Celestica's AI Portfolio Push Confirms Growth but Ignores Valuation Excess

What happened

Celestica is touting its high-bandwidth AI portfolio, including 400G to 1.6T Ethernet switches, as critical for powering energy-efficient AI data centers amid surging hyperscaler demand. This aligns with the company's strategic pivot to Hardware Platform Solutions (HPS) and communications, driving 23% year-to-date revenue growth in 2025. However, the stock trades at approximately 47 times trailing earnings and 45 times EV/EBITDA, reflecting a 724% premium over a discounted cash flow valuation anchor. The DeepValue report underscores persistent risks like high customer concentration, with top ten customers accounting for 80% of Q3-25 revenue, and cyclicality in the low-moat electronics manufacturing services industry. Thus, while the AI focus validates operational momentum, it does not address the overvaluation or structural vulnerabilities that support a 'POTENTIAL SELL' stance.

Implication

The news highlights Celestica's leverage to secular AI data center trends, which could sustain revenue growth and margin expansion in the communications segment. However, the stock's rich multiples, at ~47x P/E, imply near-perfect execution is priced in, leaving no margin of safety for any missteps. High customer dependence—with 80% of revenue from the top ten—amplifies downside risk from hyperscaler capex cuts or competitive insourcing. Solid free cash flow and modest leverage offer some financial cushion, but fail to justify the premium in a cyclical, structurally competitive industry. For conservative investors, this reinforces the need to wait for a material derating towards normalized cash flow multiples before considering an entry.

Thesis delta

The news confirms Celestica's growth narrative in AI and data centers, aligning with HPS expansion trends noted in the DeepValue report. However, it does not mitigate core risks such as overvaluation, customer concentration, or industry cyclicality. Therefore, the 'POTENTIAL SELL' recommendation remains unchanged, as the growth validation is insufficient to offset the lack of margin of safety.