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

Wall Street expects Navitas Semiconductor to fall 55% over the next year, with a median price target of $8 and only 22% of analysts rating it a buy. Marvell Technology is seen more favorably operationally, but its valuation has pushed the median target to $126, about 24% below the current price despite 82% buy ratings. The article is a mixed-to-bearish analyst-driven take on two AI chip stocks, suggesting caution rather than an immediate catalyst-driven move.

Analysis

The market is treating these two names very differently because it is pricing in two distinct earnings paths: Marvell is being valued as a near-term AI beneficiary whose multiple has outrun the certainty of forward growth, while Navitas is being marked as a pre-profit transformation story with execution risk still dominating the model. The second-order implication is that AI infrastructure capital is increasingly selective: investors are rewarding names with visible design wins, backlog, and margin leverage, and punishing “next cycle” beneficiaries until their revenue mix proves durable.

For Marvell, the key risk is not demand but duration of expectations. If hyperscaler spend remains strong, the stock can still re-rate higher on estimates, but at this valuation even a modest guide-down, mix shift, or normalization in AI enthusiasm could compress the multiple faster than fundamentals deteriorate. The cleaner tell over the next 1-2 quarters is whether margin expansion keeps pace with revenue growth; if it stalls, the stock becomes vulnerable to a derating even before any top-line miss.

Navitas is a classic timing mismatch: the market is discounting current revenue weakness as if it were terminal, while the bull case depends on a 2027 earnings inflection that the street is not willing to underwrite today. The contrarian angle is that the bear case may already be over-embedded if the Nvidia relationship turns into a repeatable qualification pathway rather than a one-off branding event. Still, this is a multi-quarter story, and the intermediate risk is that dilution, cash burn, or slower-than-expected customer adoption forces the company to raise capital before the operating leverage shows up.

The broader read-through is that AI power and networking ecosystems remain attractive, but investors should prefer compounders with current free cash flow over optionality names with distant payoff. In practice, that argues for buying dips in the higher-quality infrastructure leaders and treating Navitas as a venture-style position rather than a core equity holding.

Key Decisions for Investors

• Buy MRVL on a 5-10% pullback over the next 2-4 weeks; prefer a starter position now and add only if next guidance confirms continued margin expansion. Risk/reward is attractive if estimates move higher, but trim aggressively if the P/E stays elevated without upward revisions.
• Avoid initiating a full-size long in NVTS until there is evidence of customer conversion in data center revenue over the next 2-3 quarters; if you want exposure, use a small satellite position only. Treat it as a 2027 call option, not a 2026 earnings trade.
• Pairs trade: long MRVL / short NVTS for the next 6-12 months. This captures the market’s preference for current AI monetization over future optionality and should work unless NVTS delivers an earlier-than-expected inflection.
• For more tactical investors, use MRVL call spreads into any post-earnings volatility compression over the next 30-60 days rather than outright stock. This limits downside if the market decides to de-rate high-multiple AI infrastructure names.
• Set a watchlist trigger on NVTS only if the stock washes out another 15-20% while fundamentals stabilize; that would create a better asymmetry for a 2027-style re-rating. Before then, downside risk from financing or delayed adoption likely dominates.