Celestica Inc. (CLS)

Priced as of 2026-09-11 — market data goes stale within days; treat this as a starting point, not a current quote.

Celestica's Sum-of-the-Parts (SOTP) central fair value works out to **$263.74 per share**, about 24% below today's $346.55. The reason is not that the switch business is weak — it is that the market appears to be paying a near-duopoly growth multiple across the *whole* company, while only about 56% of Celestica's revenue actually sits in that near-duopoly. The other 44% is a small-follower Artificial Intelligence/Machine Learning (AI/ML) server business that the Stage 7 Job Audit called "Blocked," plus a Dog segment. Two specific findings drive the gap. First, a metric cross-check required by the new rules shows that borrowing Accton Technology's and Fabrinet's revenue multiples wholesale does not work: Celestica's Connectivity & Cloud Solutions (CCS) segment earns an 8.61% segment margin while Accton earns a 13.68% operating margin — a revenue multiple carries margin the business does not have. Second, the AI/ML server slice grew from 20.9% of CCS revenue in Q2 2025 to 30.4% in Q2 2026, so the Stage 4 warning that the mix could shift toward the weak position is actively happening. The honest caveat: the price sits **inside** the fair-value band, not above it, so this method does not fully resolve Celestica — conviction is capped at Moderate.

Fair-value band

Rather than a single point estimate, this valuation reports a range: every discretionary judgment call set to its defensible low end, its midpoint, and its high end.

Price sits inside the band — the method does not resolve this name. The decision below comes from the central estimate, but the honest read is that it depends on assumptions this analysis could not settle.

Portfolio at a glance

Key assumptions & swing factors

• Rule B, discount once not twice: the Job Audit's pessimism was applied through comp-set selection ONLY, and no haircut was applied to any multiple. The switches slice uses Accton Technology, which sells the same white-box switches to the same hyperscalers under the same Ethernet-generation re-qualification cycle and therefore carries the identical Contingent 'Star on borrowed time' dependency in its own market price; the servers slice uses follower and AI-server ODM multiples (Quanta, Sanmina, Jabil) rather than growth-leader multiples, which is the Blocked verdict expressed once. The conservative case drops Fabrinet, the one comp that is not equally contingent, which is comp-set selection at the low end rather than an invented haircut.
• The Rule A cross-check rejected the raw growth-peer revenue midpoint of 3.145x for the switches slice: translated onto Celestica's disclosed 8.61% CCS segment margin it implies 36.5x EV/operating-income, above the 24.0x-30.2x peer range (Accton 24.0x, Fabrinet 30.2x). The range was narrowed per Rule A step 2 to 2.33x. This is a metric-basis correction under Rule B2, NOT a second Job Audit discount.
• Rule B2 basis flag: Celestica's segment income is non-GAAP (it excludes stock-based compensation and amortisation of intangible assets) while the peer operating margins used in the cross-check are GAAP. The mismatch flatters Celestica, so the cross-check's rejection of 3.145x is a floor, not a ceiling.
• Largest swing factor: Celestica does not disclose segment margin by end market, so the Communications slice is valued on the CCS blended 8.61% margin. If that slice actually earns 10% (plausible given Hardware Platform Solutions revenue of roughly $1.9bn in Q2 2026), the central multiple becomes 2.71x and fair value becomes $298.45 per share, a gap of -13.9% — still AVOID but materially narrower.
• Rule C, freshest base: the central case runs on Q2 2026 annualised run-rate revenue ($18.79bn total), not FY2025. The conservative case uses trailing-twelve-month revenue to June 30 2026 ($15.59bn) and the generous case uses the raised FY2026 company outlook of $20.5bn (Celestica, July 27 2026). On the FY2026 guidance base at central multiples fair value is $291.78 per share, a gap of -15.8% — the decision does not flip on the base choice.
• Inherited Stage 4 flag: CCS's Star label rests on a revenue-weighted average across two internally divergent share positions (~90% combined share with Accton in white-box switches versus a small-follower position inside an 11.5%-combined bucket in AI/ML servers). This valuation prices the two slices separately rather than blending them, but the weighting is itself inherited. The mix is actively shifting toward the weak slice: Enterprise was 20.9% of CCS revenue in Q2 2025 and 30.4% in Q2 2026.
• Inherited Stage 4 flag: ATS's Dog label is rank-based with no sourced share percentage for Celestica or any competitor in any sub-vertical. ATS is 8.5% of central Total Fair EV, so the exposure is small. Stage 7 found Aerospace & Defense to be a well-defended Niche keeper, which is why the Kimball liquidation multiple is used as the floor endpoint of a range rather than as the answer — expressed once, not added as an uplift.
• ATS cross-check pushed UP, not down: ATS's 6.16% segment margin exceeds every mature EMS peer's operating margin (Kimball 3.97%, Benchmark 3.74%, Jabil 5.06%, Sanmina 5.49%), and the peer EV/operating-income midpoint of ~19.1x implies 1.17x EV/Revenue against the 0.74x revenue midpoint used. Central was held at 0.74x because the Dogs rule caps a Dog below the sector average and only one of three ATS sub-verticals is a confirmed defended niche; the cross-check is carried into the generous case at 1.00x.
• Conviction is capped at Moderate for two independent reasons: the price sits inside the Rule D band rather than above its generous end, and the largest single SOTP component (the Communications slice, 80% of central Total Fair EV) rests on a Stage 4 borderline call.
• SOTP sanity check, gap named not closed: market EV of $40.5bn is 31% above central Total Fair EV of $30.9bn. Holding the server and ATS slices at central multiples, today's price implies 3.24x EV/Revenue for the Communications slice — essentially Accton's own 3.28x — despite Celestica's CCS earning 8.61% (non-GAAP) against Accton's 13.68% GAAP operating margin. The market appears to be pricing all of Celestica as if it were the near-duopoly switch business. No multiple was adjusted to close this gap.
• No placeholder valuations were used; every SBU rests on researched, sourced figures.
• This decision reflects market prices as of September 11, 2026 and should be re-run, not assumed current, before acting on it. Celestica reports Q3 2026 results and holds its Investor and Analyst Day on October 27, 2026, which may resolve the end-market margin disclosure that is this valuation's largest gap.

Portfolio-level tensions

• Shared-factory-capacity risk (unconfirmed but plausible): CCS discloses detailed facility investment (Dallas-Fort Worth, Thailand) while ATS discloses no dedicated footprint, suggesting ATS may draw on shared corporate manufacturing capacity that CCS's hyperscaler ramp could compete for
• Capital-allocation question: with no Cash Cow, it is unresolved in company disclosures whether ATS's harvest-driven margin gains (from A&D portfolio pruning) should be earmarked toward closing CCS's Blocked AI/ML-server capability gap, or reinvested in ATS's own growing corners (Capital Equipment, Industrial, HealthTech)
• One meter design across the whole portfolio: all five scored jobs are paid the same way — a fee charged once per unit built, with only the switches job (via HPS) partially capturing extra design-IP value — meaning the Question 5 (meter) weakness identified job-by-job is one company-wide structural feature, not five independent risks
• A live risk to the Star label itself: if the AI/ML-server sub-position's share of CCS revenue keeps growing while it remains Blocked, the revenue-weighted majority underlying CCS's Star classification could flip toward the weaker position before the next scheduled pipeline run