CLS · Celestica, Inc. · Stock research

Current Thesis

The picks-and-shovels leg here is unchanged in kind and larger in size than when this name was last written up: Celestica builds the racks, 800G/1.6T switching and custom hardware platforms that hyperscalers order by the data-center row, and the 2026-07-27 Q2 print was another quarter where the operating line ran ahead of the guide. Revenue $4.70B, +62% YoY from $2.89B; adjusted EPS $2.54 versus $1.39 a year earlier; both above the top of the guidance ranges. Management lifted FY2026 to $20.5B revenue and $11.30 adjusted EPS, from $19.0B and $10.15 set three months earlier, with adjusted operating margin to 8.4% and free cash flow to $600M.

The tape has not followed. The 2026-08-14 close of $335.05 sits 29.1% below the $472.40 52-week high and the three-month return is -6.6% — a stretch over which the FY EPS guide rose from $10.15 to $11.30. That gap between an accelerating earnings line and a falling price is the whole situation: a multiple reset inside a working fundamental. On 2026-08-05 the company announced a $3B equity offering and priced it at $310 on 2026-08-06 (9,677,419 shares), with the 1,451,612-share underwriter option fully exercised the same day and net proceeds of roughly $3.39B earmarked for working capital, capital expenditures and general corporate purposes. What an investor is buying now is not discovery — it is the proposition that $310 marked where $3B of institutional paper cleared, and that the base built above it survives to a late-October Q3 print carrying a $5.25–$5.55B revenue guide.

The narrative is maturing. The narrative is well known and still working, but flow has moderated visibly. Dating it: the June 2026 break from the high, a beat-and-raise on 2026-07-27 that produced no new high, an equity raise cleared 34% below that high on 2026-08-06, a Barclays target cut to $406 on 2026-08-07 in the same week the EPS guide went up, and mainstream list coverage on 2026-08-14 bundling the name with fifteen others as an AI-buildout play "still 20% below highs." That is late-middle. It is also not dead — CCS grew 84% YoY in Q2 and Scotiabank reinstated at Sector Outperform with a $500 target on 2026-08-11.

Bullish and bearish views on Celestica, Inc.

The model's bull view on Celestica, Inc. (CLS), in brief: Q2 2026 (2026-07-27): revenue $4.70B, +62% YoY; adjusted EPS $2.54 vs $1.39. The bear view: A 62% revenue quarter and an FY EPS guide raised to $11.30 did not put the stock back above its June high. Both cases follow in full.

Bull Case

• Q2 2026 (2026-07-27): revenue $4.70B, +62% YoY; adjusted EPS $2.54 vs $1.39. Both above the high end of the company's own guidance, attributed to higher-than-anticipated customer demand and operating leverage.
• CCS revenue $3.81B, +84% YoY, at an 8.7% segment margin versus 8.3% in Q2 2025. The hyperscaler-facing segment is compounding and expanding margin at the same time, which cuts against the standard objection that EMS volume arrives without incremental profit.
• Hardware Platform Solutions ~$1.9B, +58% YoY (Q2 2026). The rack-scale integration line inside CCS — the dollars most levered to a single customer build cycle — is still growing off a materially larger base than the ~$1.7B booked in Q1 2026.
• FY2026 guidance raised twice in 2026: to $19.0B/$10.15 on 2026-04-27, then to $20.5B/$11.30 on 2026-07-27, with adjusted operating margin 8.4% and free cash flow $600M. The $11.30 FY guide stands against a 2026-08-14 close of $335.05.
• Management framed 2027 revenue growth as accelerating beyond the ~65% expected for 2026, with adjusted EPS growing faster than revenue (2026-07-27 release). That is a forward claim by the company and carries no evidence yet beyond the guide itself.
• The $3B raise is capacity-directed rather than balance-sheet repair: proceeds stated for working capital, capital expenditures and general corporate purposes, sized against multi-year customer demand. ATS also improved — $0.89B, +8% YoY, margin 6.3% from 5.3%.

Bear Case

• A 62% revenue quarter and an FY EPS guide raised to $11.30 did not put the stock back above its June high. Over the three months into 2026-08-14 the return is -6.6%; the multiple absorbed the entire earnings upgrade and more.
• Share count rose by up to 11,129,031 while the FY guide explicitly assumes no further repurchases or issuances for the rest of 2026.
• Sell-side is dispersed and moving in both directions: Barclays maintained Overweight but cut its target to $406 on 2026-08-07; Scotiabank reinstated at $500 on 2026-08-11. Third-party aggregation put the 15-analyst average near $426 as of 2026-08-11. A 23% spread between two of the most recent targets is not a consensus.
• Category multiple ceiling. An EMS/ODM integrator carries a structurally lower terminal multiple than the silicon designers it serves; the de-rating since June is consistent with the market repricing growth it already believes.
• Customer concentration remains the air-pocket risk. Hyperscaler demand is top-heavy; a single program push-out took roughly a fifth out of this stock inside a week in 2024, and a $20.5B revenue guide does not cushion that.
• No company-specific catalyst for roughly ten weeks. Nothing on the calendar before the Q3 print forces a re-rate; the interval is governed by sector headlines.

