The factory can be outsourced. The economics can’t.
Lets first get hold the term: Contract manufacturing is a business arrangement in which a company outsources the production of its products to a third-party manufacturer. The contract describes a relationship, not a business model. Depending on manufacturer’s contribution to the product, the economics can look completely different, from low-margin electronics assembly to high-value semiconductor fabrication and pharmaceutical CDMO services.
When the supply chains seized up during the pandemic of COVID-19, the situation exposed a fragile reality hidden in plain sight. Consumers observed that the brand on the product is rarely the company making. Investors realised the harsh truth that one missing part from an unknown supplier in remote part of the world could shut down a global production supply chain.
Behind every finished product lay many of invisible companies buried in a bill of materials. This fragmented, highly vulnerable chain was not an accident of modern manufacturing. As Part 1 argued, it is precisely the model the world deliberately built over the last few decades.
If we look at everyday products such smartphones, or a pack of daily medicine the brand on the packaging is always at the front and center, but the company that makes the product is hidden under the leaf.
This creates a tricky problem for anyone analyzing the value chain: two companies may both make products for a third party, but have very different underlying economics.
One company might simply be assembling the components to a customer’s blueprint. Another might engineer the product, invent the manufacturing process, or own the underlying intellectual property.
Some even manufacture drugs where their proprietary process is baked directly into the customer’s FDA filings
Technically, all of them are contract manufacturers. Economically, they are different businesses.
Fundamentally:
The Customer owns the product.
The Manufacturer owns the production.
Lets understand this with some examples
For Hon Hai Technology Group (Foxconn) reported revenue of around US$260 billion for FY2025, a gross margin of 6.15% and an operating margin of 3.20%. Net profit was roughly US$6 billion.
For same year, Taiwan Semiconductor Manufacturing Company (TSMC) reported revenue of US$122.4 billion, less than half of Foxconn’s but having net income of US$55.2 billion.
TSMC earned roughly 9 times Foxconn’s profit on under half its revenue.
Both are contract manufacturers. Both build products they do not own or sell. Neither appears on the box. And yet TSMC captured about 45 cents of every dollar, while the Foxconn only a little over two.
Understanding why so is the subject of this report. To get there, we first have to be precise about who is actually standing on each side of the contract
1. OEM – The Customer
To untangle the modern supply chain, we must start with the someone whose name is always on the product: called the OEM, or Original Equipment Manufacturer in electronics, consumer goods and industrial equipment world.
In the automobile world, “OEM” represents the carmaker, the brand that designs the vehicle and sells it.
In electronics the “OEM” term is used to describe the supplier that builds equipment for someone else to rebrand, which is very nearly the opposite.
Bosch is often called an “OEM supplier” to companies who are themselves called “OEMs.”
The label OEM is confusing, therefore we should look strictly at the economic role.
The OEM is the company that owns the product proposition.
OEM is the architect of product.
It decides what product needs to exist, outlines the exact specifications, and owns the relationship with the final buyer.
Dell, for example, dictates a laptop’s performance, design, and shelf price, but completely outsources the physical assembly.
This model is ubiquitous. A automaker sets the vehicle’s architecture and brand strategy while sourcing individual components from specialist suppliers. The badge on the bonnet rarely belongs to the company that manufactured the parts underneath it.
The modern OEM therefore has decoupled from factory floor involved in making its product, its economic function is to decide what should be made, how it should perform, and who ultimately buys it.
Which leaves a different question for somebody else to answer: How do we actually make this?
2. Build to Spec: The Simplest Contract
The baseline arrangement is brilliantly simple: A customer provides the product with a design, a specification and quality standards.
The manufacturing partner provides a factory, a workforce, equipment, procurement relationships and production expertise. They sign a contract.
The customer requests: here is what we need; build it for us, at this scale.
The manufacturer does not own the brand neither it dictates the design, rarely choses the components which goes in, because the customer has often already qualified and approved them. Its job is to convert a specification into physical output consistently, at agreed scale, and to a defined quality tolerance.
Distilled to its essence, Contract Manufacturing is a service business dressed in industrial clothing.
What makes contract manufacturing interesting is how much the same basic skeleton can vary once you examine who takes responsibility for what:
Procurement: The manufacturer may simply assemble customer-supplied components, or it may also take responsibility for sourcing them.
Engineering: The manufacturer may operate purely as a production partner, or contribute to the design and engineering of the product.
Process: The manufacturer may follow a customer-defined production process, or develop and optimize the processing techniques itself.
Capital: The manufacturer may use customer-owned tooling and equipment, or invest in and own the critical tooling, equipment, or proprietary technology required for production.
