Celestica (NYSE: CLS) Deep Research: Is AI Networking Growth Already Priced In?

Celestica has undergone one of the most important business-model transitions in the electronics manufacturing industry.

For years, investors primarily viewed the company as an electronics manufacturing services provider: a contract manufacturer that assembled complex products for customers while earning relatively thin margins. That description is no longer sufficient. Celestica is increasingly designing, integrating, manufacturing, and supporting high-speed networking, AI compute, storage, power, and rack-scale infrastructure for hyperscalers and other data-center customers.

The change matters because the economic value of Celestica’s work has moved upward.

The company is no longer paid only for factory capacity and supply-chain execution. A growing share of its opportunity comes from Hardware Platform Solutions, or HPS, where Celestica contributes engineering, platform design, system integration, software support, thermal management, validation, and manufacturing at scale. Those capabilities are especially relevant as AI data centers transition from 400G and 800G networking toward 1.6T Ethernet, liquid-cooled racks, co-packaged optics, and increasingly complex scale-up and scale-out architectures.

That transition has already appeared in the financial statements.

Celestica reported first-quarter 2026 revenue of $4.05 billion, up 53% year over year. Connectivity & Cloud Solutions revenue increased 76% to $3.24 billion, HPS revenue rose 63% to approximately $1.7 billion, and adjusted operating margin reached 8.0%. Management subsequently raised its 2026 outlook to $19.0 billion of revenue and $10.15 of adjusted earnings per share.

The investment question is therefore no longer whether Celestica has AI exposure.

The real question is whether the company can sustain unusually high revenue growth, protect its improved margins, manage customer concentration, convert rapid expansion into cash flow, and justify a valuation that has moved far above traditional contract-manufacturing multiples.

Direct answer: Celestica is one of the clearest public-market beneficiaries of the shift toward open, high-speed AI networking infrastructure. Its strongest advantages are platform engineering, manufacturing execution, hyperscaler relationships, and early positioning in 800G, 1.6T, liquid-cooled, and rack-scale systems. However, CLS is not a semiconductor-style monopoly. It remains a customer-concentrated, capital-intensive systems supplier operating between powerful hyperscalers and powerful chip vendors. The stock’s future return depends less on proving that AI networking demand exists and more on proving that 2026–2027 growth can remain strong without margin compression, program delays, excessive working-capital use, or a valuation de-rating.

Celestica Investment Thesis at a Glance

01What Does Celestica Actually Do?

Celestica operates through two reportable segments:

1. Connectivity & Cloud Solutions, or CCS
2. Advanced Technology Solutions, or ATS

Understanding the difference is essential. Celestica should not be analyzed as a single undifferentiated manufacturing company because the two segments have different end markets, growth rates, competitive dynamics, and margin profiles.

Connectivity & Cloud Solutions

CCS contains Celestica’s communications and enterprise businesses.

The communications business includes networking platforms, optical products, and related infrastructure. The enterprise business includes servers, storage, and AI/ML compute programs. This segment has become the company’s primary growth engine because hyperscalers are investing heavily in data-center networking, custom compute, AI racks, and higher-bandwidth systems.

In Q1 2026:

• CCS generated approximately $3.24 billion of revenue.
• CCS revenue increased 76% year over year.
• CCS represented roughly 80% of total company revenue.
• CCS segment margin reached 8.6%.
• HPS revenue was approximately $1.7 billion, up 63%.

These figures show that Celestica’s current investment case is primarily a CCS case.

Advanced Technology Solutions

ATS includes:

• Aerospace and defense
• Industrial
• HealthTech
• Capital equipment

ATS generated approximately $810 million of Q1 2026 revenue and a 6.0% segment margin. Revenue was roughly flat year over year, but margin improved from 5.0%.

ATS is not the main reason investors have re-rated CLS as an AI infrastructure stock. However, it remains strategically important for three reasons.

First, it diversifies Celestica away from data-center customers. Second, it exposes the company to industries with long product cycles and demanding engineering standards. Third, ATS may provide earnings stability if hyperscaler demand becomes more volatile.

The segment is therefore best viewed as a portfolio stabilizer rather than the main source of near-term upside.

Hardware Platform Solutions Is Not a Separate Reportable Segment

Investors should avoid a common analytical error: treating HPS as if it were a third reportable segment.

HPS is a solution portfolio within Celestica’s broader operations. It includes Celestica-designed or co-designed platforms in networking, compute, and storage. HPS matters because it generally represents higher engineering content and greater strategic participation than traditional build-to-print manufacturing.

The distinction is important:

• Traditional EMS: the customer owns most of the design and uses the manufacturer primarily for production and supply-chain execution.
• HPS model: Celestica may provide platform design, electrical and mechanical engineering, thermal architecture, software integration, validation, component qualification, manufacturing, and lifecycle services.

That deeper role can support better margins and stronger program stickiness. It also increases development cost, execution risk, and exposure to product-roadmap changes.

02Why AI Networking Has Become a Critical Bottleneck

AI infrastructure is often discussed as if accelerator supply were the only constraint. GPUs and custom accelerators remain central, but large AI clusters cannot deliver useful performance unless data moves efficiently between processors, memory, storage, and other racks.

As clusters grow, networking becomes more important for four reasons.

GPU Utilization Depends on the Network

An expensive accelerator that waits for data or synchronization is underutilized capital.

Training workloads require large numbers of accelerators to exchange information repeatedly. Congestion, packet loss, latency, topology limitations, and insufficient bandwidth can reduce effective cluster performance even when the underlying chips are powerful.

This means that AI infrastructure spending cannot stop at compute. The system also requires:

• High-bandwidth switches
• Network-interface cards
• Optical transceivers
• Copper interconnects
• High-speed printed circuit boards
• Low-latency software
• Power-delivery systems
• Liquid cooling
• Rack-level integration
• Monitoring and orchestration

Celestica participates in several of these layers through networking platforms and system integration.

