Wall Street Says These 2 AI Stocks Could Fall 20% or More. Time to Sell?

Wall Street sees downside of about 24% for Marvell Technology to a $126 median target and 55% for Navitas Semiconductor to an $8 median target, reflecting caution despite both companies’ AI exposure. Marvell has surged 95% year to date on 43% annual revenue growth, strong guidance, and a $2 billion Nvidia investment, but its 56 P/E valuation is seen as stretched. Navitas is more challenged near term as its pivot away from consumer chips toward data centers has दबressed revenue, though analysts expect a possible earnings inflection in 2027.

Analysis

The market is separating the AI infrastructure stack into two very different trade types: a quality compounder with near-term multiple compression risk, and a transition story with optionality but no earnings visibility. MRVL sits in the dangerous zone where fundamentals are strong enough to attract momentum capital, but valuation is now implying flawless execution through at least the next several quarters; that makes it vulnerable to any guide-down, margin noise, or simply a pause in AI capex ordering. NVTS, by contrast, is a classic “future winner, current loser” setup where the equity can stay broken longer than the business thesis looks logical, because the re-rating depends on design-win conversion and meaningful revenue inflection that likely won’t be visible for multiple quarters.

Second-order effects favor the incumbent ecosystem more than the names highlighted. If Marvell’s AI exposure keeps compounding, suppliers tied to optical, advanced packaging, and high-speed interconnect should see a broader halo, while weaker networking peers face a tougher bar for differentiation. For NVTS, the pivot into data center power conversion is strategically attractive, but it also puts the company in a crowded field where share gains depend on qualification cycles and customer concentration; that means the upside case is highly path-dependent and can be delayed by one or two large platform decisions.

The contrarian read is that analyst targets likely understate the duration of AI infrastructure spending, but overstate how quickly that spending translates into linear upside for smaller names. MRVL’s downside is more about multiple mean reversion than business deterioration, which usually argues for buying on a 10-15% pullback rather than chasing strength. NVTS may be investable only as a staged position or call-option proxy into 2027, because the stock can easily re-rate lower again before the data center story becomes visible in reported numbers.

Near term, the real trading risk is not that AI demand disappears; it is that expectations get too far ahead of booking cadence and gross margin leverage. That creates a window where MRVL can correct 15-25% even if the fundamental story remains intact, while NVTS could drift lower or go sideways until investors get proof of conversion. In this tape, patience should be used as a timing tool: the better entry is likely after the first post-earnings reset, not on the headline optimism.

Key Decisions for Investors

• MRVL: avoid initiating new longs at current levels; use a 10-15% pullback or a post-earnings reset to build a position, with a 6-12 month horizon and ~2:1 upside/downside if AI demand remains intact.
• MRVL: consider a short-dated put spread or collar into earnings if implied volatility is cheap relative to the risk of a valuation reset; structure for a 15-20% correction rather than a fundamental collapse.
• NVTS: treat as a staged optionality trade only, not a core equity long; small starter position or long-dated call options for 2027 upside, but size it assuming 12-18 months of dead money risk.
• Pair trade: long MRVL / short a lower-quality AI hardware name with weaker fundamentals if you want to express AI infrastructure leadership without broad beta; the relative-value edge is in execution quality, not sector direction.
• If you want exposure to the Nvidia ecosystem without paying full multiple for MRVL, wait for a market-wide AI shakeout and rotate into semis on weakness rather than chasing the current momentum window.