Regulatory: The manufacturer may simply produce to the customer’s specifications, or play an active role in securing and maintaining the customer’s regulatory approvals.
Switching Costs: The relationship may be relatively easy to move to another factory, or become difficult to replicate elsewhere because of embedded tooling, process know-how, qualifications, regulatory approvals, and accumulated expertise.
The subjects determine the economics of the business far more reliably than the words printed in of the contract.
3. EMS: When Manufacturing Becomes a Service
The baseline model stays intuitive until we apply it to modern electronics, at which point the complexity becomes almost absurd and the industry responds with a specialised structure called EMS, or Electronics Manufacturing Services.
An EMS provider sits at the narrow end of an extraordinarily wide supply chain. A smartphone contains hundreds of discrete components sourced from suppliers across dozens of countries and the make of the phone – EMS – provider may procure those parts, finance and manage inventory, operate highly automated surface-mount and assembly lines, test the finished device, package it, and ultimately ship it into the customer’s distribution network.
The operational demands for EMS players are therefore extensive. Foxconn, for example, operates more than 240 campuses across 24 countries and can employ roughly one million people during peak seasons. And yet for the enormous amount of work Foxconn performs, it captures remarkably little of the product’s final price.
The arithmetic is easiest to see in a device everyone knows. Check the teardown of iPhone 17 Max Pro (512 GB) to see the breakdown of value capture between brand, providers and assemblers.
What the EMS provider genuinely contributes is the conversion: the transformation of those procured inputs into a finished, tested, packed product. That conversion is difficult, capital-hungry and unforgiving of error. It is also, in economic terms, razor thin. The metrics are confusing to a naive investors when the first encounter an EMS business.
Lets take a look at Dixon Technologies (NSE:DIXON), India’s most prominent EMS player. In FY26, Dixon earned ₹48,893 crore in revenue. But, its net profit for the year was just ₹845 crore, about 1.7%.
That means for every ₹100 of product that moved through its factories, Dixon earned roughly ₹2. That is not an execution error. That is the exact structure of this business model.
Revenue in an EMS company is, to a significant degree, a measure of throughput. It tells us how much product moved through the factory, and not how much economic value the manufacturer retained. Because components dominate the cost base, revenue can move for reasons that have nothing to do with the manufacturer’s own performance.
Riding on the same Dixon example, in March 2026, the company came under pressure as chip prices pushed input costs higher while consumer demand for mobile and IT hardware remained soft. When component prices go up, it mechanically inflate the revenue, but profit margins shrink. This does not necessarily mean the company is making more profit.
This points to a broader issue in India’s EMS growth story: a meaningful portion of current profitability is policy-derived rather than contract-derived.
Our friend – Dixon has disclosed more than ₹1,100 crore of PLI income awaiting formal disbursement. The government’s Production Linked Incentive (PLI) scheme is a deliberate industrial policy designed to accelerate electronics manufacturing in India, and it has worked. But the support is also time-bound.
For investors, this difference matters. A company may report strong profits because its manufacturing contracts are genuinely profitable, or because the government is temporarily supporting its economics through subsidies. These are two very different reasons for making money.
That is why comparing EMS companies based only on revenue multiples can be misleading. Two companies may have similar revenues but very different working-capital needs, and dependence on government incentives.
4. ODM: When the Manufacturer Designs.
Now when change one clause in the contract where instead of handing over a complete design blueprint, the customer asks: we want a device that does these things, at this price point, by this date. You work out how to build it.
The manufacturer is no longer only executing instructions, It is engineering the solutions. This is the ODM, or Original Design Manufacturer model, and the distinction matters because design sits considerably closer to the economic centre of a product than assembly does.
In this model, the manufacturer designs the circuit board, qualifies the components, and engineers the thermal solutions. The brand still owns the customer. But the manufacturer now owns the knowledge physically embedded in the product.
The clearest historical demonstration is the Taiwanese notebook industry. Quanta, Compal, Wistron, Inventec and Pegatron did not merely assemble the world’s laptops; they progressively absorbed the engineering required to design them. The result is one of the most concentrated manufacturing structures in modern industry.
Even today, Taiwanese ODMs account for roughly 80–90% of global notebook production, with Quanta at around 22% share and Compal near 18%. A laptop brand today can decide what it wants a machine to do without maintaining most of the engineering capability needed to make one.
This is why moving to “ODM” is so often pitched as the natural, highly profitable next rung for an EMS company.
The pitch however needs a caveat: Taking on more responsibility does not automatically produce better economics.
Engineering capability requires engineers, laboratories, tooling, development spend and technical infrastructure – all of which arrive as cost before any of it arrives as pricing power.