The Market Is Moving From 400G and 800G Toward 1.6T

Each networking generation increases system complexity.

A 1.6T switch is not simply an 800G switch with a higher number printed on the front. Higher speeds create more difficult signal-integrity, power, cooling, optical, mechanical, and manufacturing challenges.

Celestica’s DS6000-series platforms use Broadcom Tomahawk 6 silicon and provide up to 102.4 Tbps of non-blocking switching capacity through 64 ports of 1.6TbE connectivity. The product family includes air-cooled and hybrid-cooled designs intended for different rack standards.

The importance of this transition is not merely the product specification. It demonstrates Celestica’s ability to productize leading-edge merchant silicon quickly and deliver systems suitable for hyperscale deployments.

Ethernet Is Expanding in AI Back-End Networks

NVIDIA’s InfiniBand ecosystem has been highly influential in AI networking. However, hyperscalers and industry participants also have strong incentives to develop open Ethernet-based alternatives.

Those incentives include:

• Supplier diversification
• Open standards
• Broader hardware choice
• Integration with existing data-center networks
• Potentially lower cost
• Greater control over software and architecture

Ethernet’s expansion does not guarantee that every Ethernet supplier wins. It does expand the addressable market for companies able to design and manufacture high-performance open networking platforms.

Celestica is positioned in that ecosystem through merchant-silicon platforms, SONiC support, OCP-aligned systems, and partnerships with major technology suppliers.

AI Infrastructure Is Moving Toward Rack-Scale Systems

The unit of competition is moving from individual servers toward integrated racks and clusters.

A rack-scale system may combine:

• Accelerators
• CPUs
• Memory
• Scale-up switches
• Scale-out switches
• Power shelves
• Cooling distribution
• Cables
• Management systems
• Storage
• Software
• Mechanical integration

This creates a larger role for companies that can bridge design and high-volume manufacturing. Celestica’s opportunity is to become a system-level partner rather than only a device-level assembler.

03Celestica’s Position in the AI Infrastructure Stack

Celestica does not control the entire stack. It depends on technology roadmaps established by semiconductor vendors, cloud customers, optics suppliers, and standards bodies.

Its value lies in converting those roadmaps into deployable systems.

This table reveals both the strength and limitation of the business.

Celestica is valuable because advanced systems are difficult to engineer and manufacture. However, it does not own the dominant accelerator architecture, switching ASIC, optical standard, or cloud workload. Its bargaining power is therefore meaningful but not absolute.

04The Core Bull Case for Celestica Stock

The bullish CLS thesis rests on six connected ideas.

4.1 Celestica Has Escaped the Lowest-Value Part of EMS

Traditional electronics manufacturing can be a difficult business.

Customers often own the design, suppliers compete on cost, contracts may not guarantee volume, and margins can remain low despite large revenue. Celestica spent years repositioning away from lower-margin programs and increasing its engineering and platform content.

That transformation is visible in the margin trend.

CCS segment margin improved from 6.2% in 2023 to 7.4% in 2024, 8.2% in 2025, and 8.6% in Q1 2026. The improvement suggests that mix, scale, productivity, and HPS growth are changing the earnings profile.

This does not turn Celestica into a software company. It does indicate that the business is no longer accurately valued using the weakest contract-manufacturing comparisons.

4.2 Celestica Is Leveraged to Multiple AI Networking Transitions

The company is exposed to several overlapping upgrade cycles:

• 400G to 800G
• 800G to 1.6T
• Air cooling to liquid and hybrid cooling
• Pluggable optics toward LPO and CPO
• Standalone switching toward rack-scale integration
• Traditional data-center networks toward AI scale-out fabrics
• Proprietary architectures toward open Ethernet alternatives
• Device manufacturing toward system-level HPS programs

The number of transitions matters because it reduces dependence on a single product launch. If Celestica executes well, it can participate in multiple generations of infrastructure spending.

4.3 Design Wins Can Create Multi-Year Revenue

A hyperscaler design win is more valuable than a one-time equipment order.

Once a system enters qualification and production, switching suppliers can be expensive and operationally risky. Customers must consider engineering validation, manufacturing yield, component qualification, software compatibility, logistics, capacity, and quality.

This can make successful programs relatively sticky over their lifecycle.

However, investors should not confuse stickiness with guaranteed revenue. Celestica states that customer agreements generally do not guarantee a fixed level of business. Hyperscalers can change forecasts, dual-source programs, negotiate pricing, delay deployments, or redesign architectures.

The correct conclusion is that design wins can improve visibility, not eliminate volume risk.

4.4 Revenue Growth Is Converting Into Earnings Growth

Celestica’s 2025 revenue increased 28% to $12.39 billion, while adjusted EPS increased 56% to $6.05. Management’s current 2026 outlook calls for $19.0 billion of revenue and $10.15 of adjusted EPS.

Using those figures:

• Expected revenue growth is approximately 53%.
• Expected adjusted EPS growth is approximately 68%.

EPS is growing faster than revenue because of operating leverage, mix, and margin expansion.

That relationship is central to the investment case. If revenue grows but margins decline, the valuation framework changes quickly. If both revenue and margin remain strong, earnings can continue compounding faster than sales.

4.5 Celestica Is Expanding Capacity Before Major 2027 Ramps

Management expects approximately $1 billion of 2026 capital expenditures, or about 5% of its current revenue outlook.

The spending is intended to support demand in 2026 and 2027, particularly in CCS. Celestica is adding manufacturing and engineering capacity across multiple regions.

The bullish interpretation is that customers are giving Celestica enough visibility to justify unusually large investment.

The bearish interpretation is that the company is committing capital near the strongest point of the cycle. If customer demand changes, new facilities can create depreciation, underutilization, and working-capital pressure.

