Fab Equipment

Equipment is what every process, every material, and every design rule in this glossary ultimately runs on: the physical machines a fab buys, installs, and depreciates over a decade, each one capable of exactly one class of operation at a precision most other industries never need. A design can specify an atomically sharp interface or a line a few nanometers wide; equipment is what decides whether that specification is a real number or a wish.

One Wafer, Hundreds of Machine Visits a leading-edge logic wafer passes through 1,000+ process steps before it is a finished die Litho scanner Etch plasma chamber Deposition CVD / PVD / ALD Implant ion beam Metrology measure, decide loop: pass / rework / scrap A wafer revisits this sequence dozens of times A modern logic process repeats litho-etch-deposition cycles to build each successive layer — interconnect alone can mean 15+ metal layers, each one its own full pass through the sequence. Why one tool class can set the ceiling for the whole fab A fab's capacity is the minimum throughput across every tool class it depends on, not the average. An EUV scanner costs on the order of $150–400M and runs on the order of 100-150 wafers/hour; if a fab has too few of them, every other tool in the building can sit idle waiting for its turn. Equipment purchase decisions are capacity decisions years before the fab produces a single die.

A fab is, physically, a sequence of single-purpose machines a wafer visits over and over. Lithography scanners project a circuit pattern onto photoresist; etch tools remove material the pattern exposes; deposition tools — chemical vapor deposition, physical vapor deposition, atomic layer deposition — add material back in controlled thicknesses, down to single atomic layers when the application demands it; ion implanters drive dopant atoms into the silicon at precise energy and dose; metrology and inspection tools measure the result and decide whether the wafer proceeds, reworks, or scraps. None of these steps builds a transistor alone — a leading-edge logic process repeats this sequence over a thousand times, and a scrapped wafer at step 900 loses everything invested in the preceding 899.

Equipment cost and throughput set the economics of the entire node, not just one step. An EUV lithography scanner is the clearest example: a single tool can cost on the order of one hundred fifty to four hundred million dollars, ASML is effectively the sole supplier, and each unit processes on the order of a hundred to a hundred fifty wafers an hour. A fab's maximum output is bounded by whichever tool class it owns the fewest of relative to demand — buying equipment is a capacity decision made years before a single die ships, and it is why foundry capacity, not chip design talent, has repeatedly been the binding constraint on how much of a given node the whole industry can produce in a year.

Two Equipment Worlds: Capital Machines vs. Software Tools both are called "equipment" in the industry, and they obey opposite economics Fab equipment physical, one per unit of capacity • Scanners, etchers, deposition chambers • Cost: tens to hundreds of $M per tool • Depreciated over 5-10+ years • Adding capacity = buying more tools • Supply is the hard constraint (ASML, etc.) governs wafers per hour, the physical floor EDA / design equipment software, copies scale almost for free • Synthesis, place-and-route, signoff tools • Cost: license seats, not fabrication capex • A new seat ships instantly, not in years • Adding capacity = buying more licenses • Supply is practically unlimited governs designs per year, the intellectual floor Same word, opposite bottleneck A wait for a lithography scanner is measured in years of order backlog; a wait for an EDA license is measured in a purchase order.

The word "equipment" splits into two economies that happen to share a name. Fab equipment is physical capital: a finite number of scanners, etchers, and deposition chambers that can only be built and delivered on a manufacturer's own production schedule, so capacity additions take years and a shortage ripples through every customer at once. EDA equipment — the software tools that turn a design into a manufacturable file, covered in full under the `design` entry — is licensed, not fabricated; a new engineering seat is a purchase order, not a years-long supply chain. A fab can be capacity-constrained by its scanner count while its design partners are constrained by nothing more than how many EDA licenses they choose to buy.

Metrology and inspection equipment is the category that decides whether every other tool gets to keep running. Measuring critical dimension, overlay, and film thickness after each step is how a fab catches a drifting process before it scraps a full lot rather than after; inspection tools hunting for particle defects do the same for yield. These tools do not add material or pattern anything — they are pure decision-makers sitting between every process step, and a fab under-invested in metrology finds out about a problem only once it shows up as lost yield three hundred wafers later.

CategoryExamplesWhat it adds or changesBinding constraint
LithographyDUV, EUV scannersPattern, nothing elseTool count, ASML allocation
EtchPlasma, RIE chambersRemoves material per patternChamber-hours, chemistry supply
DepositionCVD, PVD, ALD systemsAdds material, atomic-layer controlPrecursor purity, chamber-hours
ImplantIon beam, plasma implantersDopant placement and doseBeam current, tool count
Metrology/inspectionCD-SEM, overlay, defect scannersNothing physical — pure measurementThroughput vs. process drift speed
Design (EDA)Synthesis, P&R, signoff toolsTurns intent into a mask-ready fileLicense seats, not fabrication capacity

Equipment is where every other glossary entry in this series becomes real or stays theoretical. A material exists because some deposition or implant tool can place it; a process exists because some chamber can run its recipe; a design exists because some signoff tool can verify it. None of those entries describe a bottleneck on their own — the equipment behind each one does, and it does so on its own schedule, at its own price, measured in tool-years rather than good ideas.

Take fab equipment further

Ask the copilot about this term, or have our engineers assess it against your process.