AI Mega-Caps Push SPY to Records; Stock Picker Frustration Mounts

The sustained rally in US equities has been driven almost entirely by a handful of mega-cap artificial intelligence plays, leaving active portfolio managers and broader market participants behind. Bloomberg reported that just 1 in 4 active managers are beating the S&P 500 in 2026, a frustration that echoes the concentration peak of 2023-2024. NVDA, META, GOOGL, and MSFT have accounted for a disproportionate share of SPY gains, with NVDA alone adding 20% in seven days. Meanwhile, mid-cap and small-cap indices, represented by the Russell 2000, are lagging as capital rotates into large-cap technology names.

Market breadth metrics are deteriorating even as headline indices touch all-time highs. The market structure now mirrors the most acute concentration periods in recent memory: mega-cap dominance is shielding index performance while underlying equity health (measured by advancers, market-cap-weighted breadth, and non-mega-cap participation) is weakening. This creates a systemic risk if sentiment shifts away from the AI trade, as margin calls and rotation dynamics could be swift and brutal. Retail traders have openly acknowledged being sucked into the momentumThe empirical fact that winners keep winning over the medium term., with social media mentions of "NVDA full porting" and emotional decision-making at market tops.

The structural issue is that AI capex demand is real and durable, but valuation multiples for NVDA and peers are pricing in perfect execution and continued outsized demand share for years. A modest miss on earnings or guidanceCompany-issued forecasts of future financial performance., or evidence that in-house alternative chips are gaining traction faster than expected (as ARM and smaller vendors compete for inference workloads), could unwind the concentration quickly. JPMorgan's Asset Management division warned that the high-grade corporate bond rally faces two stumbling blocks: even bigger AI capex surges and waning retail participation, both of which could destabilize credit and equity markets.

The bull case emphasizes that AI infrastructure is a genuine multi-year capex cycle with winners and losers, and that owning the leaders (NVDA, META) is justified by earnings growth and margin expansion. However, the concentration is creating a crowded trade with elevated execution risk and limited margin of safety for new entrants. The market's ability to absorb a surprise or earnings miss without a sharp correction is questionable.