Some of the Year's Hottest AI ETFs Are in Big Trouble Today

The Nasdaq 100 fell more than 10% from its October closing high, entering correction territory, as chip, AI infrastructure, and data-center stocks sold off. The Roundhill Memory ETF dropped about 15% Thursday, while leveraged semiconductor ETF SOXL fell more than 30%; SOXX turned negative year to date and WTAI neared wiping out its 2026 gains. Investors cited election-related concerns over AI spending, but commentators also pointed to forced selling and uncertainty after technology earnings, with the longer-term outlook for data-center development remaining debated.

Analysis

The selloff looks more like a positioning and confidence shock than evidence, by itself, that AI infrastructure demand has rolled over. Forced de-risking can push high-beta names below levels justified by near-term earnings, but it also raises the bar for every capex announcement: spending without visible utilization or customer commitments may no longer reassure investors.

The important second-order risk is project economics, not a blanket ban on data centers. Local opposition can lengthen permitting and require developers to fund power, safeguards, or community measures. That can delay orders and shift costs even if projects proceed. VRT is more directly exposed to data-center power and cooling deployments; PWR may have a relative buffer if grid investment broadens beyond data centers, though project timing remains a risk. Memory and storage names (MU, SNDK, STX, WDC) add a separate cyclical supply/pricing exposure; a weaker AI buildout could compound any industry-specific inventory or pricing pressure. AMD’s exposure is not interchangeable with those component cycles.

Days: flow-driven volatility and leveraged ETF decay can exaggerate losses. Over 1–3 months, monitor hyperscaler capex guidance, order/backlog trends, project cancellations, and permitting or power-contract developments. Over 6–18 months, the key question is whether mitigation requirements raise costs and delay returns enough to redirect investment. The political narrative alone does not establish that outcome. A thesis of durable demand is weakened by capex cuts, cancellations, or deteriorating orders; it is strengthened by sustained commitments and improving project visibility.

Key Decisions for Investors

• Avoid chasing the selloff with SOXL or other daily leveraged long exposure: volatility and daily reset effects can erode returns even if the underlying sector later rebounds.
• Keep a watchlist, not an automatic dip-buy, in AMD, MU, SNDK, STX, and WDC. Reassess after company guidance and industry pricing/inventory data clarify whether the move is flow-led or earnings-led.
• Consider a relative-value expression—long PWR versus short VRT—only if data show grid work broadening while data-center-specific orders or project timing weaken. The thesis is falsified if PWR’s own backlog/order outlook deteriorates or VRT’s demand remains resilient; absent that evidence, no pair trade is warranted.
• Track hyperscaler capex commitments, data-center cancellations and permitting/power agreements over the next 1–3 months. A reversal in commitments would argue for reducing exposure; stable spend plus improving project visibility would support staged re-entry rather than buying solely on the correction.