Celestica Inc.

Thesis

Celestica is the purest large-cap manufacturing play on hyperscaler AI networking and rack integration — CCS segment revenue grew 76% y/y in Q1 2026 and management guides $19.0B for 2026 — and we expect revenue to compound roughly 16%/yr through 2031. But at $339 the stock already discounts near-flawless execution: our base case is worth roughly $160 in present value against a $339 price, so this publishes as a watchlist thesis, not a buy. The operating thesis is falsified if CCS growth stalls to zero for two consecutive quarters or a top-three hyperscaler program is lost; the valuation becomes actionable below roughly $170.

Why Now

The inflection is in the filings, and it is steep. Q1 2026 revenue was $4,047.0M, up 53% from $2,648.6M a year earlier (Q1 2026 earnings release, 8-K exhibit filed 2026-04-27, for the 10-Q period ended 2026-03-31). Inside that, the Connectivity & Cloud Solutions (CCS) segment — hyperscaler networking switches plus AI/ML compute and storage — did $3.24B, up 76% y/y, with segment margin expanding to 8.6% from 8.0%. Hardware Platform Solutions, Celestica's own-designed platforms rather than build-to-print work, reached roughly $1.7B in the quarter, up 63%, meaning Celestica-designed hardware is now about 42% of total company revenue. Management raised the 2026 outlook from $17.0B to $19.0B in revenue and from $8.75 to $10.15 in adjusted EPS in a single quarter — a $2B guide-up 90 days after the prior guide was set (Q4/FY2025 release, 2026-01-28; Q1 2026 release, 2026-04-27).

The full-year 2025 numbers show this is acceleration, not a one-quarter blip: FY2025 revenue was $12.39B, up 28% from $9.65B in 2024, with Q4 2025 alone at $3.65B, up 44% y/y (Q4/FY2025 press release, 2026-01-28; FY2025 Form 10-K filed 2026-02-27). Sequencing the quarterly growth rates — +28% for FY2025, +44% in Q4 2025, +53% in Q1 2026, with Q2 2026 guided to $4.15–4.45B — growth is speeding up as AI programs ramp, not decaying.

The demand side is corroborated by the customers' own filings: the big four hyperscalers (Microsoft, Amazon, Alphabet, Meta) have guided roughly $700–725B of combined 2026 capex, up ~77% from about $410B in 2025, with roughly three-quarters of it AI-related. Celestica also disclosed a new design win in the Q1 release — a 1.6-terabit co-packaged-optics Ethernet switch for a hyperscaler's AI scale-out network, ramping production in 2027 — which extends program visibility into the next networking generation. On 2026-04-27 the company upsized its credit facility to approximately $2.5B explicitly to fund the working capital of this scale-up.

The reason to write this now rather than later: the stock has corrected roughly 28% from its 52-week high of $474 to $339, Q2 2026 results land within days of publication, and the honest question is whether the drawdown has created a five-year entry point. Our math below says it has not — yet.

Business Quality

Celestica has spent a decade migrating from commodity electronics assembly toward an ODM model, and the mix shift is now the story. HPS — platforms Celestica designs itself, primarily networking switches and storage systems for hyperscalers — is ~42% of revenue and growing 63% y/y. That design layer is why CCS segment margin runs at 8.6% while traditional EMS competitors (Flex, Jabil, Benchmark) operate at 4–5% operating margins. Gross margin is still thin at 10.8% (Q1 2026: $437M gross profit on $4,047M revenue) — this is manufacturing, not software — but returns on capital are strong because the balance sheet is light: net debt of roughly $0.4B against annualized net earnings of ~$850M, and Q1 2026 GAAP net margin of 5.2%.

The moat is execution and co-design entanglement rather than pricing power. Winning an 800G or 1.6T switch program means years of joint engineering with the hyperscaler's silicon and thermal roadmap (the new CPO win involves co-packaged optics and liquid cooling); those programs do not move to a competitor mid-generation. The counterweight is brutal customer concentration: the FY2025 10-K discloses that the top ten customers were 79% of revenue and three CCS customers individually represented 32%, 14%, and 12%. These customers are far larger than Celestica, dual-source deliberately, and reset pricing at every generation transition. Celestica's margin gains to date suggest it is winning more value per box, but there is no world where it accrues pricing power over Microsoft or Meta-scale buyers.

