History Says the Nasdaq-100 Does This in September. AI Investors Should Take Note.

September has historically been a weak month for major indexes, with the S&P 500 averaging declines from 1928–2023 and the Nasdaq-100 down an average 2.1% in September (2017–2025), with negative months 67% of the time. Recent AI-heavy Nasdaq-100 constituents have also seen sharp September drawdowns (e.g., Nvidia -11.8% in Sep 2023, -19.5% in Sep 2022, -7.4% in Sep 2021). The article argues that trying to time bearish periods is costly over the long run and suggests investors focus on staying invested/adding on pullbacks rather than selling ahead of October.

Analysis

This is less a fundamental bearish call on AI than a reminder that crowded duration trades get the harshest de-grossing when macro volatility rises. The marginal seller matters: NVDA and the rest of the AI complex are held as consensus growth exposure, so even a modest risk-off tape can compress multiples faster than estimates move, especially when positioning is already extended.

Second-order, the weakest links are usually the semi-capex chain rather than the cash-generators. If investors trim AI exposure, ASML, AMD, and MU can underperform NVDA on a beta-adjusted basis because they are priced more on future wafer starts and memory recovery than on near-term cash flow; that makes them more vulnerable to a narrative reset if order momentum softens. Conversely, large-cap software/mega-cap balance-sheet names should be relatively better defended because they can absorb a higher discount rate and are less directly exposed to a one-month factor unwind.

The contrarian point is that this seasonality is widely known, so it can become self-limiting: systematic flows and dealer hedging often create the first flush, not a persistent downtrend. The real watch item is whether September weakness is accompanied by a change in earnings revisions or AI capex commentary; without that, the move is likely a tactical de-risking episode, not the start of a structural top. A break back to fresh highs in QQQ/SMH after the first two weeks of September would likely invalidate the bearish seasonal setup.

For most long-only investors, the better response is hedging, not liquidation. The best risk/reward is a time-limited volatility hedge into the next macro/earnings catalysts, then reassess after the first drawdown rather than preemptively selling strong franchises.

Key Decisions for Investors

• Buy a 30-45 DTE QQQ or SMH put spread into September macro events as a tactical hedge; target a 2-3x payout if the index sells off 3-5%, with defined premium risk.
• If concentrated in AI semis, short a basket hedge: long-quality mega-cap software/defensives versus short SMH for 2-6 weeks; the thesis is factor de-grossing, not semiconductor fundamental collapse.
• Use NVDA as the highest-beta hedge candidate rather than AMD/ASML: short-dated downside protection is more efficient here because the stock’s positioning and index weight amplify seasonal flow pressure.
• Do not add to longs on the first weak session; wait for either a second-leg selloff or a stabilization in breadth. The trade is only valid if QQQ fails to reclaim its 20-day moving average after the initial dip.
• Falsifier/watch item: if NVDA earnings commentary or AI capex guidance remains firm and SMH holds its prior month low, treat September weakness as noise and remove hedges.