Customers with scale generally negotiates that value away. Designs eventually standardizes with time Trade secrets leaks out in public. And the manufacturer often has to build capacity ahead of the contract that is meant to justify it.
The Taiwanese notebook ODMs we discussed, won the market as they achieved near-total dominance over a global product and hold decades of design IP.
Yet, for most of their history, they were stuck with low single-digit margins. They were trapped in a room with four highly capable suppliers on one side of the table, and only one mega-buyer on the other. What finally changed their fortunes wasn’t their design skill, which they had all along. It is the latest boom in AI, where factory capacity itself became scarce.
That is the lesson. Moving from assembly to design changes how much the manufacturer contributes. Whether shareholders capture any value out of it depends on scarcity, not on sophistication.
5. JDM: When Design Is a Conversation
Between pure assembly and full design sits a hybrid arrangement: JDM (Joint Design Manufacturing).
In this model the customer brings the market requirements, brand positioning, and overall product architecture. The manufacturing partner brings the detailed engineering, component selection, and tooling. The finished product emerges from iterative collaboration, not a clean hand-off.
In practice, however, the boundaries between EMS, ODM, and JDM are highly porous. Companies use the terminology inconsistently in the direction that most flatters their valuation. A manufacturer might legitimately do ODM work for 15% of its business while running pure EMS assembly for the rest.
The key question is not whether the company calls itself an ODM but what the manufacturer actually contributes that the customer cannot easily replicate or replace ,and how difficult, costly and time-consuming that replacement would be.
That distinction takes us beyond the factory floor. Because once the manufacturer’s contribution includes proprietary process knowledge, regulatory expertise or specialised scientific capability, replacing the supplier can become far more complicated than moving production from one assembly line to another.
And that is where the economics of contract manufacturing begin to change materially.
6. CDMO: When the Factory Starts Solving the Science
A pharmaceutical manufacturing relationship starts from an entirely different problem.
A drug company may have discovered a molecule, however, it is not remotely the same thing as being able to manufacture it reliably, economically and legally at commercial scale.
The chemistry has to be translated into a repeatable process. Yields must be improved, often by orders of magnitude. Impurity profiles have to be characterised and controlled. Analytical methods need to be developed and validated. Production has to travel from grams in a laboratory to tonnes in a reactor without changing the product. And every stage has to be documented to a standard that will survive inspection by regulators in multiple jurisdictions.
A specialist partner that can take on some or all of that work is called a CDMO (Contract Development and Manufacturing Organisation). The word development is doing the heavy lifting.
A CDMO is not simply renting out reactor capacity.
It contributes process chemistry and scale-up expertise that helps convert a customer’s molecule into something that can actually be manufactured commercially.
This creates a fundamentally different relationship from an electronics assembly contract, and the difference shows up most when the customer wants to leave.
If an electronics brand is unhappy with an assembler, it can often qualify an alternative within a few quarters. Switching a pharmaceutical site is a nightmare. It demands technology transfers, new process validations, comparability trials, and deep regulatory reviews before a single new dose can be sold.
The manufacturing partner is no longer merely a supplier. It has become part of the product’s industrial and regulatory history.
In 2024 when demand for its GLP-1 medicines – Ozempic and Wegovy running ahead of available supply, Europe’s most valuable pharmaceutical company Novo Nordisk decided the safest option was to buy the its CDMO partner Catalent for US$16.5 billion outright rather than risk losing the capacity. Novo Nordisk (NYSE:NVO) paid US$11 billion for three Catalent fill-finish sites in Italy, Belgium and Indiana, facilities that it was already using as a customer.
The fill-finish, the sterile filling and sealing of injectable medicines, is a long way from discovering a drug in a laboratory.
The transaction also attracted public objections from competitors, including Eli Lilly and Roche. The arguments, and the deal made one thing clear: specialised manufacturing capacity can become strategically important enough that the customer would rather own the factory than risk losing access to it.
The Indian equivalent of that capability, if not that transaction, is Divi’s Laboratories (NSE:DIVISLAB). In FY26, Divi’s reported revenue of ₹10,388 crore and an operating profit of ₹3,489 (33.59%). The net profit margin was roughly 25.10% – a net profit of ₹2,607 crore.
Exports accounted for 88% of revenue, while custom synthesis represented roughly 53% of the product mix
Those margin is not a pricing accident. It is the reflection of decades of process chemistry, backward integration, regulatory experience and relationships around specific molecules that can take years to build.
That is the distinction between a factory and a capability.
However – this does not make CDMO a perfect business. Pharmaceutical manufacturing suffers from brutal pricing cycles, regulatory risks that can shut down a plant overnight, and heavy capital requirements that never quite disappear.
