This Under-the-Radar Semiconductor Stock Is Beating Nvidia and Broadcom in 2026

Navitas Semiconductor’s stock has surged in 2026 after launching a DC-DC power delivery board in March that converts 800 volts to 6 volts in one stage, aimed at improving efficiency for Nvidia-linked AI data centers. Morgan Stanley raised its price target to $12.50 from $4.20 and Baird to $20 from $9 in May, reinforcing bullish sentiment. However, the article notes the stock trades at 92x sales versus a five-year average P/S of 11.8, suggesting elevated expectations and limited near-term upside.

Analysis

The important signal here is not that a small power-optimization vendor has a good product; it is that the AI infrastructure spend chain is widening from compute chips into every bottleneck around power delivery. That tends to create a second-order winner set in “picks-and-shovels of the picks-and-shovels”: if hyperscalers standardize around higher-voltage rack architectures, the value accrues first to enabling component vendors, then to integrators, and only later to broader semiconductor exposure. In that sense, the rally is less about near-term revenue quality and more about a narrative re-rating that can run ahead of fundamentals for 1-2 quarters.

The market is likely underestimating how quickly multiple compression can happen once a stock becomes a consensus AI proxy. The extreme sales multiple implies the equity is now priced like a de facto strategic asset, so any evidence of design-win slippage, slower conversion from evaluation to production, or a competing power-delivery architecture from larger incumbent suppliers could trigger a sharp drawdown even if top-line growth remains strong. The risk is asymmetry: upside from here is driven by multiple expansion on already-lofty expectations, while downside can be 30-50% on a single missed milestone or a broader risk-off rotation in AI beneficiaries.

For the larger AI names, this is mildly supportive for the ecosystem but not a direct fundamental inflection. The better expression is via relative value: the market is increasingly paying for scarcity in AI infrastructure enablers, which can temporarily lift adjacent names and small-cap secondaries, but it also raises the bar for the incumbent leaders to justify their own AI capex trajectories. If the power-efficiency thesis gains traction, suppliers with exposure to data-center electrical infrastructure may see a multi-month sentiment tailwind even before revenue contribution is material.

The contrarian takeaway is that the market may be extrapolating one product announcement into a durable platform position. In practice, AI power architecture is likely to be winner-take-most only after standards settle, and that process usually takes longer than momentum investors expect. Consensus is missing timing risk: the story may be right structurally, but the stock can still be a bad entry here because the discount rate is being overwhelmed by narrative premium.

Key Decisions for Investors

• Avoid initiating a new long in NVTS at current levels; if long already, trim into strength and look to re-enter only after a 20-30% pullback or after a confirmed production-design-win update over the next 1-2 quarters.
• Express the theme with lower downside: buy NVTS call spreads 3-6 months out rather than stock, targeting a 1.5-2.0x payoff if the market continues to re-rate AI power enablers while capping premium paid.
• Pair trade: long AVGO / short NVTS for 2-4 months. AVGO offers AI exposure with materially lower multiple risk and broader monetization, while NVTS carries the highest narrative-to-fundamental gap; reward on the spread is a normalization of relative valuation.
• For event-driven traders, watch for analyst revisions and customer qualification updates as the next catalyst window; upside works only if the company converts product visibility into order flow within a single budget cycle, not just a story-stock rerating.
• Keep a tactical long bias in NVDA as the ecosystem beneficiary, but treat the move as sentiment-positive rather than revenue-revising; size small and use any broad tech selloff to accumulate, not chase.