Customer forecasts for 2026 already exceed the $17 billion guide, and management says demand visibility now runs past 2028.
Thesis: Celestica owns a real bottleneck in hyperscaler system-level assembly where engineering complexity is rising with each node transition (400G to 800G to 1.6T to 3.2T) and very few competitors can execute at scale. Management revised CCS growth up $1B in 90 days and committed $1B CapEx to booked backlog while guiding $17B revenue that customers already forecast above. The Street is modeling deceleration from Q1's 51% growth to the full-year 37% as a slowdown. It's conservatism on a guide backed by non-cancellable bookings.
Verdict: LONG — Conviction: MEDIUM
Catalyst: Q1 2026 results confirming 51% revenue growth midpoint ($3.85-4.15B) and EPS up 71% at midpoint, which would validate the back-half math the Street fears is decelerating. 1.6T mass production ramp in H2 2026 is the second structural proof point.
Key Risk: FCF delivery: $500M FCF with $1B CapEx requires ~$1.5B operating cash flow. At 7.8% op margin on $17B that's ~$1.33B adj op income plus ~$300M D&A before working capital. Growing revenue 37% eats receivables and inventory. The math is tight and any working capital drag or margin miss breaks the FCF guide.
The Tell: Mionis unprompted on the competitive moat: 'some of the business that we decided not to play the pricing game on in 2025 have come back to us in 2026 because others could not execute on it.' He volunteered this without a pricing question. That is a Pricing Power Signal he chose to plant. It says volume isn't fungible across EMS providers and complexity is the moat.
Friction Level: MODERATE_FRICTION — Both sides agree the quarter is clean. Bull sees $1B CapEx tied to bookings as revenue floor into 2027-28. Bear sees same CapEx as a top-of-cycle bet with an FCF math problem and 36% single-customer concentration. Same facts, opposite read on forward risk.
2025Q4 · Jan 29, 2026
3 of 5 readings agreed3 LONG · 2 HOLDRead the full call
Capex 5x Last Year, All Tied to Booked Business
Celestica owns a real bottleneck in hyperscaler system-level assembly where engineering complexity is rising with each node transition (400G to 800G to 1.6T to 3.2T) and very few competitors can execute at scale. Management revised CCS growth up $1B in 90 days and committed $1B CapEx to booked backlog while guiding $17B revenue that customers already forecast above. The Street is modeling deceleration from Q1's 51% growth to the full-year 37% as a slowdown. It's conservatism on a guide backed by non-cancellable bookings.
Mionis unprompted on the competitive moat: 'some of the business that we decided not to play the pricing game on in 2025 have come back to us in 2026 because others could not execute on it.' He volunteered this without a pricing question. That is a Pricing Power Signal he chose to plant. It says volume isn't fungible across EMS providers and complexity is the moat.
Q4 beat high end on both revenue ($3.65B vs high-end guide) and EPS ($1.89). Raised full-year outlook to $17B and $8.75. Q1 2026 guides +51% revenue and +71% EPS at midpoint. Third consecutive beat-and-raise cycle with CCS accelerating, not decelerating.
CapEx from $201M in 2025 to ~$1B in 2026, fully funded by operating cash flow. Management explicitly ties it to 'booked business' and 'record bookings year in 2025.' $1B committed to Texas and Thailand capacity that only makes sense against multi-year hyperscaler commitments.
CCS growth for 2026 revised from $3.5B to $4.5B in 90 days. 2027 CCS growth now implied at $7B off a higher base. Three switch generations (400G, 800G, 1.6T) running concurrently without cannibalization. Demand visibility extended past typical 4-quarter window into 2028.
Mionis stated business declined on price in 2025 returned in 2026 because competitors couldn't execute. ASPs rising on networking due to silicon pass-through. Programs getting more complex with HPS design-led engagement shifting margin mix upward.
Primary source for Google TPU with joint commitments for future generations. 1.6T win at third hyperscaler was a direct extension of 400G then 800G performance. Mionis: 'very few of our competition can produce these products at scale' combined with engineering depth.
Management repeatedly ties $1B CapEx to 'booked business' and 'record bookings.' However, hyperscaler EMS orders are cancellable by design, which tempers the fortress claim. Not pure Backlog Fortress but strong given the specificity of program-level investments.
+82
CONFIDENT"Mionis volunteered the pricing power story unprompted: business declined on price in 2025 'came back to us in 2026 because others could not execute.' CFO directly stated customer forecasts exceed their guide. Specific numbers on CCS growth revision ($3.5B to $4.5B in 90 days), 10 active 1.6T programs, margin floor of 30bps."
