Arista Networks Vs. Celestica: The White Box Wins

Arista Networks expects AI-driven growth to remain robust but guided for gross margins to stay flat at 63% through CY27 as supply-chain cost inflation pressures profitability. The company is prioritizing supply security rather than passing higher costs to customers, limiting margin upside versus Celestica. Celestica is viewed as better positioned for margin expansion and faster revenue growth from rising white-box Ethernet-switch demand while trading at lower valuation multiples.

Analysis

ANET's issue is not AI Ethernet demand but incremental-profit conversion: a prolonged gross-margin ceiling shifts valuation support from multiple expansion toward revenue upside and operating-expense discipline. If component costs remain elevated while hyperscaler customers retain purchasing leverage, each upside revenue revision is worth less to EPS than in prior AI-networking cycles. The key near-term question is whether ANET can offset gross-margin pressure through mix, software/services attachment, or opex leverage; absent that, consensus EPS estimates for CY26-27 likely remain vulnerable despite healthy order activity.

CLS has a more favorable operating setup if white-box adoption broadens beyond isolated hyperscaler programs. Greater disaggregation transfers value from branded network-system vendors toward manufacturing, integration, and supply-chain execution; that can create a multi-quarter revenue and utilization uplift for CLS while placing price pressure on ANET's hardware portfolio. This is not a zero-sum outcome initially, since overall AI cluster networking spend can grow, but the competitive signal becomes material if hyperscaler capex growth no longer produces commensurate ANET gross-profit growth.

The contrarian view is that supply assurance is economically rational rather than purely margin-destructive: network downtime and deployment delays carry far larger costs for AI customers than modest switch pricing. ANET could regain pricing power if advanced Ethernet supply tightens or if customers prioritize validated low-latency fabrics over lower-cost white-box systems. The thesis is falsified if ANET's gross-margin outlook improves without a demand slowdown, or if CLS fails to translate white-box volume into sustained operating-margin and free-cash-flow expansion over the next two earnings cycles.

Key Decisions for Investors

• Establish a 3-6 month market-neutral pair: long CLS / short ANET, sized by beta rather than equal dollars. The expected payoff is relative multiple and estimate dispersion as CLS captures white-box volume while ANET's CY26-27 EPS upside is capped by gross-margin assumptions; exit if ANET raises its medium-term gross-margin framework or CLS reports weak program ramps.
• For long-only technology exposure, underweight ANET into the next earnings report unless channel checks show component-cost relief or higher software/services mix. Reassess after management provides quarterly gross-margin trajectory rather than relying on annual revenue growth, since margin stabilization is the gating variable for renewed multiple expansion.
• Use ANET call spreads only after confirmation that gross margin has bottomed: buy 6-9 month upside spreads rather than outright calls to limit exposure to multiple compression. A credible catalyst would be sequential gross-margin expansion paired with unchanged hyperscaler demand; without that evidence, AI-networking upside is better expressed through CLS.
• Monitor hyperscaler Ethernet procurement awards, merchant-silicon availability, and CLS backlog-to-revenue conversion over the next 1-3 months. If white-box deployments remain limited to pilots or CLS utilization does not improve, reduce the pair because the relative thesis depends on verified volume conversion rather than announced AI demand.