Setup & Price Structure

Reference points, all from the adjusted daily series: last close $335.05 (2026-08-14), 52-week high $472.40, distance from high -29.1%, three-month return -6.6%, RSI(14) 53.3. Nothing about that RSI reading is stretched in either direction — it is the middle of the range, which fits a name that has stopped falling without starting to trend.

The structure that matters is now defined by three levels. Below, $310 is where 11.1M shares were placed on 2026-08-06 — a marked clearing price with real size behind it, and the level that decides whether the deal book is offside. The pre-May 2026 breakout shelf near $330 sits just under the current close, so the tape is pinned directly on it. Above, the $390.11/$410.93 shelf cut during the June 2026 break is the reclaim zone; until a daily close prints back above $390 with a higher low behind it, June reads as an unrepaired top rather than a correction, and the June-to-August range remains the operative structure.

Crowding and positioning observables, stated as observables: equity issued into weakness at $310 (2026-08-06); a target cut to $406 seven days after a beat-and-raise (2026-08-07); options-flow retail coverage on 2026-08-10 ("whale alerts" screens); mainstream list inclusion on 2026-08-14 grouping the name with fifteen peers on the shared attribute of trading below their highs. No earnings date falls inside 30 days, so the blackout condition that governed the July note has lapsed. Price sits above the offering price and below every 2026 breakout shelf — a base under construction, with no confirmation yet that it has completed.

Catalyst Calendar (next 30 days)

• 2026-08-26 — NVIDIA Q2 FY2027 results (company-confirmed date, call 5pm ET). Not a Celestica event, but the single largest scheduled read-through on AI-rack and networking build rates inside the window; the whole cohort has repriced on prior NVDA demand commentary.
• ~2026-10-26 (est.) — Q3 2026 print. Guide is $5.25–$5.55B revenue and $2.88–$3.08 adjusted EPS, adjusted operating margin 8.4% at the midpoint. First hard test of the raised $20.5B/$11.30 FY2026 outlook and the first quarter reported on the enlarged share count.
• ~late Oct 2026 (est.) — Alphabet, Microsoft, Meta and Amazon Q3 capex updates, historically clustered within days of the Celestica print. Hyperscaler capex guidance is the primary CCS swing factor.

Elapsed catalysts

• ~2026-09 (est.) — September broker technology conferences. Any updated CCS or 2027 framing outside the print cycle would land here. Celestica participation unconfirmed as of 2026-08-16. (passed 10d ago)

What Would Change Our Mind

The structure breaks if the post-raise base fails. The $3B placement at $310 on 2026-08-06 gave this name a marked institutional clearing price; a weekly close below $310 puts that entire book underwater and turns the June–August range into distribution rather than repair. That is the gradeable line.

Second, the fundamental line. The Q3 guide floor is $5.25B revenue with adjusted EPS of $2.88 at the low end; revenue landing under $5.25B, or any trim to the $20.5B / $11.30 FY2026 outlook at the October print, contradicts the acceleration claim outright. CCS is the segment to read — Q2 was $3.81B at 8.7% margin, and CCS YoY growth decelerating toward 50%, or margin printing below 8.7%, would matter more than a headline revenue beat.

Third, the life-cycle label. This flips to saturated if a second beat-and-raise fails to produce a close above $390, or if further equity supply arrives before the Q3 print. Running the other way, a daily close above $390 with a higher low left behind in the $340–$360 zone would re-open the June high as a reference and argue the de-rating is finished.

Correlation Notes

• Trades as a high-beta expression of hyperscaler capex. The dominant correlation is to AI-infrastructure capex headlines from Alphabet, Microsoft, Meta and Amazon, and secondarily to the AI-networking complex (ANET, AVGO) rather than to legacy EMS peers.
• periodic mispricing in both directions, and sharper drawdowns when momentum funds rotate out of the theme.
• The 2026-08-26 NVIDIA print is the nearest scheduled event capable of moving the whole cohort; the name has no company-specific release inside the window to differentiate it from the group.
• Post-raise, the float is larger by up to 11,129,031 shares, which mechanically dampens the reflexive squeeze dynamics that amplified the 2026 advance.