Capex is therefore both evidence of confidence and a major thesis risk.

4.6 Celestica Is Moving Into More Strategic Programs

Three program areas illustrate the move upward.

1.6T Networking

Celestica’s DS6000 and DS6001 platforms use 1.6TbE connectivity and 102.4 Tbps switching capacity. This positions the company for the next high-speed Ethernet generation.

Co-Packaged Optics

In Q1 2026, Celestica disclosed a hyperscaler award for the design and manufacturing of a CPO Ethernet switch using 1.6T switch silicon, co-packaged optical interconnects, and liquid cooling. Production is expected to begin ramping in 2027.

CPO is important because electrical interconnects become more difficult at higher bandwidths and longer distances. Integrating optics closer to the switch silicon may improve bandwidth density and power efficiency, but it also increases packaging, thermal, serviceability, and manufacturing complexity.

AMD Helios

Celestica and AMD announced a collaboration on scale-up networking switches for AMD’s Helios rack-scale AI platform. The systems are intended to connect next-generation AMD Instinct MI450-series GPUs using the Ultra Accelerator Link over Ethernet architecture.

Helios gives Celestica exposure to a potential open, non-NVIDIA rack-scale ecosystem. It also introduces execution risk because adoption depends on AMD’s accelerator roadmap, customer deployment decisions, software maturity, and the development of the surrounding ecosystem.

05What Makes Celestica Different From a Traditional Contract Manufacturer?

The word “manufacturer” can obscure the differences between suppliers.

Celestica’s higher-value work combines five capabilities.

Engineering

High-speed networking platforms require expertise in:

• Signal integrity
• Power integrity
• Thermal design
• Mechanical design
• PCB layout
• Firmware
• Software integration
• Optical compatibility
• Compliance
• Product validation

As data rates increase, small design errors can cause unacceptable performance, power, or reliability problems.

New Product Introduction

A prototype is not the same as a production platform.

New product introduction requires Celestica to translate a design into a repeatable manufacturing process. That includes test development, tooling, component qualification, yield improvement, quality controls, and supply-chain planning.

Speed matters because hyperscalers want to deploy new architectures quickly.

Supply-Chain Management

AI infrastructure contains many specialized components with different lead times and supplier risks.

Celestica must coordinate:

• Switching silicon
• CPUs and controllers
• PCBs
• Connectors
• Power supplies
• Fans and pumps
• Optics
• Cables
• Memory
• Mechanical parts
• Cooling components

The supply chain is not a secondary task. It is part of the product.

Manufacturing Scale

A successful platform must be produced in large volumes without sacrificing quality.

Celestica’s global manufacturing network helps customers regionalize supply, manage tariffs, reduce logistics risk, and scale programs across multiple geographies.

Lifecycle Support

Platform value can extend beyond initial production through testing, repair, logistics, software support, and product lifecycle services.

The combination of these capabilities creates a higher barrier than basic assembly. It does not create an unassailable moat, because competitors can invest in similar capabilities and customers can support multiple suppliers.

06Financial Analysis

6.1 Revenue and Earnings Trend

The direction is clear: revenue, EPS, and operating margin have all improved.

The analytical challenge is determining how much of the improvement is structural and how much is cyclical.

6.2 What the 2026 Revenue Outlook Requires

Celestica reported $4.047 billion of Q1 revenue and guided to $19.0 billion for the full year.

That means the company must generate approximately $14.95 billion in the remaining three quarters, or an average of roughly $4.98 billion per quarter.

That average is about 23% above Q1 revenue.

The math does not mean the outlook is unrealistic. Management expects additional program ramps during the year. It does show that the annual target requires continued sequential growth rather than simply maintaining the Q1 run rate.

Investors should compare each quarter with the implied trajectory, not only with the prior year.

6.3 Margin Quality

Adjusted operating margin reached 8.0% in Q1 2026.

For a traditional EMS company, that level is strong. For a stock valued as an AI infrastructure leader, it is still a relatively thin margin.

This creates an important asymmetry.

A movement from 8.0% to 8.5% can add meaningful operating income because revenue is large. A decline from 8.0% to 7.0% can also remove a large amount of earnings.

For example, on $19 billion of revenue:

• An 8.1% adjusted operating margin implies approximately $1.54 billion of adjusted operating income.
• A 7.1% margin would imply approximately $1.35 billion.
• The difference is approximately $190 million before interest and tax.

The stock therefore has substantial margin sensitivity.

6.4 CCS and ATS Should Be Modeled Separately

A consolidated margin can hide important mix changes.

CCS had an 8.6% segment margin in Q1 2026, while ATS had a 6.0% margin. If CCS continues to grow faster, mix can lift the company margin even without improvement inside each segment.

However, the same mix effect increases dependence on hyperscaler spending and customer concentration.

A strong model should track:

1. CCS revenue growth
2. CCS segment margin
3. ATS revenue growth
4. ATS segment margin
5. HPS revenue
6. Communications revenue
7. Enterprise revenue
8. Customer concentration

6.5 Free Cash Flow and Capital Intensity

Celestica generated $458.3 million of free cash flow in 2025. Management expects approximately $1 billion of 2026 capital expenditures.

Free cash flow already subtracts capital expenditures, so investors should not subtract the $1 billion from a future free-cash-flow figure a second time. The key issue is that higher capex makes cash generation more dependent on operating cash flow and working-capital discipline.

Working capital can move sharply in a fast-growing manufacturing business.

Potential uses of cash include:

• Inventory for program ramps
• Receivables from large customers
• New facilities
• Equipment
• Engineering investment
• Employee growth
• Supplier commitments
• Share repurchases
• Debt repayment

A quarter with strong reported earnings but weak operating cash flow may not invalidate the thesis if the difference reflects temporary customer-backed growth. Repeated weak cash conversion would be more concerning.

