A semiconductor factory is only as capable as the machines inside it. One tool deposits a film, another removes material, another prints a pattern and another searches for defects that are too small to see. Each step must work with the others. A machine that improves one process but reduces total factory throughput can be a technical achievement and a commercial disappointment at the same time.
The investment question
Equipment suppliers benefit when chipmakers add capacity or require more sophisticated manufacturing. The investment question is how much a supplier earns from each technology transition, and whether that growth survives changes in customer spending. Semiconductor revenue, factory capital expenditure and equipment revenue are related, but they are not interchangeable measures.
Our central view is that process difficulty supports attractive niches, while timing and customer concentration preserve cyclicality. A supplier qualified for a critical step can enjoy durable relevance, service revenue and substantial switching barriers. It still depends on a small number of manufacturers deciding when to build, upgrade and operate their factories.
How the equipment stack works
Lithography transfers patterns onto light-sensitive material. Deposition adds thin films; etching removes selected material; cleaning prevents contamination; implantation changes electrical properties. Metrology measures structures, while inspection searches for defects. These functions repeat throughout fabrication. Packaging introduces additional requirements around bonding, interconnects, redistribution layers and testing.
The suppliers therefore have distinct exposures. ASML is central to advanced lithography. Applied Materials spans several materials-engineering processes. Lam Research is prominent in etch and deposition. KLA focuses on process control, including inspection and metrology. Tokyo Electron and other specialists address additional process steps. Describing all of them as identical picks-and-shovels investments misses where their products enter the manufacturing flow.
ASML’s EUV technology overview explains why optics, light sources and precision stages are inseparable parts of a lithography system. The manufacturing customer evaluates resolution alongside throughput, availability, process integration and cost. Better resolution is commercially useful only when it supports an economical production route.
Market structure and competitive advantage
| Equipment category | Customer objective | Source of staying power |
|---|---|---|
| Lithography | Print required patterns economically | Optics, precision and process ecosystem |
| Deposition and etch | Build and shape complex structures | Materials knowledge and qualified recipes |
| Inspection and metrology | Find problems and control variation | Detection capability and production data |
| Packaging and test | Deliver reliable integrated devices | Alignment, bonding and qualification |
Qualification is a major barrier. Changing a tool can require adjustments elsewhere in the process, and the customer risks lost production while validating the change. Suppliers learn from their installed bases and improve recipes, maintenance and software. That accumulated production experience can be more defensible than an isolated laboratory specification.
Nevertheless, incumbency is not permanent. A new transistor structure or memory architecture can change the required process and reopen competition. Customers may also support alternatives to reduce dependence on a single supplier. Investors should distinguish installed-base strength from the ability to win the next generation of critical steps.
Economics: capacity, intensity and service
Equipment demand has two broad drivers. Capacity investment adds machines to produce more output. Technology investment changes the machines and process steps required to produce a more advanced device. A supplier can outperform total wafer growth if its content per wafer rises, but it can underperform if a new process reduces the number of steps it supplies.
This is why smaller transistors do not automatically mean every tool category grows equally. A lithography improvement that reduces repeated patterning may eliminate some associated deposition and etch work, while new three-dimensional structures increase demand elsewhere. The relevant measure is net process opportunity after the manufacturing flow changes.
Service provides a second revenue stream through maintenance, replacement parts, upgrades and support. It tends to relate to the installed base and tool use rather than only to new factory construction. It can therefore soften a new-equipment downturn, although lower utilisation and restrictions on servicing particular customers can still affect revenue.
An illustrative comparison makes the economic test clear. A tool costing 15% more but delivering 25% more qualified output per year has 8% lower purchase cost per unit of annual output, before operating and financing costs. Customers ultimately care about this broader cost of ownership. Higher selling prices are defensible when the productivity or yield benefit outweighs them.
AI and hyperscalers: a delayed transmission
Hyperscaler spending reaches equipment suppliers through several decisions. A cloud operator orders systems; chip and memory suppliers translate those requirements into production plans; foundries and memory manufacturers then decide whether existing capacity is sufficient. Equipment orders can precede the final deployment by a considerable period. Their timing also reflects factory construction, installation and qualification.
AI affects both logic and memory. Advanced accelerators require demanding manufacturing processes, while high-bandwidth memory adds complexity in memory fabrication and packaging. Larger integrated systems also increase the financial cost of defects, supporting investment in process control. These are mechanisms for equipment demand, not a reason to apply a single AI growth rate to every supplier.
The indirect customer concentration matters. Several equipment customers can be expanding to serve the same hyperscaler programmes. Apparent diversification across foundries and memory suppliers may therefore conceal common exposure to a handful of downstream investment decisions. The strongest analysis follows demand through to the computing workload rather than stopping at a factory purchase order.
Current market debates — September 2026
Recent reporting supports a strong investment environment. Applied Materials’ fiscal third-quarter 2026 results and Lam Research’s June-quarter release describe strong financial performance, with Lam explicitly linking demand to AI. KLA’s fiscal 2026 fourth-quarter results provide a complementary view from process control. These results demonstrate supplier activity, not the eventual profitability of every customer expansion.
The near-term debate is whether unusually strong orders and margins reflect a higher sustainable spending level or some acceleration of future demand. The constructive case points to simultaneous logic, memory and packaging transitions. The countercase points to overlapping customer commitments, regional duplication and the risk of later digestion. Installation schedules, customer utilisation and repeat orders help distinguish the two.
High-numerical-aperture EUV is another important debate. Its higher resolution can simplify some patterning flows, but adoption depends on process economics and customer readiness. ASML’s 2025 strategy review discusses productivity and the potential move from multiple patterning to single patterning. A tool delivery, a customer qualification and broad production insertion are different milestones and should be reported separately.
Structural debates: concentration and substitution
The first structural question is whether rising complexity increases equipment spending faster than semiconductor output. Three-dimensional transistors, stacked memory and advanced packaging support that possibility, but suppliers must win the specific processes that gain importance. An expanding industry can still redistribute spending between vendors.
The second concerns market access and regional supply chains. Export restrictions and customer efforts to develop local alternatives can alter the addressable market, including services. Rules change, so any precise exposure estimate needs a dated assessment of current restrictions and the company’s disclosures. A broad claim that all China revenue is either protected or permanently lost is analytically inadequate.
Third, equipment vendors increasingly provide integrated process solutions and software. Better coordination can improve yields and deepen customer relationships, but manufacturers may resist excessive dependence on one supplier. The tension between integration and supplier diversity is part of the long-run competitive structure, particularly when a process transition changes who controls the most difficult step.
What to watch
Track orders, shipments and revenue separately; distinguish new systems from services; and compare supplier growth with customer capital expenditure and utilisation. Examine which process transitions increase the supplier’s content and which could remove steps. Monitor customer concentration and installation delays alongside headline backlog.
The strongest franchise is one that repeatedly earns a place in the customer’s next production flow and then supports that equipment economically for years. A large order book is valuable, but continued process relevance is what makes it durable.
Explore this sector
Semiconductors & chipmaking — sector overview
Related sectors: AI infrastructure & data centres · Data platforms & analytics
