Semiconductor crowded trade conviction drops to 53% from 82%, BofA survey

Editorial illustration: Three hands touch an oversized semiconductor chip on a dark pedestal, while two hands on the right hover apart from it, one with its palm raised.

In brief

  • Semiconductors held most-crowded-trade rank for fourth consecutive month in BofA Global Fund Manager Survey
  • Manager conviction collapsed to 53% in September from 82% in July despite unchanged crowding rank
  • Bond yield anxieties replaced AI bubble fears as primary tail risk concern for fund managers
  • Consensus erosion signals fragmenting conviction despite sustained positioning in semiconductor longs

The Conviction Collapse

The decline tells a story of momentum fading even as positioning holds. In April 2026, roughly 25% of fund managers identified semiconductors as the most crowded position. By May, that figure nearly tripled to 73%. It hit 80% in June before peaking at 82% in July. The drop from 82% to 53% in two months marks a sharp erosion of consensus, yet semiconductors remain atop the crowded-trade list.

A term has emerged on trading desks to describe this paradox: "frozen bulls." These are fund managers who believe valuations are stretched but remain unwilling to reduce exposure. They're stuck—conviction wavering, but conviction not yet broken.

Why the Shift?

AI-driven chip demand from Nvidia, AMD, and Taiwan Semiconductor turned the semiconductor sector into the defining trade of the current cycle. The iShares Semiconductor ETF (SMH) had gained 99% year-to-date as of mid-June 2026. The Philadelphia Semiconductor Index notched multiple all-time highs during the same period.

But sentiment is turning. In June 2026, 28% of fund managers identified an AI bubble as a top tail risk. That figure jumped to 45% in July, one of the fastest-rising concerns in recent survey history. By September, AI bubble fears were overtaken by anxieties about rising bond yields.

Higher bond yields increase the discount rate applied to future earnings, which hits long-duration growth stocks—like semiconductor companies—disproportionately hard. Net overweight positioning in technology stocks fell to 18% as of July.

The Crowded-Trade Paradox

The decline in consensus from 82% to 53% does not mean fund managers suddenly turned bearish on semiconductors. It means fewer consider it the single most consensus position. The trade remains number one, but it's a weaker number one—conviction fragmenting as bond yields rise and tail risks shift.