6.6 Return on Invested Capital

Celestica reported strong adjusted ROIC in 2025.

ROIC is useful because it asks whether higher earnings justify the capital required to produce them. Rapid revenue growth is less valuable if the business must continually add large amounts of low-return inventory, facilities, and equipment.

The most important ROIC question is forward-looking:

Can Celestica maintain attractive returns after the 2026–2027 capacity expansion enters the asset base?

If new capacity supports durable high-margin programs, returns can remain strong. If capacity utilization falls, depreciation and fixed costs can reduce returns.

6.7 GAAP and Non-GAAP Differences

Celestica reports both GAAP and adjusted metrics.

Investors should understand the major adjustments, including:

• Stock-based compensation
• Amortization
• Restructuring
• Total return swap fair-value adjustments

The total return swap can create volatility in reported GAAP results that does not directly reflect operating performance. Non-GAAP results may therefore provide a cleaner view of operations.

However, adjusted numbers should not be accepted without scrutiny. Stock-based compensation is an economic cost, and repeated restructuring may be relevant to normalized earnings.

A balanced approach is to use adjusted operating metrics for business trend analysis while separately tracking dilution, cash taxes, capital spending, and GAAP cash flow.

07Customer Concentration: The Most Important Risk

Celestica’s customer concentration is unusually high.

In Q1 2026:

• The top ten customers represented 84% of revenue.
• The three largest customers represented approximately 35%, 15%, and 15%.
• Those three customers therefore represented roughly 65% of revenue.
• The largest customers were in CCS.

This concentration has two interpretations.

The Positive Interpretation

Large hyperscalers can support:

• Multi-year programs
• Large production volumes
• Advanced engineering investment
• Global capacity planning
• Fast technology transitions
• Better factory utilization

A successful relationship with a major cloud customer can transform the scale of a supplier.

The Negative Interpretation

The same customers have enormous bargaining power.

They can:

• Reduce forecasts
• Delay a deployment
• Renegotiate pricing
• Add a second source
• Move production
• Redesign a platform
• Change component suppliers
• Insourc parts of the system
• Shift capital spending between technologies

Because a single customer represents about one-third of revenue, even a moderate change can affect the consolidated model.

How to Monitor Customer Concentration

Investors should not assume that lower concentration is always better in the short term.

If the largest customer grows rapidly, concentration may rise while revenue and earnings improve. The problem is the risk embedded in that growth.

A useful monitoring framework is:

Public disclosures generally do not identify every large customer. Investors should avoid presenting rumored customer identities or exact program allocations as confirmed facts unless the company or customer discloses them.

08Product and Technology Analysis

8.1 800G Networking

800G was the product generation that helped establish Celestica’s position in high-speed AI networking.

The DS5000 platform offers 64 ports of 800GbE and 51.2 Tbps of switching capacity. Celestica has also developed other 800G platforms for different network positions and form factors.

The 800G opportunity remains relevant because infrastructure transitions take time. Not every data center moves immediately to the newest generation, and 800G can continue growing while 1.6T begins deployment.

The key question is whether Celestica can protect share as the market broadens.

8.2 1.6T Networking

The DS6000 series represents the next major upgrade.

Key characteristics include:

• Broadcom Tomahawk 6 switching silicon
• Up to 102.4 Tbps of non-blocking capacity
• 64 ports of 1.6TbE
• 224G SerDes
• Support for high-speed copper and optical interconnects
• Air-cooled and hybrid-cooled configurations
• Alignment with open Ethernet and OCP ecosystems

The investment significance is not that Celestica invented the switch ASIC. It did not. The significance is that Celestica is among the system suppliers turning leading-edge silicon into production-ready platforms.

A successful 1.6T ramp would provide evidence that its 800G leadership is transferable to a new generation.

8.3 Co-Packaged Optics

At higher bandwidths, traditional pluggable optical architectures face power, density, and signal-integrity challenges.

CPO moves optical components closer to the switch ASIC. Potential benefits include:

• Lower electrical reach
• Higher bandwidth density
• Lower power per bit
• Better scaling for future systems

Potential disadvantages include:

• More complex manufacturing
• Thermal-management challenges
• Serviceability questions
• Reliability requirements
• Tighter coordination between silicon and optics
• Higher qualification burden

Celestica’s hyperscaler CPO program is important because it places the company in a highly complex integration role. Investors should still treat 2027 production as a forward-looking program rather than recognized revenue.

8.4 Liquid Cooling

Networking equipment is often discussed separately from cooling, but the two are becoming inseparable.

Higher bandwidth and denser systems increase power consumption and heat. Celestica’s hybrid-cooled and liquid-cooled designs show that thermal engineering is becoming part of the networking platform.

Liquid cooling can increase Celestica’s engineering content and system value. It also introduces:

• Leak and reliability risks
• More complex assembly
• Facility compatibility requirements
• New test procedures
• Additional component dependencies

8.5 Scale-Up Versus Scale-Out Networking

Investors should distinguish two network layers.

Scale-up networking connects accelerators within a tightly coupled system or rack. It requires very high bandwidth and low latency.

Scale-out networking connects multiple systems or racks across a larger cluster.

Celestica historically has been most visible in high-speed Ethernet scale-out systems. The AMD Helios collaboration and other rack-scale work could expand its role in scale-up networking.

That expansion is strategically important because scale-up systems can carry higher design value. It is also competitive because NVIDIA, AMD, Broadcom, hyperscalers, and networking vendors are all shaping the architecture.

8.6 SONiC and Open Networking

SONiC is an open-source network operating system widely associated with cloud and data-center networking.

Open networking separates hardware and software choices more than traditional vertically integrated systems. This can benefit platform suppliers by allowing customers to use merchant silicon and standardized hardware.