Where EMS provider is primarily being paid to execute a manufacturing process, a CDMO can become part of the process itself. The longer the relationship persists, the more technical knowledge, validation work and regulatory history can become embedded between the customer and the manufacturing site.
And once that happens, replacing the factory is no longer a procurement decision. It becomes a development decision, a regulatory decision and, ultimately, a business-risk decision.
7. CRDMO: When the Relationship Begins Before the Product Exists
The next step extends the relationship backwards in time.
What happens when a customer doesn’t merely need help manufacturing a drug, but also discovering and developing one?
We will now discuss CRDMO, Contract Research, Development and Manufacturing Organisation.
A CRDMO can participate across a far longer stretch of the pharmaceutical value chain: early discovery chemistry and biology, lead optimisation, process development, analytical development, clinical-phase manufacturing and eventually commercial supply. The relationship may begin five or ten years before there is a commercial product to manufacture at all.
Syngene International is India’s prime example in this category, spanning discovery through development and manufacturing across small molecules and biologics, with its business split roughly 59% research services and 41% CDMO. The scale of the regulatory apparatus that sits underneath this is easy to miss: Syngene cleared 85 client and regulatory audits across FY26.
That accumulated position changes what outsourcing means. The customer is not renting a factory; it is buying scientific capability. And the manufacturer is not merely delivering physical output; it is accumulating years of knowledge about a specific molecule, a specific process and a specific development programme. Switching costs of that kind are among the most durable in industrial business.
CRDMO business is a form of concentration risk wearing a nicer suit.
Syngene’s FY26 report proves the point. Revenue grew just 3% to ₹3,739 crore, while operating EBITDA declined 12% to ₹918 crore and the operating EBITDA margin fell from 29% to 25%.
Management attributed much of the pressure to a destocking cycle at a large biologics customer, alongside higher employee costs and the operating expense of newly commissioned capacity that had not yet reached full utilisation.
A deep, sticky, multi-year, hard-to-replace relationship is a wonderful asset right up until the moment that customer’s product stalls. The same integration that makes you hard to remove makes you impossible to replace quickly on your own side of the table.
That is the unvarnished reality of the CRDMO model. It can yield superior economics, but the manufacturer has to bleed for them through continuous scientific execution, flawless regulatory audits, and aggressive customer diversification. The acronym alone guarantees nothing.
8. Private Label: Owning the Product, Not the Customers
The same structural separation appears in a completely different part of the economy, and at a scale most people underestimate.
Walk through a supermarket. The brand on a packet of biscuits, a bottle of shampoo or a floor cleaner frequently does not own the plant that made it. In many cases the manufacturer produces to the retailer’s specification while the retailer owns the brand, the shelf, the price point and the customer.
This is private-label manufacturing, and it is no longer a discount niche.
In Europe, store brands are no longer just a “cheap alternative”, they are becoming the market standard.
Massive Market Share: Across 17 European countries, nearly $39 out of every $100 spent on groceries goes to private label products.
In eight of those countries, store brands make up more than 40% of the market. Switzerland leads the pack at 52.4% (more than half of all grocery sales), followed closely by the UK at 44.3%.
Private label sales grew by 4.1%, while traditional manufacturer brands like Nestlé or Unilever only grew by 2.7%. This means store brands are expanding at nearly double the rate of major brands. n the US, store brands account for roughly 20% (a fifth) of all grocery sales.
More tellingly, the proposition is no longer simply about being cheaper: 47% of surveyed shoppers said price was no longer the primary reason they chose own-brand products, while 52% considered their quality equivalent to branded alternatives.
The model exists because retailers and brand owners want to control their price architecture and category range without having to build a factory for every product they sell.
Hindustan Foods sits inside this ecosystem in India, manufacturing across home care, personal care, foods and beverages for both marquee FMCG customers and emerging brands.
The economics here are different again from electronics. There is far less technological complexity than in a semiconductor fab or a peptide plant.
What matters instead is capacity utilisation, geographic footprint relative to customer distribution, category breadth, customer diversification, and the discipline to keep working capital and interest costs from consuming the operating profit.
Hindustan Foods has invested over ₹700 crore in capex during its recent expansion, and elevated interest costs and working-capital pressure have been a recurring theme in its results even as profit has grown steadily.
A manufacturer of toothpaste, detergent and packaged foods does not need to own a consumer brand to build a large and durable business.
It also does not automatically receive the economics of one.
That distinction – between owning the product and owning the customer -recurs through every model in this series.
