Celestica Just Posted AI-Era Numbers. Now Comes the Hard Part.
Date Published
TL;DR
Quick Summary
- Celestica’s Q4 2025 was huge: $3.65B revenue (+44% YoY) and $1.89 adjusted EPS, both above guidance and expectations.
- AI data center hardware is now the star, with cloud and hardware platform revenue growing 60–70%+ year over year.
- Management raised its 2026 outlook to $17B revenue and $8.75 adjusted EPS, backed by hyperscaler demand and a $1B capex plan tied to Google TPU and other AI builds.
- The stock still slipped after-hours as investors weighed rich expectations, customer concentration, and the risks of building so much capacity into an already hot cycle.
- Celestica is evolving from a generic manufacturer into a key AI infrastructure builder, but its fate is now tightly linked to cloud AI spending cycles.
#RealTalk
Celestica just delivered the kind of AI-fueled numbers that justify its massive run, but also lock it into the mood swings of hyperscale data center spending. This is what it looks like when an old-school manufacturer steps onto the main AI infrastructure stage—high growth, high expectations, and much less room for error.
Bottom Line
Celestica’s latest quarter shows a company successfully reinventing itself around AI infrastructure, with growth and guidance that line up with the biggest cloud capex plans in the market. The upside case leans on those customers continuing to spend aggressively and Celestica executing on a very ambitious capacity build-out. The risk side revolves around valuation, customer concentration, and whether AI hardware demand can stay this hot through 2027 and beyond. For investors tracking AI’s real-world plumbing—not just the headline chipmakers—this is a name worth understanding at a business-model level, not just a chart level.
Earnings season just handed Celestica Inc. (CLS) one of those report cards you’d want framed.
On January 28, 2026, the Toronto-based hardware and supply-chain specialist delivered Q4 2025 revenue of $3.65 billion, up 44% year over year, and adjusted EPS of $1.89, ahead of expectations and even above its own guidance range. For all of 2025, revenue climbed 28% to $12.4 billion, while adjusted earnings per share jumped 56%. This is not a sleepy contract manufacturer anymore; this is an AI data center infrastructure story wearing an old-school electronics badge.
Why the growth looks different this time
Celestica’s roots are in building things for other people: servers, switches, storage systems, aerospace components, energy gear. Historically that’s been a low-margin, don’t-get-too-attached-to-any-customer kind of business.
But 2025 pushed Celestica into a different lane. Its Connectivity & Cloud Solutions arm, which includes AI-heavy hardware, generated about $2.86 billion in Q4 revenue, up 64% from the prior year. Inside that, Hardware Platform Solutions – think custom high-performance platforms for big cloud players – reached roughly $1.4 billion, up 72%. In plain English: the highest‑value, AI-adjacent stuff is what’s growing fastest.
The AI capex wave is very real here
If you want to understand why Celestica suddenly matters to every AI macro thread on your feed, look at its new outlook. Management now expects 2026 revenue of about $17 billion and adjusted EPS around $8.75, up from prior guidance of $16 billion and $8.20.
That upgrade doesn’t come from hoping the AI cycle continues; it comes from orders in hand and multi‑year roadmaps with hyperscale customers. Celestica also disclosed that three customers made up roughly a third of total revenue in 2025, which pretty much screams “mega cloud + AI players.” That concentration is a risk, but it’s also the clearest tell on where the demand is coming from.
The Google TPU twist – and $1 billion of capex
Celestica also leaned straight into one of the spiciest rumors around the stock: its relationship with Google. The company confirmed it’s expanding U.S. manufacturing capacity to support production of Google’s Tensor Processing Unit (TPU) systems and other complex data center hardware.
To make that happen, Celestica plans roughly $1 billion of capital spending in 2026, around 6% of its latest annual revenue outlook. That money is going into expanded facilities in Texas, new buildings in Thailand, upgrades in Mexico and Japan, and fresh design centers in Asia and the U.S. For a business that used to run pretty lean on investment, this is a full send into the AI infrastructure cycle.
Of course, building capacity ahead of demand is always a judgment call. If hyperscalers keep spending aggressively on AI clusters through 2027 and beyond, Celestica’s expanded footprint looks brilliant. If they slam the brakes, you’re staring at a lot of underused square footage and depreciation.
Why the stock can slide on a “beat and raise”
Despite the big numbers, Celestica’s shares dipped in after-hours trading on January 28. That sounds irrational until you remember how far the stock has run: the name has been a multi‑bagger over the past two years, and expectations were already sky‑high.
When a stock is priced for perfection, even great news has to fight gravity. Some investors will simply use a strong quarter as an excuse to lock in gains. Others are looking at that beefed‑up capex plan, customer concentration, and cyclicality in hardware and asking: how long can this pace last?
What this business is becoming
Strip away the ticker for a moment and you’re left with a clear story: Celestica is evolving from a generic electronics manufacturer into a core builder of AI data center plumbing – switches, storage, custom compute platforms, and now TPU systems.
That transition shows up in margins slowly trending higher, revenue growth tied to cloud and AI infrastructure cycles, and its presence in thematic ETFs like FFTY, FIVG, and WTAI. It’s not a pure-play AI chip designer and it doesn’t have a consumer brand, but it’s increasingly sitting in the middle of where AI workloads actually live: racks, cables, cooling, power, and boxes full of very expensive silicon.
For investors, the question from here is less “Did they beat?” and more “How durable is this positioning if the AI build‑out normalizes?” Celestica just proved it can execute into the boom. The next few years will show whether it can navigate whatever comes after.