Rich evidence: specific revenue/EPS/margin numbers, named CapEx geography breakdown, program counts (10 1.6T programs, 5 ramping H2), customer concentration disclosed (36/15/12%), CFO walked through FCF math. Only gap is working capital assumptions for the $1B CapEx year.
Q1 2026 results confirming 51% revenue growth midpoint ($3.85-4.15B) and EPS up 71% at midpoint, which would validate the back-half math the Street fears is decelerating. 1.6T mass production ramp in H2 2026 is the second structural proof point.
FCF delivery: $500M FCF with $1B CapEx requires ~$1.5B operating cash flow. At 7.8% op margin on $17B that's ~$1.33B adj op income plus ~$300M D&A before working capital. Growing revenue 37% eats receivables and inventory. The math is tight and any working capital drag or margin miss breaks the FCF guide.
“In the fourth quarter, revenue of $3.65 billion was up 44% and above the high end of our guidance range, driven by very strong demand in our CCS segment.”
“Now moving on to our guidance for 2026, first-quarter revenue is projected to be between $3.85 billion and $4.15 billion, representing growth of 51% at the midpoint.”
Figures are extracted from the transcripts and checked in code: every quote appears word for word. Descriptive only: none of this feeds the verdict.
Q4 beat high end on revenue and EPS with 44% growth and 70% EPS growth. Record 7.7% op margin. Q1 2026 guides +51% revenue and +71% EPS. Management raised the full-year outlook to $17B and explicitly said customer forecasts exceed that number. CCS growth for 2026 was revised up $1B in 90 days (from $3.5B to $4.5B). 1.6T design win with a third hyperscaler is a direct extension of a 400G-to-800G performance track record. 10 active 1.6T programs with 5 ramping H2. Three switch generations running concurrently with no cannibalization. $1B CapEx is 5x last year and tied to booked business with customers aligning multiyear capacity roadmaps. Margin floor of 30bps expansion with gross margin already up 30bps. ROIC 43% up 14 points. Cash cycle days improved 8 days. The pricing power story is real: declined business came back because competitors can't execute at scale.
The FCF math is tight: $500M FCF with $1B CapEx requires $1.5B OCF. At 7.8% op margin on $17B revenue that's ~$1.33B adj op income. D&A adds maybe $300M. Growing revenue 37% means working capital is a use during the ramp. Cash conversion needs to exceed 100% of operating income. Concentration is extreme: 36% of revenue from Google where Mionis confirmed they're 'not sole-sourced' and hyperscalers maintain second sources for BCP. The CEO's emphatic denial of competitive threats ('never been stronger,' 'no new entrants') reads as defensive on a question investors are already pricing. Demand language is at maximum intensity: 'unprecedented,' 'high-confidence,' visibility into 2028. Building 2M+ square feet across Texas and Thailand into a hyperscaler AI CapEx cycle is a top-of-cycle capacity bet. At 7.7% op margin there's no cushion if a single hyperscaler renegotiates.
Say/Do Gap
$500M FCF with $1B CapEx requires ~$1.5B OCF. At ~7.8% op margin on $17B, adj op income is ~$1.33B. Even with D&A add-back and improving cash cycle days, the conversion required is near 100% of operating income during a 37% revenue growth year. Management says achievable, arithmetic says very tight.
Unsustainable Trend Confidence
'Unprecedented' visibility, demand extending into 2028, customer forecasts exceeding guide, 'high-confidence view' language. Every descriptor at ceiling. Valid while hyperscaler CapEx grows; breaks hard if any of three 10%+ customers revises down.
Backlog Quality
$1B CapEx justified by 'booked business,' but EMS orders from hyperscalers are cancellable by design. Management frames it as 'very little risk' but doesn't quantify cancellation terms. Building 1.7M+ square feet on capacity commitments Celestica doesn't fully control.
Zero Pricing Power Despite Constraints
7.7% op margin on a business where Google is 36% of revenue and second sources exist. Mionis confirmed 'share is largely awarded on performance' which means share can shift on performance. No structural pricing power despite being capacity-constrained.
LONG at medium conviction. The asymmetry is this: management revised CCS growth up $1B in 90 days, committed $1B CapEx to booked backlog, and explicitly said customer forecasts exceed their own $17B guide. The Street is reading Q1's 51% growth against the full-year 37% as deceleration. It's conservatism on a rising base. The structural driver is real: engineering complexity rising with each node transition means fewer competitors can execute, and Celestica is winning share on performance. Own it through Q1 2026 to confirm the 51% guide. Risk is the FCF delivery and Google concentration. Size for MEDIUM conviction, not full position, until the $500M FCF guide proves out with $1B CapEx in the same year.