Celestica can support hardware platforms and customized software capabilities around open networking.

However, open systems may also increase competition and price transparency. The same openness that expands the addressable market can reduce vendor lock-in.

09The AMD Helios Opportunity

AMD Helios is one of the most strategically interesting optionalities in the Celestica thesis.

The platform is intended to combine next-generation AMD Instinct accelerators, CPUs, networking, power, cooling, and rack-level architecture. Celestica is collaborating on scale-up networking switches using an Ethernet-based architecture.

Why Helios Matters

Helios could create a new rack-scale ecosystem outside NVIDIA’s tightly integrated platform.

If hyperscalers want:

• A second accelerator supplier
• Open standards
• Greater control over hardware
• Ethernet-based scale-up networking
• Supplier diversification

then AMD’s platform could gain adoption.

Celestica could benefit because it provides engineering, manufacturing, and supply-chain capabilities that help turn the design into a deployable system.

Why Helios Is Not Yet a Proven Earnings Stream

The opportunity depends on several variables:

• AMD MI450 performance
• Software ecosystem maturity
• Customer qualification
• UALoE development
• Production timing
• Rack reliability
• Customer capital budgets
• Competitive response from NVIDIA and other suppliers

The platform is expected to become available to customers in late 2026. Until production volume and customer adoption become visible, Helios should be treated as a valuable option rather than the base case for current earnings.

10Competitive Landscape

Celestica competes across multiple categories, so there is no single perfect comparable company.

Arista Networks

Arista sells branded networking systems and software and operates with much higher margins.

Arista competes for data-center Ethernet spending, but its business model differs from Celestica’s. Arista captures software, brand, and system economics. Celestica is more heavily involved in open platforms, design manufacturing, and hyperscaler-specific systems.

Arista is therefore a strategic competitor and ecosystem comparison, but not a direct valuation equivalent.

Accton and Edgecore

Accton is one of the more relevant ODM and open-networking competitors.

It has experience in white-box and open networking and can compete for high-speed switch designs. The main competitive variables include engineering speed, customer relationships, supply-chain execution, cost, manufacturing scale, and platform breadth.

Quanta, Wiwynn, and Foxconn

These companies are significant in server, rack, and hyperscaler infrastructure manufacturing.

They may compete more directly as AI infrastructure procurement shifts toward rack-scale systems that combine compute, networking, power, and cooling.

Celestica’s networking platform strength is an advantage. Larger manufacturing peers may have scale advantages or deeper relationships in other system categories.

Jabil and Sanmina

Jabil and Sanmina are broader manufacturing-services comparisons.

They provide context for capital intensity, margins, customer relationships, and the economics of advanced manufacturing. Celestica’s faster AI-driven growth and HPS mix currently differentiate it from traditional EMS peers.

NVIDIA

NVIDIA is not a traditional ODM competitor, but it controls a highly integrated AI networking and compute ecosystem.

NVIDIA’s Spectrum-X and InfiniBand offerings can compete with open Ethernet architectures supported by Celestica. At the same time, Celestica can participate in systems built around NVIDIA or other vendors, depending on the customer program.

Broadcom

Broadcom is a critical silicon supplier and ecosystem partner.

Celestica’s DS6000 uses Broadcom Tomahawk 6. Broadcom captures semiconductor economics, while Celestica captures system-design and manufacturing economics.

The relationship illustrates Celestica’s position: it can benefit from leading silicon without owning the silicon itself.

Competitive Summary

11Does Celestica Have a Moat?

Celestica’s moat is real but narrower than the moat of a dominant semiconductor or software platform.

Elements of the Moat

Engineering Knowledge

High-speed system expertise is accumulated over multiple product generations. Signal integrity, thermal design, manufacturing yield, and optical integration are difficult to master quickly.

Customer Qualification

Hyperscaler qualification creates switching costs. A new supplier must meet demanding performance, quality, capacity, security, and delivery requirements.

Speed to Market

Technology cycles are short. A supplier that can move from silicon availability to a qualified production platform quickly creates economic value.

Global Manufacturing and Supply Chain

Customers require geographic flexibility, component sourcing, capacity, and reliable delivery.

Platform Breadth

Celestica participates in networking, compute, storage, power, cooling, and rack-level solutions. Broader system capability can increase strategic relevance.

Limits of the Moat

Customer Power

Hyperscalers are large enough to negotiate aggressively and fund alternative suppliers.

Merchant Silicon

When multiple vendors use the same switching silicon, product differentiation may narrow.

Open Standards

Open architectures expand Celestica’s opportunity but can also make supplier substitution easier.

Capital Requirements

Competitors with sufficient scale can invest in engineering and facilities.

Limited Control Over End Demand

Celestica does not control cloud workloads, accelerator demand, or hyperscaler capital budgets.

The best description is therefore:

Celestica has an execution moat and qualification moat, not an absolute technology monopoly.

12Valuation Analysis

A strong company can still be a weak investment if the valuation assumes too much.

As of July 15, 2026, CLS traded near $338.78, with a market capitalization of approximately $39.2 billion. Market data changes continuously, so readers should verify current figures before making decisions.

Trailing Valuation

At the market-data snapshot used for this article:

• Trailing EPS was approximately $8.25.
• Trailing P/E was approximately 41 times.

That multiple is far above traditional EMS valuations.

The premium reflects expectations for:

• AI networking growth
• 1.6T ramps
• HPS expansion
• Margin improvement
• 2027 program visibility
• Rack-scale AI opportunities

Forward P/E

Using management’s 2026 adjusted EPS outlook of $10.15:

Formula: Forward P/E = \frac{$338.78}{$10.15} \approx 33.4×

A forward P/E in the low-30s can be justified by rapid earnings growth if that growth is durable.

It becomes difficult to justify if 2026 represents a peak growth year and 2027 earnings flatten.