9. The Foundry: When Manufacturing Technology Becomes the Moat
Semiconductors push the outsourcing model to its logical extreme, and then completely invert it.
In most industries, a company that becomes unhappy with its manufacturing partner can, at least in principle, build a factory of its own. That option is not realistically available when the factory is a leading-edge semiconductor fab.
The capital requirement runs into tens of billions of dollars, while yield management, process integration, equipment optimisation and continuous R&D compound over decades. You cannot simply buy those capabilities off the shelf and install them in a new building.
Out of that reality came the pure-play foundry.
When TSMC was founded in Hsinchu in 1987 under Morris Chang, it radical promise that it would only fabricate chips designed by its customers and would never design, manufacture or market semiconductor products under its own name.
That single constraint is what made the global fabless industry possible. Nvidia, Qualcomm, Apple Silicon, AMD and hundreds of smaller design houses exist in their current form because someone credible promised never to become their competitor.
The scale of what that promise now supports is difficult to hold in the head.
In 2025, TSMC deployed 305 distinct process technologies to manufacture 12,682 different products for 534 customers, across capacity exceeding 17 million 12-inch equivalent wafers.
But the crucial difference between TSMC and a conventional assembler is not scale. It is what the customer is buying.
TSMC’s advantage is not that it owns factories. It is that it owns manufacturing technology at a level no one has been able to replicate. Its process nodes, accumulated yield learning, packaging capability and the pace at which it can move a new generation into volume production are themselves the product.
A fabless designer is not renting capacity; it is buying access to a manufacturing capability that determines whether its design is commercially viable at all.
The financial signature of that is unmistakable. TSMC’s 2025 gross margin was 59.9%, operating margin 50.8% and net margin 45.1%.
Compare that with Foxconn’s 6.15% gross and 3.20% operating margin in the same year. Both companies build products designed by others. One of them owns something no customer, and so far no competitor, can substitute.
Both companies manufacture products designed by others. The difference is what sits behind the factory.
For Foxconn, the competitive advantage is primarily scale, execution, procurement and manufacturing efficiency. For TSMC, the manufacturing process itself has become a scarce technological asset.
The fact that both Intel and Samsung possess enormous capital and have still struggled to close the gap is the strongest available evidence that the moat is not the money.
The moat is not the money. It is everything that has been learned inside the factory.
This is the contract-manufacturing paradox in its purest form. The customer owns the design. The manufacturer owns the ability to realise it. And the second turns out to be scarcer than the first.
10. One Phrase, Many Businesses
By now the flaw with the phrase “contract manufacturing” should be clear.
It describes a commercial relationship, not a business model. It is exactly like using the word “employee” – a legal term shared equally by a warehouse picker and a chief executive.
| Model | Model | What the customer owns | What the manufacturer contributes | What creates economic advantage |
|---|---|---|---|---|
| OEM | Product owner | Product proposition, brand, customer relationship, often the design | May outsource all manufacturing | Brand, ecosystem, pricing power |
| EMS | Factor Execution | Product design and specifications | Procurement, assembly, testing, execution | Scale, throughput, working-capital efficiency |
| ODM | Engineering Capabilities | Product requirement and brand | Design, engineering and manufacturing | Engineering depth – if it is scarce |
| JDM | Shared Design and Engineering | Product requirement, shared development | Joint design, engineering, manufacturing | Depends entirely on contract terms |
| CDMO | Process Knowlege | Molecule and product concept | Process development and manufacturing | Process know-how, regulatory history |
| CRDMO | Research and Scientific capabilities | Research objective or early molecule | Research, development and manufacturing | Scientific capability, multi-year embedding |
| Private label | Scale and Operating capabilities | Brand, shelf and customer relationship | Manufacturing to agreed specification | Utilisation, category breadth, cost |
| Foundry | Manufacturing technology | Semiconductor design | Process technology and wafer manufacturing | Manufacturing technology itself |
The categories are not watertight. Real contracts usually overlaps, the terminology varies by industry, and large companies routinely operate across several models at once For example Dixon runs pure assembly alongside genuine ODM work; Syngene runs discovery research alongside commercial manufacturing.
Still, the taxonomy earns its place, because it exposes a progression in what the manufacturer is actually selling.
At one end, it sells factory capacity. Further along, it sells engineering capability. Further still, it may be selling process knowledge, scientific expertise or manufacturing technology.
The closer a business sits to those scarce capabilities, the more interesting its economics can become. Can. Not will.
Because there is another layer underneath, and it is the one most analysis skips. A manufacturer can contribute a great deal and still capture very little of what it creates.
The reason is buried in the contract. In Part 3 of the series we will who gets the economics of this business.
Leave a Reply