Celestica vs. Corning: Which AI Infrastructure Stock is a Better Buy?

Celestica is presented as the stronger AI-infrastructure investment, supported by projected 2026 sales growth of 64.27% and EPS growth of 87.11%, versus Corning's expected 15.78% sales and 29.76% EPS growth. CLS is benefiting from 800G Ethernet demand, 1.6T switch ramps and AI-compute programs with OpenAI, Broadcom and AMD, while trading at 20.04x forward earnings versus Corning's 42.46x. Corning has major hyperscaler agreements and gained 124.8% over the past year, but faces optical-networking competition, consumer-electronics cyclicality and a $7 million Q2 2026 solar loss after roughly $30 million of transition-related costs.

Analysis

CLS offers the cleaner AI-infrastructure operating leverage: its value-add is concentrated in system integration and network/compute assembly, where program ramps can drive both revenue mix and factory-utilization upside. The key second-order issue is customer concentration and working-capital intensity—large hyperscale ramps may inflate receivables and inventory before cash conversion catches up. The market should reward CLS if gross margin holds while revenue scales; if margins flatten despite growth, it will be treated as a lower-multiple EMS rather than an AI-platform beneficiary.

GLW has a more diversified earnings base, but that diversification dilutes its AI read-through and leaves the stock exposed to unrelated consumer-device and solar execution variables. Its optical opportunity is also not winner-take-all: APH is better positioned in higher-value interconnect assemblies and CIEN captures a more direct networking-equipment spend cycle. For GLW, the relevant catalyst is not additional AI partnership announcements but evidence that optical content per deployed compute cluster rises fast enough to offset pricing pressure and the drag from non-data-center businesses.

The valuation gap makes CLS the preferred relative-value expression over the next 1-3 months, but it is not a riskless long after a strong estimate-revision cycle. Consensus may be underestimating the duration of 1.6T and custom-rack deployment, while simultaneously overestimating near-term margin expansion from those ramps. Over 6-18 months, optical architectures moving closer to the package could be structurally positive for GLW, but commercial adoption timing remains uncertain and is unlikely to justify paying for that optionality before design-win-to-revenue conversion is visible.

Key Decisions for Investors

• Initiate a 1-3 month pair trade: long CLS / short GLW in equal dollar amounts. Thesis is superior AI revenue sensitivity and lower multiple risk at CLS versus GLW's broader cyclical and execution exposure. Target 10-15% relative outperformance; exit if CLS reports sequential gross-margin compression or working capital consumes the majority of operating cash flow.
• Add CLS only around earnings or supply-chain volatility rather than chase momentum. Monitor bookings, backlog conversion, customer concentration, and 1.6T program revenue contribution; a guide implying AI-program delays beyond two quarters falsifies the near-term rerating thesis.
• Maintain APH as the higher-quality optical/interconnect alternative to GLW for a 6-18 month AI connectivity allocation. Prefer GLW only if management demonstrates sustained Enterprise Networks margin expansion and solar losses normalize; otherwise its multiple leaves limited tolerance for a handset or solar miss.
• Use AMD and AVGO as confirmation indicators for CLS: upward accelerator-platform demand commentary or raised rack-scale deployment expectations supports the long. Conversely, any hyperscaler capex pause, custom-silicon qualification delay, or networking inventory digestion would likely hit CLS earlier and harder than diversified infrastructure peers.