Price-to-Sales

Using a market capitalization of approximately $39.2 billion and the $19.0 billion revenue outlook:

Formula: Forward P/S = \frac{$39.2B}{$19.0B} \approx 2.1×

A 2-times sales multiple is not extreme for a high-quality technology company. It is high for a business with an 8% operating margin and substantial customer concentration.

The price-to-sales ratio therefore cannot be interpreted without the margin.

Free-Cash-Flow Yield

Using 2025 free cash flow of $458.3 million:

Formula: Historical FCF Yield = \frac{$458.3M}{$39.2B} \approx 1.2%

This historical yield is low because the market expects future earnings and cash flow to be substantially higher.

The 2025 figure is also not a good estimate of normalized 2026–2027 cash flow because revenue, capex, and working capital are changing rapidly.

What the Current Valuation Requires

The valuation becomes more defensible if:

• 2026 revenue reaches or exceeds $19 billion.
• Adjusted EPS reaches or exceeds $10.15.
• CCS margin remains at least in the mid-8% range.
• 2027 revenue continues growing strongly.
• 1.6T and CPO programs ramp on schedule.
• Free cash flow improves after the capacity build.
• Customer concentration does not increase materially.
• The company avoids major pricing concessions.

The valuation becomes vulnerable if:

• 2027 growth slows sharply.
• Operating margin falls below 7.5%.
• A major customer reduces forecasts.
• Capex and working capital suppress cash flow longer than expected.
• 1.6T ramps are delayed.
• The market reclassifies CLS from AI infrastructure leader back toward an EMS multiple.

13Bear, Base, and Bull Scenarios

These scenarios are analytical illustrations, not price targets or forecasts.

Bear Case

Assumptions:

• 2026 revenue reaches only the low end of expectations.
• 2027 growth slows to low single digits.
• Adjusted operating margin falls toward 7.0%–7.3%.
• A major hyperscaler moderates spending or adds a second source.
• 1.6T and CPO programs ramp more slowly.
• Capex remains elevated while utilization disappoints.
• The market applies a mid-teens to low-20s earnings multiple.

What would cause it:

• AI infrastructure spending digestion
• Program loss or forecast reduction
• Pricing pressure
• Component or facility delays
• Weak Helios adoption
• Ethernet share loss
• Greater competition from ODMs

The bear case does not require AI demand to disappear. It only requires growth and margin to fall below the assumptions embedded in the stock.

Base Case

Assumptions:

• 2026 revenue is near $19 billion.
• Adjusted EPS is near management’s $10.15 outlook.
• CCS remains the main growth engine.
• 1.6T ramps contribute in the second half and into 2027.
• 2027 revenue grows at a healthy but slower rate.
• Adjusted operating margin remains around 8%.
• Free cash flow improves as expansion spending normalizes.
• The stock retains a premium technology-manufacturing multiple.

The base case assumes Celestica has structurally improved but does not become a semiconductor-like margin business.

Bull Case

Assumptions:

• 2026 results exceed current guidance.
• 2027 demand visibility converts into another large revenue increase.
• Multiple 1.6T platforms ramp successfully.
• CPO and liquid-cooled systems create higher engineering content.
• AMD Helios wins meaningful deployments.
• AI/ML compute and networking both expand.
• CCS margin moves toward or above 9%.
• Free cash flow rises sharply after the capacity build.
• Customer diversification improves.
• The market values Celestica as a durable AI systems platform.

The bull case requires more than strong demand. It requires Celestica to retain economic value as customers and chip suppliers negotiate over the profit pool.

14Key Catalysts

Q2 2026 Earnings

Management guided to Q2 revenue of $4.15 billion to $4.45 billion and an adjusted operating margin of approximately 8.0% at the midpoint.

The most important questions will be:

• Is revenue near the high end?
• Does CCS remain the primary driver?
• Is HPS revenue approaching or exceeding $2 billion?
• Does margin remain at or above 8%?
• Is annual guidance raised again?
• Is 2027 visibility becoming more specific?

1.6T Production Ramps

Product availability is not the same as volume production.

Evidence of customer qualification, program ramps, shipments, and 2027 orders would strengthen the thesis.

New Hyperscaler Design Wins

Additional 1.6T, CPO, AI compute, or rack-scale programs could broaden the revenue base and improve visibility.

AMD Helios Availability

Late-2026 availability would validate the project timeline. Customer deployments and production orders would be more important than the initial launch.

Margin Expansion

A sustained CCS segment margin above 8.5% would suggest that HPS mix and operating leverage remain strong.

Free-Cash-Flow Improvement

Investors will eventually expect earnings growth to convert into cash after the major capacity build.

Customer Diversification

New customers or broader program distribution would reduce the risk attached to the largest hyperscaler relationship.

15Major Risks

15.1 Customer Concentration

The largest customer represented approximately 35% of Q1 2026 revenue, and the top three represented about 65%.

This is the most important risk.

15.2 Hyperscaler Capital-Expenditure Cycles

Celestica’s growth depends on customers continuing to invest in AI and cloud infrastructure.

A spending pause can affect multiple programs at once.

15.3 Pricing Pressure

Large customers can negotiate lower prices as volumes scale or add competing suppliers.

Revenue growth does not guarantee margin stability.

15.4 Program-Ramp Risk

New systems can experience:

• Qualification delays
• Component shortages
• Yield problems
• Software issues
• Cooling problems
• Facility delays
• Customer design changes

15.5 Technology Transition Risk

Celestica must execute across rapidly changing technologies:

• 1.6T Ethernet
• 224G SerDes
• LPO
• CPO
• Liquid cooling
• Scale-up Ethernet
• Rack-scale AI

Missing a generation can damage share and customer relevance.

15.6 Dependence on Semiconductor Vendors

Celestica depends on switching silicon, CPUs, accelerators, controllers, and other components that it does not control.

