Data Centers Need Faster Connections. These 6 Optical and Networking Stocks Could Benefit.

AI data-center build-out is driving strong demand for high-speed networking, optical connectivity and specialized chips. Credo reported fiscal Q1 2027 revenue of $479 million, up 114% year over year; Astera Labs more than doubled revenue in Q2 2026 and said Scorpio switches would become its largest product family one quarter earlier than previously expected. Lumentum's fiscal 2026 revenue rose 83% to $3.01 billion, while the article identifies Arista Networks, Coherent and Marvell as additional suppliers positioned to benefit from the connectivity demand.

Analysis

The investable question is not whether AI needs more bandwidth, but which connection layers get funded as architectures evolve. Spending can migrate between in-rack electrical links, switches, and optical links; a copper-to-optical transition is therefore not a uniform win for all six. Short-reach electrical products can remain complementary even as optical gains distance and bandwidth share, while faster links may simply move the bottleneck to switching, power, or deployment capacity.

The strongest second-order risk is timing: a small number of large infrastructure buyers can pause or redesign deployments, leaving component suppliers with inventory and utilization pressure before the longer-term demand thesis breaks. Rapid sales growth cited for individual products is not proof of durable margins, cash conversion, or repeat orders. Verify customer concentration, backlog conversion, product mix, gross-margin trajectory, and any evidence of dual-sourcing before paying for growth.

At days-to-weeks horizons, the basket is vulnerable to positioning and hyperscaler capex guidance. Over 1–3 months, earnings evidence on orders and margins matters more than architecture narratives. Over 6–18 months, optical adoption and custom network designs could reshuffle value among suppliers; competition and customer bargaining may prevent infrastructure growth from translating proportionally into supplier profits. The contrarian risk is treating bandwidth demand as a straight-line revenue forecast: more efficient networking can reduce required equipment per unit of compute, and AI spending can be deferred even if the need remains real.

Key Decisions for Investors

• Do not chase the six-name theme as an equal-weight basket. Consider a staged, modest long in ANET and MRVL on broad semiconductor/networking weakness; add only if upcoming results confirm order conversion and stable margins. This favors exposure to switching and connectivity silicon without assuming every high-growth component supplier sustains its current trajectory.
• Keep CRDO, ALAB, COHR, and LITE on a catalyst watchlist rather than treating reported growth or management optimism as sufficient confirmation. Before initiating, verify customer concentration, backlog or design-win conversion, inventory, and product-level margins; these are key missing data for assessing durability and downside.
• Falsify the bullish thesis if major cloud providers cut or defer AI infrastructure plans, or if relevant suppliers report weakening orders, rising inventory, or material margin deterioration. Reassess the staged long rather than averaging down; architecture substitution could also shift spend away from a specific connection layer without ending overall AI investment.