Marvell Technology Has Soared 247% in 2026. Here's How Much Upside It Can Deliver Over the Next 3 Years

Marvell Technology has surged 247% this year as investors price in accelerating demand from AI infrastructure, with management targeting 40% revenue growth in fiscal 2027 to $11.5 billion. The company expects DCI optical revenue to double to $1 billion by fiscal 2028 and custom silicon revenue to rise by more than 100% next year, supporting the case for further upside despite a rich valuation. The article argues Marvell could still outperform over the next three years if AI-related growth exceeds expectations.

Analysis

MRVL is effectively a levered call on two scarce nodes in AI infrastructure: custom silicon design wins and optical bandwidth. The market is still underwriting this as a single-name hardware rerating story, but the better read is that Marvell is becoming a “picks and shovels” bottleneck supplier to hyperscalers that want to reduce inference cost and move data faster inside the rack and across the cluster. That makes the earnings profile more durable than a typical chip cycle, but also more lumpy because revenue will likely be gated by a small number of very large program ramps.

The second-order effect is on competitive positioning. If custom ASIC adoption keeps taking share from merchant accelerators in some workloads, the winners are not just MRVL and the hyperscalers; they are also network and optical vendors with the right power, signal integrity, and co-packaging roadmap. That pulls forward demand for the whole interconnect stack, while pressuring companies that remain exposed to generic AI compute pricing where ASPs are more vulnerable to commoditization. In other words, the hidden beneficiary is the infrastructure layer around compute, not just compute itself.

The main risk is not valuation in isolation; it is execution against very rich expectations over the next 2-3 quarters. At this multiple, any delay in customer qualification, a slower-than-expected revenue conversion curve, or evidence that hyperscalers are reallocating capex toward in-house silicon less aggressively than assumed could compress the stock quickly. Conversely, the setup becomes self-reinforcing if the company prints upward guidance revisions tied to a visible ramp in optical and switching orders, because that would validate the market’s assumption of sustained scarcity value.

Consensus may be underestimating how much of the upside is already embedded in the ‘AI beneficiary’ label, but may also be underestimating the duration of the demand wave if optical remains the binding constraint. The stock likely has more upside over 12-24 months if management can prove a multi-year backlog conversion story; the cleanest inflection will be evidence that these programs are broadening beyond a handful of anchor customers. Until then, this is a high-quality momentum name, but one that should be traded around rather than blindly accumulated into strength.

Key Decisions for Investors

• Buy MRVL on a pullback of 8-12% from recent highs; target a 6-9 month hold with upside tied to the next two guide/reports, but use a hard stop if management signals program slippage or margin pressure.
• Express the AI infra thesis via a pair: long MRVL / short a higher-beta merchant compute name over 3-6 months, betting that bandwidth and custom silicon scarcity outlasts generic accelerator enthusiasm.
• For investors unwilling to chase spot, buy 6-9 month MRVL call spreads to capture a continued re-rating while defining downside; best risk/reward if implied vol remains elevated after the recent rally.
• Monitor AVGO and NVDA as relative-victory hedges: if MRVL strength is driven by custom ASIC and interconnect share gains, any weakness in peers on the same theme is a signal to rotate rather than add to the basket.
• If MRVL gaps higher on results, take partial profits into the move and roll into longer-dated exposure; at current multiples, the stock is more sensitive to guidance credibility than to headline growth alone.