Supply limitations or roadmap delays can restrict revenue.

15.7 Competitive ODM Risk

Hyperscalers often prefer multiple suppliers.

A competitor can win second-source share or displace Celestica in a future generation.

15.8 Capex and Utilization Risk

Approximately $1 billion of expected 2026 capex is large relative to Celestica’s historical spending.

If demand slows, utilization and returns can fall.

15.9 Working-Capital Risk

Rapid growth can consume cash through inventory and receivables.

Accounting earnings may grow faster than free cash flow.

15.10 Margin Reversion

An 8% adjusted operating margin is a major improvement but remains vulnerable to mix and pricing.

A one-percentage-point decline would materially reduce earnings.

15.11 AMD Helios Execution Risk

Helios depends on AMD’s accelerator competitiveness, software, networking standards, and customer adoption.

15.12 Open Networking Is Not Automatically High Margin

Open systems expand supplier opportunity but can increase commoditization and price competition.

15.13 Geopolitical and Tariff Risk

Celestica operates a global manufacturing network and depends on international component flows.

Tariffs may be recoverable from customers, but operational complexity remains.

15.14 Power and Data-Center Constraints

AI infrastructure deployments can be delayed by grid connections, transformers, cooling, water, permits, and construction.

Celestica can have customer demand without immediate site readiness.

15.15 Valuation Risk

The stock has already been substantially re-rated.

Even strong results may not produce strong returns if they only meet elevated expectations.

16What Would Invalidate the Bullish Thesis?

A useful investment thesis must include explicit failure conditions.

The CLS thesis would weaken materially if several of the following occur:

1. Revenue misses guidance for multiple quarters.
2. 2026 revenue guidance is reduced below $19 billion.
3. Adjusted operating margin falls below 7.5% without a clear temporary cause.
4. CCS segment margin falls below 8%.
5. HPS revenue stops growing despite strong industry AI spending.
6. 1.6T programs fail to enter meaningful production by 2027.
7. The CPO program is delayed substantially beyond the announced schedule.
8. AMD Helios availability or customer adoption slips materially.
9. The largest customer exceeds 40% of revenue while diversification weakens.
10. A large customer insources or moves a major program to another ODM.
11. Free cash flow remains weak after the 2026 capacity build.
12. New facilities operate below expected utilization.
13. Inventory and receivables rise much faster than revenue.
14. The company requires materially more debt to fund growth.
15. The market shifts toward architectures where Celestica has limited participation.

One weak quarter does not automatically invalidate the thesis. A combination of slowing growth, declining margins, delayed programs, and poor cash conversion would be more significant.

17Quarterly Monitoring Dashboard

18Celestica Versus Other AI Infrastructure Investments

Celestica occupies a different part of the value chain from chip, networking-software, and data-center-equipment companies.

Celestica can be attractive for investors seeking AI infrastructure exposure beyond semiconductors. It should not be expected to generate semiconductor-level margins.

For adjacent research, see:

• Best AI Infrastructure Stocks
• Best Data Center Stocks
• Best AI Chip Stocks
• SK hynix HBM Deep Research
• Best Power Grid Stocks

19How to Analyze Celestica Earnings

Investors should read each earnings report in the following order.

Step 1: Check Revenue Against Guidance

Do not focus only on year-over-year growth. Compare actual revenue with the company’s prior guidance and the run rate required for the annual target.

Step 2: Separate CCS and ATS

Determine whether growth comes from networking, AI compute, storage, or other businesses.

Step 3: Check HPS Revenue

HPS growth indicates whether higher-value platform programs are scaling.

Step 4: Analyze Margin Before EPS

EPS can be affected by tax, share count, and adjustments. Operating and segment margins reveal the quality of revenue.

Step 5: Review Customer Concentration

A strong quarter driven by one customer can improve earnings while increasing risk.

Step 6: Read Cash Flow and Working Capital

Track inventory, receivables, payables, customer advances, operating cash flow, and capex.

Step 7: Review Program Language

Look for changes in wording around:

• 1.6T ramps
• CPO
• AI/ML compute
• Hyperscaler demand
• Helios
• 2027 visibility
• Capacity expansion

Step 8: Recalculate Valuation

Update forward P/E, free-cash-flow yield, and scenario assumptions using the new guidance.

20Frequently Asked Questions

Is Celestica an AI stock?

Celestica is an AI infrastructure supplier rather than an AI software or semiconductor company. Its AI exposure comes primarily from high-speed data-center networking, AI/ML compute platforms, storage, rack integration, power, cooling, and manufacturing for hyperscalers and technology customers.

Why has Celestica stock risen so much?

The company has reported rapid revenue growth, expanding margins, strong HPS demand, hyperscaler program wins, and increasing exposure to 800G, 1.6T, AI compute, and rack-scale infrastructure. Investors have re-rated Celestica from a traditional EMS valuation toward an AI infrastructure valuation.

Is Celestica a manufacturer or a technology company?

It is both. Celestica still operates a capital-intensive manufacturing and supply-chain business, but a growing share of its value comes from engineering, platform design, system integration, software support, and product industrialization.

What is Celestica’s CCS segment?

Connectivity & Cloud Solutions includes the company’s communications and enterprise businesses. It contains networking, optical, server, storage, and AI/ML compute activity and is currently Celestica’s main growth engine.

What is Celestica’s ATS segment?

Advanced Technology Solutions includes aerospace and defense, industrial, HealthTech, and capital equipment. ATS provides diversification and engineering exposure outside data centers.

What is HPS revenue?

Hardware Platform Solutions revenue comes from Celestica-designed or co-designed platforms in networking, compute, and storage. HPS is not a separate reportable segment, but it is important because it generally contains more engineering and platform value than traditional manufacturing.

