The S&P 500 Rose More Than 78% Over the Past Three Years. This Is What Happened the Last Time Two Times It Posted Such Strong Gains

The S&P 500 rallied roughly 78% from 2023 through 2025 (with returns of >20% in 2023 and 2024 and 16% in 2025), marking a rare three‑year surge of more than 75%; the last comparable multi‑year runs occurred around 2021 (≈90% three‑year gain) and 1999 (≈98%), which were followed by sharp subsequent declines (a 19% drop in 2022 and the dot‑com crash after 1999). The piece highlights that the 2023 recovery was aided by AI enthusiasm (e.g., ChatGPT) but cautions that strong recent returns are not a reliable sell signal and recommends reallocating away from richly valued names toward more modestly valued or dividend‑paying stocks rather than market timing.

Analysis

Market structure: The recent >75% three-year S&P run concentrated returns in mega-cap tech (AI hardware/software winners like NVDA) and market infrastructure (exchanges, data providers such as NDAQ) while compressing returns for small caps, cyclicals, and low-growth value names. Passive/ETF inflows amplified concentration risk—supply (float) of high-quality free-float shares is tight versus demand from indexes and quant/flow strategies, raising liquidity and skew risk. Cross-asset: continued equity strength keeps real yields low, flattening credit spreads but raising sensitivity to a policy shock; options markets show low skew but can gap vol up quickly on macro prints.

Risk assessment: Key tails are a Fed policy surprise (hawkish pivot causing a 10–20% equity drawdown within 3 months), a rapid slowdown in AI capex (GPU oversupply or export controls), or regulatory actions on platform monopolies; any could erase concentrated P/E premium. Immediate (days) risk = volatility spikes around CPI/FOMC and NVDA earnings; short-term (weeks–months) = re-rating risk if EPS beats fail to meet stretched multiples; long-term (years) = secular winners benefit if AI drives durable revenue. Hidden dependencies include reliance on buybacks to support EPS and index rebalancing mechanics that can amplify moves.

Trade implications: Prefer selective longs in market infrastructure and cash-flow-rich dividend names while hedging index concentration. Use limited, defined-risk options to preserve upside: 3–6 month spreads on NVDA to capture AI upside and 1–3 month SPY put spreads as tail insurance. Rotate 3–5% from broad SPY exposure into exchange exposure (NDAQ) and dividend ETFs to lower downside beta while keeping tactical growth exposure.

Contrarian angles: The 1999 parallel is noisy—1999’s earnings growth and capital structure differ from 2023–25 (buybacks + higher free-cash-flow today), so indiscriminate “sell everything” is risky. Consensus may underprice the durability of AI-driven revenue for middleware/platforms even as it overprices low-quality momentum names. Unintended consequence: mass hedging/ETF de-risking could create liquidity vacuums; position sizing and cost-controlled hedges matter more than binary market-timing.

Key Decisions for Investors

• Establish a 2–3% long position in NDAQ within 2 weeks (target 12–18% upside in 6–12 months); set stop-loss at -10% and scale up into any 8–12% pullback to capture sustained derivatives/volumes and buyback tailwinds.
• Deploy a defined-cost tail hedge: buy a 3-month SPY 5% OTM put / 10% OTM put spread sized so premium ≤1.5% of portfolio value within 7 trading days; if SPY falls ≥10% realize hedge and redeploy into quality names.
• Buy a 6-month NVDA 20%/35% OTM call spread sized to 1% of portfolio to capture continued AI adoption; take profits at +50% on the spread or cut if NVDA trades -15% from entry.
• Reduce gross S&P 500 ETF (SPY/VOO) weight by 3–5% within one month if equity share >50% of portfolio and redeploy proceeds: 60% to NDAQ (above) and 40% to dividend ETF VIG to lower forward beta and add cash-flow yield.
• Establish a 2% tactical long in NFLX only on a pullback ≥12% within 3 months or on confirmed positive subscriber/ARPU catalysts; target 25–35% upside over 12 months with a 12% stop-loss.