spend is broadening beyond accelerators into HBM and server build-outs. That also makes it more fragile on the downside: if pricing or mix disappoints into the Sept. 23 print, the stock can re-rate much faster than NVDA. The contrarian point is that the market may be underestimating how quickly MU can turn from beneficiary to tell on AI digestion, while overestimating the predictive power of a calendar effect that is historically small.

Key Decisions for Investors

• NVDA: buy incremental exposure only on a 5-8% September pullback or post-Fed risk-off tape; 1-3 month upside remains intact if 10Y yields stay contained. Falsify if the stock loses support after guidance or if rates break materially higher.
• MU: treat the Sept. 23 print as the key catalyst, not the month itself; prefer a staged entry after results unless channel checks confirm HBM pricing is still tight. If management implies decelerating ASPs or inventory build, expect 10-15% downside over 2-4 weeks.
• Pair trade: long NVDA / short QQQ for a 1-2 month window if you want AI exposure with market hedge; this isolates idiosyncratic capex strength from broader September de-risking. Exit if NVDA starts underperforming semis on a strong tape, which would signal the trade is no longer about seasonality but about fundamentals.
• Avoid shorting NVDA or MU on September seasonality alone; the better bearish expression, if inflation reaccelerates, is a rates-driven hedge rather than a single-name equity short. Watch the 10-year yield and the Fed statement as the real catalyst set.