Why are 1.6T switches important for Celestica?

AI clusters require more network bandwidth as accelerator counts and model sizes grow. The transition to 1.6T creates demand for new switch platforms with higher bandwidth, advanced optics, better cooling, and more difficult system engineering.

Does Celestica make networking chips?

No. Celestica designs and manufactures systems using silicon from suppliers such as Broadcom and other technology vendors. It captures platform and manufacturing economics rather than semiconductor economics.

What is the Celestica CPO opportunity?

Celestica has received a hyperscaler program for a co-packaged-optics Ethernet switch using 1.6T switch silicon and liquid cooling. Production is expected to ramp in 2027. CPO could increase Celestica’s engineering value but remains a forward-looking program.

What is AMD Helios?

Helios is AMD’s rack-scale AI platform. Celestica is collaborating on scale-up networking switches intended to connect next-generation AMD Instinct accelerators through an Ethernet-based architecture. Customer availability is planned for late 2026.

Who are Celestica’s largest customers?

Celestica discloses customer concentration but generally does not publicly identify every large customer. In Q1 2026, the three largest customers represented approximately 35%, 15%, and 15% of total revenue. Rumored identities should not be treated as confirmed unless officially disclosed.

What is the biggest risk to CLS stock?

The biggest operating risk is customer concentration. The biggest stock risk is that the valuation assumes continued rapid growth, strong 2027 demand, and stable margins. A customer slowdown or margin decline could cause both earnings estimates and the valuation multiple to fall.

Is an 8% operating margin strong?

It is strong relative to Celestica’s historical EMS profile. It is still thin compared with semiconductor and software companies. Because the revenue base is large, small changes in margin can materially affect earnings.

Is Celestica’s growth capital intensive?

Yes. The company expects approximately $1 billion of 2026 capital expenditures to support expansion. Investors should monitor utilization, working capital, operating cash flow, and returns on the new capacity.

Is CLS stock cheap?

CLS is no longer cheap on traditional EMS metrics. At the July 15, 2026 market snapshot, the stock traded at roughly 33 times management’s 2026 adjusted EPS outlook. Whether that is attractive depends on 2027 growth, margin durability, cash conversion, and program execution.

Could Celestica margins reach software or semiconductor levels?

That is unlikely under the current business model. Celestica can improve margins through HPS mix, engineering value, and scale, but it remains a manufacturing and systems-integration business serving powerful customers.

Does Ethernet need to beat InfiniBand for Celestica to succeed?

No. Celestica does not require Ethernet to eliminate InfiniBand. It needs the market for high-performance Ethernet AI fabrics to grow enough to support its programs. Multiple networking architectures can coexist.

What would be the strongest bullish signal?

The strongest combination would be continued revenue beats, CCS margin above 8.5%, clear 1.6T volume ramps, quantified 2027 growth, improving free cash flow, and broader customer diversification.

What would be the clearest warning signal?

A reduction in annual guidance combined with margin below 7.5%, delayed program ramps, rising customer concentration, and weak cash conversion would be a significant warning.

21Final Assessment

Celestica has earned its re-rating as more than a traditional electronics manufacturer.

The company has built a strategically valuable position between leading semiconductor roadmaps and hyperscaler deployment requirements. Its ability to design, integrate, industrialize, and manufacture complex networking and AI systems has become more important as data centers move toward 800G, 1.6T, liquid cooling, co-packaged optics, and rack-scale architectures.

The financial evidence is strong:

• Q1 2026 revenue increased 53%.
• CCS revenue increased 76%.
• HPS revenue increased 63%.
• Adjusted operating margin reached 8.0%.
• Management raised 2026 revenue guidance to $19.0 billion.
• Management raised adjusted EPS guidance to $10.15.

The investment risk is equally clear.

Celestica remains dependent on a small number of powerful customers. It operates with margins that are strong for manufacturing but thin relative to the premium valuation. It is investing approximately $1 billion in capacity before the full 2027 demand profile is proven. It also depends on technologies and roadmaps controlled by chip vendors, cloud companies, and networking ecosystems.

The most defensible conclusion is not that CLS is simply cheap or expensive.

It is that Celestica has become a high-growth AI systems platform whose valuation now requires execution at a much higher level than its historical business. Investors should focus on the durability of CCS growth, 1.6T production ramps, segment margins, customer concentration, and cash conversion rather than relying on the broad AI narrative alone.

A strong 2026 is largely expected.

The decisive question is whether Celestica can convert its current design wins and capacity expansion into a durable 2027–2028 earnings base without losing margin or bargaining power.

Primary Sources

1. Celestica Q1 2026 Financial Results
2. Celestica Q1 2026 Form 10-Q
3. Celestica FY2025 and Q4 2025 Financial Results
4. Celestica 2025 Annual Report
5. Celestica 2025 Investor and Analyst Day Presentation
6. Celestica DS6000-Series 1.6TbE Announcement
7. Celestica and AMD Helios Collaboration
8. Celestica Networking Platform Overview
9. Celestica Quarterly Results
10. Celestica Investor Relations

Editorial Note

This article distinguishes among three evidence levels:

• Reported facts: figures and program details disclosed by Celestica or an official partner.
• Management outlook: forward-looking guidance and expected production timelines that remain subject to execution.
• SnowballHare analysis: interpretations, calculations, valuation frameworks, scenario analysis, and monitoring thresholds developed from the reported evidence.

Customer identities, customer-specific revenue, product allocation, and future production volume are not treated as confirmed unless disclosed by the relevant company.

Market prices and valuation multiples change continuously. The price and market-capitalization snapshot in this article is dated July 15, 2026 and should be updated before publication if the page goes live later.

This material is for education and research only. It is not personalized investment advice, a recommendation to buy or sell securities, or a guarantee of future performance. Investing in individual stocks involves the risk of permanent capital loss.