Morgan Stanley Flags "Clearly Overbought" Semiconductors as AI Capex Cycle Nears Inflection

On July 10, 2026, Morgan Stanley Wealth Management's Chief Investment Officer Lisa Shalett delivered a pointed caution on semiconductor equities, arguing that mounting evidence of eroding chipmaker pricing power has pushed the sector into "clearly overbought" territory. Her comments landed on the same day SK Hynix completed its record-breaking USD 26.5 billion Nasdaq listing, the largest US IPO by a foreign company, creating a striking juxtaposition between peak market enthusiasm and a high-profile Wall Street warning that the AI capital expenditure cycle may be entering an early deceleration phase.

The Core Warning: Pricing Power Is Slipping

Speaking in a Friday interview, Shalett zeroed in on a structural shift unfolding inside AI data centers. The technology stack is being redesigned, she said, with hyperscale cloud providers increasingly incorporating lower-cost, in-house custom silicon. That displacement chips away at the pricing power of merchant chipmakers who have spent the past two years capturing outsized margins on the back of supply bottlenecks. Shalett framed this as a familiar industry pattern: when supply constraints allow a narrow set of vendors, certain memory chipmakers among them, to extract excess profits, engineers respond by engineering cheaper alternatives. The implication for investors is that the extraordinary pricing leverage that powered semiconductor earnings through 2024 and 2025 may not be as durable as consensus models assume, and the margin compression could arrive faster than the market has discounted.

Valuation Stretched: SOX P/E Has Tripled Since 2022

Shalett had flagged the "clearly overbought" condition in an investment note earlier in the week, and she used the interview to reinforce it with data. Multiple gauges, from broad semiconductor ETFs to the Philadelphia Semiconductor Index (SOX), corroborate the overbought reading. Bloomberg compiled data show the SOX price-to-earnings ratio has expanded more than threefold since 2022, meaning equity prices have run far ahead of earnings growth. Such a disconnect is historically unsustainable without a corresponding acceleration in fundamental demand, and Shalett is signaling that the opposite, a moderation in demand growth, is the more probable path. For traders, a tripling of the index P/E in under four years is not merely a statistical curiosity; it represents a narrowing margin of safety that leaves the sector acutely vulnerable to any disappointment in AI spending trajectories.

Meta's Strategic Pivot Hints at Capex Reassessment

A particularly telling datapoint Shalett highlighted was Meta Platforms' recent adjustment to its AI strategy. CEO Mark Zuckerberg told an interviewer this week that he is weighing whether leasing some of Meta's AI infrastructure to external customers could generate higher value than pure internal use. Shalett read this as a signal that mega-cap technology firms are beginning to interrogate the pace, speed, and return on investment of their multi-hundred-billion-dollar capex programs, and are exploring ways to monetize infrastructure ahead of schedule. If one of the largest AI infrastructure builders is actively exploring externalization, it suggests the industry is transitioning from a build-at-any-cost mindset toward disciplined capital allocation. That behavioral shift is exactly the kind of leading indicator that often precedes a broader deceleration in aggregate spending growth.

SK Hynix IPO: Peak Enthusiasm Meets Fragile Valuations

The timing of Shalett's warning against SK Hynix's landmark Nasdaq debut is significant. The memory giant raised USD 26.5 billion, the largest foreign-company US IPO on record, underscoring that capital flows into the AI theme remain abundant. Yet on its home Korean market, SK Hynix shares have dropped roughly 26% from their recent peak, a stark reminder that abundant capital and durable valuations are not the same thing. Shalett acknowledged that money chasing the AI theme is still plentiful but insisted the structural dynamics, custom silicon adoption and cost engineering, are what ultimately determine whether premium pricing holds. The IPO thus serves as a microcosm of the broader tension in the sector: headline demand is strong, but the marginal buyer is increasingly price-sensitive and the underlying margin story is deteriorating at the edges.

"Early Innings" of AI Capex Deceleration

Shalett's most consequential framing was her conclusion that we are in the "early innings" of a deceleration in AI capital expenditure growth. This is a nuanced call: it does not predict an immediate collapse in spending, but rather a slowdown in the rate of growth. In the early phase of such a deceleration, reported earnings often still look robust because spending lags commitments, but sentiment and multiple compression can lead fundamentals lower. The analogy to a baseball game's early innings implies there is considerable runway ahead, and that the most acute repricing may still be in front of the market. For positioning, this favors a shift from indiscriminate AI-beta exposure toward quality names with defensible moats, and a careful reassessment of leveraged bets on the most extended semiconductor valuations.

Cross-Asset Ripple Effects for Crypto Markets

While Shalett's analysis is equity-centric, the implications cross into crypto. AI-themed digital assets, decentralized GPU compute networks, and tokens tied to AI inference demand have traded with meaningful correlation to the semiconductor and AI capex cycle. A sentiment adjustment in chips can spill into these niches quickly, sometimes with amplified volatility given their thinner liquidity. Conversely, any rotation away from crowded AI equity exposure can, in some regimes, redirect marginal capital into alternative stores of value such as Bitcoin. The net effect is ambiguous and regime-dependent, which is precisely why crypto traders should treat the semiconductor overbought signal as a leading risk indicator rather than a binary directive. Prudent steps include monitoring SOX and semiconductor ETF price action for breakdowns, stress-testing AI-token correlations, and ensuring portfolio leverage is sized for a regime where dispersion, not uniform upside, becomes the dominant feature.