Capital allocation has been shareholder-aligned: no dividend, consistent buybacks (diluted weighted-average shares fell from 116.9M to 115.7M y/y even during a capex ramp), and debt used sparingly. Cash generation is the soft spot in hypergrowth — 2026 free cash flow is guided to just $500M on $19B of revenue (2.6%), because receivables ($3.17B) and inventory absorb cash as programs ramp. That is acceptable while growth is funded at high incremental returns, but it means the equity's cash yield today is ~1.3%, and the FCF margin must roughly double post-ramp for the valuation to work on cash flow terms.

Competitively, Celestica sits in the strongest niche of the AI-hardware food chain below the chipmakers: white-box switches and integrated racks are gaining share from branded networking (Cisco, Arista lose share to ODM-direct at hyperscalers), and Celestica is the western-listed leader of that trade versus Taiwanese rivals (Accton, Quanta, Wiwynn, Foxconn). The durable risk is that value migrates again — co-packaged optics pushes more content into silicon vendors, and hyperscalers could take switch design fully in-house the way they took server design.

Financial Base

Reference figures as of 2026-07-23: trailing-twelve-month revenue of $13,788M, 115.7M diluted shares, net debt of $394M, and a share price of $339.10. Sources:

• Q1 2026 earnings release with condensed FS (8-K exhibit, filed 2026-04-27; period ended 2026-03-31) — revenue $4,047.0M (+53% y/y); CCS segment $3.24B (+76%), segment margin 8.6%; HPS ~$1.7B (+63%); diluted weighted-avg shares 115.7M; cash $378.0M; borrowings under credit facility and finance lease obligations $772.4M ($25.9M current + $746.5M non-current); net earnings $212.3M
• Q4/FY2025 press release (2026-01-28) and FY2025 Form 10-K (filed 2026-02-27) — FY2025 revenue $12.39B (+28% vs $9.65B in 2024); Q4 2025 revenue $3.65B (+44% y/y); top-10 customers 79% of FY2025 revenue; three CCS customers at 32%, 14%, and 12% of total revenue
• TTM revenue construction: FY2025 $12,390M − Q1 2025 $2,648.6M + Q1 2026 $4,047.0M = $13,788.4M
• Share price $339.10 at close 2026-07-23 (GuruFocus market data; 52-week range $153.11–$474.03); implies market cap ~$39.2B, EV ~$39.6B, 2.87× EV/TTM sales, ~33× 2026 guided adjusted EPS of $10.15
• 2026 guidance (Q1 2026 release, 2026-04-27): revenue $19.0B, adjusted EPS $10.15, adjusted operating margin 8.1%, free cash flow $500M; credit facility upsized to ~$2.5B on 2026-04-27

Scenario Valuation

Five-year scenario ranges under the published methodology — the tables below are computed from these exact parameters at publication and sealed with this text. What each scenario assumes:

• Loss (6%): AI capex bust combined with loss or insourcing of a top-three hyperscaler program (top customer alone is 32% of revenue); CCS shrinks back toward 2024 scale and the market pays a trough EMS multiple. $100 is ~29% of the current price and ~13× a bust-case ~$7-8 EPS.
• Bear (24%): 2026 lands near guide but hyperscaler capex digests in 2027-28 (the 77% y/y capex surge proves unsustainable); revenue dips then partially recovers to ~$17.6B by year 5. Exit at 0.8× EV/Sales — the upper end of Flex/Jabil's historical range, credited for Celestica's better margin structure. FV $118/share, PV $73.
• Base (50%): AI infrastructure buildout continues but decelerates: $19B in 2026 (guided) fading to single-digit growth by 2030, reaching ~$29B in year 5. FCF margin doubles from today's 2.6% to 6% as working-capital absorption ends (in line with guided ~6.2% adjusted net margin); 17× P/FCF exit matches mature high-quality manufacturers (Flex/Jabil trade 12-16×, premium for ODM mix). Modest continued buybacks. FV $258/share, PV $160 — below today's $339.
• Bull (20%): Hyperscaler capex keeps compounding through 2030; Celestica holds share through the 1.6T/CPO generation and full-rack integration scales, reaching ~$40B revenue in year 5 with 7% FCF margin (implied exit ~1.4× sales at 20× FCF). FV $511/share, PV $317 — an 8.5%/yr IRR from today's price even in the bull case.

Published range (present value): $73.44 – $317.13 · probability-weighted expected value $167.24 against $339.10 at drafting (0.49×) · base-case IRR −5.3%/yr. Gates: base IRR ≥ 15% and EV ≥ 1.3× — both passed at publication. Ranges are graded at the 1-year checkpoint and the 5-year horizon. This is a range, not a target.

