ADI's Communications Segment Surges on AI Demand: What's Ahead?

Analog Devices' fiscal Q3 2026 Communications revenue rose 84% year over year and 18% sequentially to $655 million, with data-center sales—80% of segment revenue—more than doubling in both optical and power applications. ADI expects optical circuit-switching revenue to roughly double in fiscal 2026 and again in fiscal 2027, supported by AI infrastructure demand and the Empower Semiconductor acquisition. Fiscal 2026 consensus EPS implies 66.4% year-over-year growth and has been revised upward over the past 30 days, though ADI trades at a premium 10.17x forward P/S versus the industry's 8.00x.

Analysis

ADI’s AI exposure is becoming economically material enough to alter its historical analog-cycle valuation framework, but the market should distinguish high-quality content growth from durable share capture. Power-management content is likely the more defensible profit pool: higher rack power density raises qualification barriers and makes design wins sticky across server generations. Optical circuit switching has greater upside optionality but carries architecture risk, particularly if co-packaged optics or custom hyperscaler designs reduce the addressable market for merchant components.

Over the next 1-3 months, the relevant catalyst is whether management converts design activity into booked backlog and raises the data-center revenue trajectory without relying on wireless normalization. ADI’s premium valuation leaves little tolerance for a sequential deceleration, gross-margin dilution from acquired power products, or evidence that growth is concentrated in one or two hyperscaler programs. The key falsifier is a weaker-than-expected Communications outlook or a failure of data-center mix to lift consolidated operating leverage at the next earnings release.

MRVL is the cleaner high-beta expression of optical bandwidth expansion, while ALAB is more exposed to rack-scale interconnect and PCIe/CXL adoption. A non-obvious competitive outcome is that accelerating AI power budgets can favor ADI even if optical standards fragment: every architecture still requires increasingly localized conversion and regulation, whereas connectivity vendors face protocol and packaging transitions. Conversely, the consensus may be underpricing supply-chain concentration and qualification cycles; optical-switching revenue targets are management objectives, not independently verified demand commitments.

The contrarian view is that ADI may be a better relative-value long than a standalone momentum long. Its industrial and automotive recovery can provide earnings support if AI infrastructure spending pauses, while pure-play connectivity multiples are more vulnerable to a hyperscaler capex digestion episode. That defensive characteristic merits owning ADI against a richer AI-connectivity basket rather than chasing it after strength.

Key Decisions for Investors

• Initiate a 6-12 month pair: long ADI / short an equal-dollar basket of MRVL and ALAB. Target 10-15% relative upside if power-content growth remains durable while optical/interconnect multiples normalize; exit if ADI’s Communications growth decelerates sharply for two consecutive quarters or gross-margin guidance deteriorates.
• Do not add outright ADI exposure ahead of the next earnings print at an elevated sales multiple. Accumulate only on a 8-12% pullback or after verification that data-center bookings/backlog support the following two quarters; use a 10% stop from entry absent an upward revision to consolidated earnings power.
• Maintain MRVL as the higher-beta bandwidth-upgrade vehicle only if hyperscaler capex guidance remains intact. A reduction in cloud capex plans, delayed 1.6T deployment, or slower scale-out revenue conversion would warrant reducing MRVL before ADI, given MRVL’s greater optical-cycle sensitivity.
• Treat ALAB as a watch item rather than a fresh recommendation until PCIe 6/CXL revenue mix and customer concentration are disclosed with greater clarity. A sustained ramp in Scorpio and evidence of multi-customer adoption would validate a 12-18 month long; an isolated program ramp would not.
• Ignore QBTS for this theme: there is no demonstrated revenue linkage between AI data-center optical/power spending and quantum-computing demand, making any sympathy move non-fundamental.