MRVL's Networking Business Grows on the Back of AI: What's Ahead?

Marvell said demand for its 800G optical DSPs remained strong in fiscal Q2 2027 and its 1.6T optical DSP business ramped rapidly; management expects 1.6T adoption to accelerate further in fiscal 2028. The company remains on track to more than double fiscal 2027 scale-out switching revenue, while consensus earnings estimates imply growth of 48% in fiscal 2027 and 76% in fiscal 2028. Shares were up 88.1% year to date versus 11.4% for the semiconductor industry, but Marvell trades at 8.58x forward sales versus the industry's 3.14x average.

Analysis

The key question is not whether AI networking demand grows, but whether Marvell converts that demand into durable, high-value content before hyperscalers optimize around cost, power, and competing architectures. Optical adoption is not a straight-line call: copper can defer optical substitution inside clusters, while scale-across links may support optical demand independently. If copper reach improves or cluster build-outs pause, the timing of the 1.6T ramp—not the long-term bandwidth thesis—becomes the near-term earnings risk.

Marvell’s sharp outperformance and premium sales multiple leave less room for execution slippage. The unchanged FY27 earnings estimate alongside only recent FY28 upward revisions suggests the stock is pricing a longer-duration acceleration; management targets are not yet proof of realized revenue or margin. Broadcom may benefit from custom silicon programs even if Marvell captures connectivity content, so treat them as adjacent exposures rather than a clean winner/loser pair. AMD’s accelerator adoption could expand the networking pool, but does not establish that Marvell wins the associated attach.

Over days, positioning and guidance credibility likely dominate. Over 1–3 months, verify reported networking growth and FY28 estimate breadth; over 6–18 months, watch 1.6T deployments, optical penetration versus copper, and customer diversification. Thesis weakens if shipment/ramp commentary slips, FY28 revisions reverse, or networking growth fails to translate into company-level guidance. A durable structural case requires realized customer deployments, not just an addressable-market narrative.

Key Decisions for Investors

• Avoid chasing MRVL after its substantial run; hold existing exposure only with a defined review around the next results and guidance update. The risk/reward is increasingly dependent on execution against elevated expectations.
• Consider a modest relative underweight in MRVL versus a broad semiconductor benchmark rather than a direct MRVL/AVGO pair: MRVL’s stated sales-multiple premium and rally create asymmetric disappointment risk, while the businesses are not interchangeable. Reassess if estimate revisions broaden and reported networking growth confirms the ramp.
• Set an alert for 1.6T revenue realization, scale-out switching customer breadth, networking-related guidance, and FY28 estimate revisions. Missing data to verify include customer concentration, product-level revenue and margins, and whether management targets convert to recognized shipments.
• Falsify the cautious stance if consecutive updates confirm broad-based deployments and improving estimates without margin deterioration; strengthen it if ramp timing slips, guidance is cut, or copper adoption delays optical conversion.