Sensitivity — present value across growth × exit multiple

The shaded cell is the one today's price of $339.10 most closely implies. Find your own cell — we show our work rather than assert a number.

Kill Conditions

1. 1. CCS segment revenue growth is zero or negative year-over-year for two consecutive quarters (10-Q segment note; baseline +76% in Q1 2026) — the AI-driven growth engine has stalled.
2. 2. Any of the three CCS customers disclosed at 32%, 14%, and 12% of FY2025 revenue (FY2025 10-K) falls below the 10% disclosure threshold in a subsequent 10-K without an offsetting new 10% customer, or the company discloses transfer/insourcing of a major hyperscaler program.
3. 3. Combined quarterly capex of Microsoft, Amazon, Alphabet, and Meta declines year-over-year for two consecutive quarters (public earnings releases) — the theme's funding source is contracting.
4. 4. Management cuts the annual revenue outlook by 10% or more in any quarterly update (e.g., 2026's $19.0B guided below $17.1B).
5. 5. CCS segment margin prints below 7.0% for two consecutive quarters (vs 8.6% in Q1 2026) — evidence hyperscalers are clawing back the ODM margin premium.

Risk Register

• Customer concentration — severe: top customer is 32% of revenue, top three are 58% (FY2025 10-K); one insourcing or dual-sourcing decision resets the model in a single quarter.
• AI capex cyclicality — severe: big-4 hyperscaler capex is guided up ~77% y/y to ~$725B in 2026; ~80% of Celestica's revenue rides on spending growing faster than the customers' AI revenues, a gap the market is already questioning.
• Valuation — severe: at 2.87× EV/TTM sales and ~33× 2026 adjusted EPS for a 10.8%-gross-margin manufacturer, multiple compression alone can produce large losses even if operations execute (the stock is already 28% off its $474 high).
• Technology value migration — moderate: co-packaged optics shifts content toward silicon/optics vendors; Celestica's 1.6T CPO win (ramping 2027) mitigates but does not eliminate the risk that the next generation carries less manufacturable value.
• Competition — moderate: Accton, Quanta, Wiwynn, Foxconn in white-box networking and racks; hyperscalers deliberately dual-source and re-bid at each generation.
• Cash conversion — moderate: 2026 FCF guided at $500M on $19B revenue (2.6%); receivables of $3.17B and ramp inventory absorb cash, and the thesis requires FCF margin to roughly double post-ramp.
• Tariffs/geopolitics — moderate: management assumes customers keep absorbing tariff costs (Q1 2026 release assumptions); a change in that pass-through breaks the margin structure.

Theme Position

Celestica's theme exposure is about as direct as the public markets offer outside the chipmakers. The CCS segment — hyperscaler networking switches, AI/ML compute racks, and storage — was $3.24B of $4.05B total revenue in Q1 2026, exactly 80% of the company, growing 76% y/y with Enterprise (AI/ML compute) up 101% and Communications (800G/400G switch ramps) up 69%. Within that, ~$1.7B per quarter is HPS: hardware Celestica itself designed, sold to hyperscalers building AI data centers. Strictly, some CCS revenue is legacy networking rather than AI, so \"~80% AI-exposed\" is the ceiling, not a precise floor — but management attributes essentially all segment growth to AI/data-center programs. The remaining 20% (ATS: aerospace/defense, industrial, health tech) was flat y/y and is diversification ballast.

Alternatives considered. Fabrinet (FN) — purer optics exposure to the same buildout, but concentrated in one customer (Nvidia optical transceivers) and trades at a similar premium; it swaps Celestica's hyperscaler concentration for single-vendor concentration without a cheaper price. Vertiv (VRT) — power and cooling is the least substitutable layer of the AI data center, but the market figured that out first; it has re-rated harder than Celestica and fails the same valuation gates by a wider margin. Flex (FLEX) / Jabil (JBL) — meaningfully cheaper (~0.7-1.2× EV/S), and Flex has a real data-center power business, but AI-exposed revenue is diluted to an estimated 20-40% of the company; they are cheaper because they are less of the thing this theme is about. Celestica was chosen as the highest-purity, highest-growth expression of the ODM-networking trade — which is precisely why its price already reflects it.

Benchmark

Beating means CLS total return from publication exceeds QQQ total return over the same 5-year window. Success means beating QQQ's total window return from publication; the scoreboard grades this thesis against that bar, live, whether we like the answer or not.