CFO gave a $25.5B revenue floor for 2027 unprompted, while leaving free cash flow guidance exactly where it was.
Thesis: Celestica owns the assembly and design bottleneck for AI networking at the moment the 800G-to-1.6T transition is starting. Component constraints (custom silicon, memory, 40+ layer PCBs) are gating their own growth, which means they have pricing power and backlog quality. The flat FCF guide is the only blemish, but it's funded working capital growth during a documented NCNR-backed ramp. The $25.5B 2027 floor on already-won programs gives visible upside to current estimates.
Verdict: LONG — Conviction: MEDIUM
Catalyst: Q2 enterprise revenue growth of ~130% as the component-constrained AI/ML compute program catches up, confirming constraints are temporary supply issues not demand cracks.
Key Risk: Hyperscaler CapEx guide-down. 65% of revenue rides three CapEx budgets (35/15/15). NCNR commitments convert $2.67B inventory and $1.5B 2027 CapEx into a cash flow problem if those forecasts slip.
The Tell: CFO volunteered the $25.5B 2027 floor unprompted when asked about growth drivers: 'the floor would be somewhere around $25.5 billion... on programs that we've won.' Giving a specific multi-billion floor well beyond guided horizon with 'programs already won' language reveals how much visibility management has. This is a commitment, not a projection.
Friction Level: MODERATE_FRICTION — Both sides agree on beat-and-raise, real program ramps, and component constraints. Disagreement centers on whether flat FCF guide amid $2B revenue raise is a Say/Do Gap signaling working capital risk or normal ramp-phase investment covered by NCNR-backed demand.
2026Q1 · Apr 28, 2026
4 of 5 readings agreed4 LONG · 1 HOLDRead the full call
Raised Revenue $2B, Raised EPS $1.40, FCF Guide Unmoved at $500M
Celestica owns the assembly and design bottleneck for AI networking at the moment the 800G-to-1.6T transition is starting. Component constraints (custom silicon, memory, 40+ layer PCBs) are gating their own growth, which means they have pricing power and backlog quality. The flat FCF guide is the only blemish, but it's funded working capital growth during a documented NCNR-backed ramp. The $25.5B 2027 floor on already-won programs gives visible upside to current estimates.
CFO volunteered the $25.5B 2027 floor unprompted when asked about growth drivers: 'the floor would be somewhere around $25.5 billion... on programs that we've won.' Giving a specific multi-billion floor well beyond guided horizon with 'programs already won' language reveals how much visibility management has. This is a commitment, not a projection.
Raised revenue from $17B to $19B (+$2B), EPS from $8.75 to $10.15, and op margin from 7.8% to 8.1% in a single print. Q1 EPS $2.16 exceeded high end of guidance. Consistent beat-and-raise pattern tied to structural 1.6T and AI/ML compute program ramps.
CEO confirmed 'a lot of the orders that we are placing supported by our customers are NCNR.' Awards shipping into 2028, lead times beyond one year, and contractual protection on capacity commitments. Backlog quality is high.
CapEx rising from $1B in 2026 to ~$1.5B placeholder for 2027, deployed into Thailand and Texas for programs shipping 2028-2029. R&D at 1,350 design engineers, up YoY, working on 2027-2028 programs. Revolver extended to 2031 with $1B increase to $1.75B.
CFO stated directly: 'We have capacities at a premium... We're not giving up price in the marketplace.' Op margin expanded 90bps to record 8.0% while on allocation. Gross margin also expanded 30bps despite component cost headwinds.
Revenue raised $2B, EPS raised $1.40, op margin raised 30bps, yet FCF guide left flat at $500M. Inventory up $485M sequentially to $2.67B. Q1 FCF was $138M against ~$260M adjusted net income. Working capital is absorbing all incremental profit.
CEO: 'We are experiencing more component shortages now than 90 days ago.' Q1 enterprise revenue missed its own target due to component constraint on AI/ML compute ramp. Constrained commodities: custom silicon, memory, 40+ layer PCBs, power, optics. Supply is the ceiling on growth.
ATS margin gain of 100bps attributed to 'portfolio reshaping' on a segment that posted flat revenue at $806M. Portfolio optimization is cutting A&D revenue while rebranding the remainder as higher quality.
+75
CONFIDENT"CEO volunteered specific component constraint details unprompted ('it wasn't a demand issue, it was actually a material supply issue'). CFO gave a $25.5B floor for 2027 beyond guided horizon. Specific numbers throughout: 10 active 1.6T programs, 1,350 design engineers, NCNR commitments, 2028 shipping awards."
