BofA poll shows tech overweight hits highest since August 2020

BofA’s October survey found global fund managers at their largest technology overweight since August 2020, with long global semiconductors the most crowded trade (45%). But the bank’s sentiment measure fell to 6.4 from 7.0, net equity longs slipped to 44% from 49%, and investors identified a disorderly rise in bond yields (31%) and an AI bubble (23%) as leading tail risks. The 161 respondents managed $453 billion; 57% did not expect cuts to AI capital spending, while 30% did.

Analysis

Positioning makes semiconductor weakness more dangerous than the underlying news alone: crowded ownership can turn routine guidance disappointments into forced, correlated selling, while low cash leaves fewer natural buyers. The second-order exposure is not limited to chipmakers; if investors start questioning hyperscaler returns on AI spend, pressure can extend to networking, data-center power and other capex beneficiaries. But survey responses are not evidence that spending is being cut, and a crowded trade can stay crowded while earnings and orders hold up.

Near term, treat the selloff as a positioning/earnings risk, not yet a confirmed break in the AI investment cycle. Over 1–3 months, watch hyperscaler capex guidance and semiconductor order/backlog revisions; over 6–18 months, the key distinction is whether deployed capacity earns acceptable returns, not simply whether spending remains high. A renewed inflation shock or disorderly Treasury selloff would undermine duration as a hedge and could pressure both growth equities and bonds.

Contrarian positioning is not automatically a buy signal: survey-based sentiment can lag price action, and low cash can amplify further downside before it creates buying power. The missing inputs are semiconductor estimate revisions, hyperscaler utilization/returns, and current relative valuation and options-implied skew; verify these before sizing a directional short.

Key Decisions for Investors

• Consider a defined-risk 2–3 month SOXX put spread only on a failed rebound or renewed relative underperformance versus the broad market. This expresses crowded-positioning downside without open-ended short risk; cap loss at premium paid. Stand down if semiconductor estimate revisions stabilize and the sector regains relative strength.
• For a lower-conviction relative-value hedge, trim semiconductor exposure against broad-market exposure rather than shorting the entire AI complex. Reassess after the next round of hyperscaler capex guidance and chip order/backlog updates; a sustained rise in revisions would falsify the underperformance thesis.
• Do not use long Treasuries as the default hedge to this trade: the hedge works best in a growth scare, but fails if inflation or term premium drives yields higher. Track long-end yields and inflation data; reduce the hedge if yields rise alongside weakening equity breadth.
• Alert: escalate from hedge to outright sector short only if capex guidance or utilization evidence turns down and semiconductor earnings revisions follow. Without that confirmation, the survey supports tighter risk limits and staged entries, not a broad claim that the AI cycle has ended.