Rich evidence: specific program counts, customer concentration disclosed (35/15/15), named constrained commodities, CapEx specifics by geography, ROIC number. Gaps: no disaggregation of scale-up vs scale-out revenue contribution, no CPO deal sizing, no breakdown of which hyperscaler drives which program.
Q2 enterprise revenue growth of ~130% as the component-constrained AI/ML compute program catches up, confirming constraints are temporary supply issues not demand cracks.
Hyperscaler CapEx guide-down. 65% of revenue rides three CapEx budgets (35/15/15). NCNR commitments convert $2.67B inventory and $1.5B 2027 CapEx into a cash flow problem if those forecasts slip.
“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.”
“Revenue in the first quarter was $4.05 billion, up 53%, just above the midpoint of our guidance range, driven by very strong demand in our CCS segment.”
Companies tend to guide low: the median company we cover reports +1.7% against its guidance midpoint, so a small beat is the norm. A transcript can drop part of a guided range, so read both quotes.
“Second quarter revenue is projected to be between $4.15 billion and $4.45 billion, representing growth of 49% at the midpoint.”
Changed less than 95% of calls
Wording compared with the Jan 29, 2026 call, ranked among 97 calls reporting on Q1 2026. A measure of how much the story moved, not a signal about the stock.
Figures are extracted from the transcripts and checked in code: every quote appears word for word. Language change is computed from word frequencies, with no model involved. Descriptive only: none of this feeds the verdict.
Beat-and-raise on structural drivers: 1.6T switch ramps, Broadcom Tomahawk 6 CPO win, AMD Helios rack, digital native rack-scale production Q1 2027. Ten active 1.6T programs ramping 2027, two customers entering mass production H2 2026. Component constraints are gating supply, not demand, which means pricing power: op margin record 8.0%, CFO says 'capacities at a premium, we're not giving up price.' NCNR-backed backlog shipping into 2028. CFO gave $25.5B floor for 2027 on already-won programs. CapEx at $1.5B for 2027 shows management putting skin in the game. Q2 enterprise growth of ~130% as constrained program catches up. ROIC hit 50%, up 18pts YoY.
Raised revenue $2B and left FCF guide flat at $500M. Inventory at $2.67B, up $885M YoY, funded partly by $1B revolver increase to $1.75B announced post-close. Q1 FCF of $138M against ~$260M adjusted net income means cash conversion is broken during ramp. 65% of revenue concentrated in three hyperscaler CapEx budgets (35/15/15). CEO admits shortages are worse than 90 days ago with lead times extending beyond one year. Memory and silicon input costs rising materially, gross margin compression deferred into guide. ATS margin gain is portfolio reshaping spin on flat revenue. NCNR commitments cut both ways if hyperscaler forecasts slip.
Say/Do Gap
Raised revenue by $2B and EPS by $1.40 but FCF guide unchanged at $500M. Working capital is absorbing all incremental profit: inventory up $485M sequentially, revolver increased $1B post-close. Legitimate if NCNR-backed demand holds, but the gap is wide.
Capacity Ceiling
CEO confirmed component shortages worse than 90 days ago, Q1 enterprise missed own target due to supply constraint. Constrained: custom silicon, memory, 40+ layer PCBs, power, optics. Supply is pacing growth. Management betting all suppliers commit to the raised guide.
Mix Shift Spin
ATS margin +100bps attributed to portfolio reshaping on flat $806M revenue. A&D revenue being cut is real portfolio optimization but presenting it as margin expansion quality is spin.
Backlog Quality
NCNR orders are non-cancellable for Celestica, but Celestica has pre-committed to component purchases against hyperscaler forecasts. If forecasts slip, inventory risk is on Celestica's balance sheet, not the customer's.
LONG. The constraint is the edge: component shortages are gating Celestica's own growth, which means they hold pricing power and their backlog is NCNR. Ten 1.6T programs, CPO on Tomahawk 6, digital native rack, and a $25.5B 2027 floor on already-won programs give visible revenue cover. The flat FCF guide is a legitimate concern but explainable as working capital investment during a documented supply-constrained ramp. Risk/reward favors long with a stop if any hyperscaler guides CapEx down. Size as a medium position, add on Q2 enterprise beat confirming the catch-up.