Who Actually Makes a Semiconductor? Mapping the Companies Behind Every Step from Design to Board Assembly
イチケン / ICHIKENElectronics creator Ichiken built this video from study notes they compiled while learning about the semiconductor industry. They found that many books and materials explain how chips are manufactured, but few let you see which companies are involved at each step. The video follows a single chip from design through materials, masks, wafers, front-end and back-end processing, testing, and finally mounting on a circuit board, and names the equipment makers, component suppliers, and materials companies active at each stage.
The focus is on cutting-edge logic semiconductors: the chips that do the computation in smartphones, PCs, and AI servers. Memory and power semiconductors come up along the way, but the processes and companies were chosen around the workflow for advanced logic. Ichiken's stated goal is that when a semiconductor company appears in the news, viewers can tell where in the process it sits.
The Overall Flow and Where Companies Fit
Ichiken starts with a process map. Three things must be prepared before manufacturing can begin: the circuit design, the photomask that carries the pattern from that design, and the wafer that forms the base of the chip. The central section is where chips are actually built. Circuits are formed on the wafer in the front-end process. While the chips are still on the wafer, they are tested and the good ones are identified. The wafer is then cut apart in the back-end process, the chips are packaged, and a final test is run before shipment. So there are two tests: one before dicing and one after the chip is finished as an electronic component.
Three types of chip-making companies cover this section: foundries, which manufacture other companies' designs; IDMs, which do everything from design to manufacturing in-house; and OSATs, which handle assembly and testing. Finished chips are then mounted on products during PCB assembly. Underneath all of this sit companies that people tend to overlook: suppliers of materials, of components that go inside the manufacturing equipment, and of factory infrastructure.
Ichiken's overview slide arranges company types from left to right as design, wafers, masks, front-end, back-end, testing, and board assembly, with Japanese companies highlighted in red. Laid out this way, Japanese firms cluster in front-end and back-end manufacturing equipment, in the components used inside that equipment, and especially in materials. The bottom rows show the chipmakers themselves (foundries, memory makers, IDMs) and the infrastructure and component suppliers that support the whole process.
For deeper reading, Ichiken recommends two resources: slides on the semiconductor industry compiled by "🐇の半導体勉強部屋" (🐇の半導体勉強部屋), which list companies by sector, from EDA and IP through wafer equipment, materials, gases, chemicals, and trading companies; and a process explanation on semiconductor-industry.com by a working process engineer known as ずーぼ, which walks through the front-end and back-end step by step.
Two Business Models: Horizontal Division of Labor vs. Vertical Integration
There are two broad ways to make semiconductors. The first is a horizontal division of labor, in which design, manufacturing, and assembly/test are handled by separate companies.
- Fabless companies only design. Examples are NVIDIA and Apple, and recently Google, Amazon, and Microsoft, companies whose core business isn't semiconductors but which design chips for their own use.
- Foundries handle manufacturing, specifically the front-end process. TSMC is the best known. Rapidus, which aims to mass-produce cutting-edge logic in Japan, is also a foundry, as is Samsung. Ichiken notes that while foundries aren't household names like their fabless customers, most people have heard of them in the news.
- OSATs handle the back-end and testing.
The second model is vertical integration: the IDM. Ichiken lists Intel, Texas Instruments, and Renesas; power semiconductor makers Infineon, ROHM, Mitsubishi Electric, and Fuji Electric; and Sony for image sensors. Memory makers such as Samsung, SK Hynix, Micron, and Kioxia also design in-house and manufacture in their own fabs, so Ichiken treats them as close to IDMs. Samsung and Intel are IDMs that also run foundry businesses for other companies.
Ichiken stresses that the two-way split in the diagram is a simplification. Many companies sit in between. A fab-lite company manufactures in its own fabs but outsources only its most advanced products to companies like TSMC. Infineon outsources some microcontroller production to foundries, and Sony outsources part of its image sensor production. Most companies decide product by product how much to make in-house, so the slide simply groups companies by which side they lean toward.
Key Players: Fabless, Foundries, IDMs, and Memory
Among fabless companies, Ichiken observes that each has its specialty: NVIDIA in GPUs, AMD in CPUs, Qualcomm in smartphone SoCs (Snapdragon), and Broadcom in custom designs such as Google's TPU, an AI-specific computing chip.
For foundries, TSMC is the name heard most often, and for cutting-edge logic there are products that seem only TSMC can make. Others include Samsung, Intel Foundry, SMIC, and Rapidus. Ichiken says they have high hopes for Rapidus, which is working hard to manufacture 2nm-equivalent chips in Hokkaido.
Among IDMs, Intel makes PC and server CPUs in its own fabs and has recently moved into contract manufacturing. Sony leads the world in camera image sensors. Renesas is a major maker of automotive microcontrollers. Infineon is a leader in power semiconductors; besides standard silicon devices, it makes power semiconductors using silicon carbide (SiC) and gallium nitride (GaN), which tolerate high voltages and fast switching. Texas Instruments leads in analog ICs for power, audio, and sensor signals. Mitsubishi Electric, Fuji Electric, and ROHM are Japanese power semiconductor makers, each with its own strengths. Some companies on the list have fab-lite elements, but Ichiken grouped them as IDMs for this video.
Memory makers are a prime example of IDMs. Ichiken explains the two main types. DRAM temporarily holds data being processed and loses it when power is cut. HBM, the high-bandwidth memory that has been in the spotlight for AI, is a type of DRAM that stacks multiple DRAM dies so large amounts of data can be exchanged at once. NAND stores photos and large data and retains it without power. The biggest players are concentrated in South Korea, Samsung and SK Hynix, which Ichiken thinks look set to be among the winners of the AI era. Micron is another well-known name, and Kioxia in Japan makes NAND. Both DRAM and NAND are increasingly built by stacking layers to raise speed and capacity. Chinese companies such as CXMT and YMTC have also been gaining ground.
Design: EDA, IP, and the PDK
Ichiken first clarifies two terms. EDA is design software, and IP refers to pre-designed circuit blocks. A modern chip contains tens of billions of transistors and can't be designed by hand, so designers use EDA tools. Circuits with well-defined functions, such as CPU cores or USB interfaces, are bought as ready-made IP and combined on the chip. Fabless companies like NVIDIA and Apple and IDMs like Intel and Renesas both use EDA and IP. The difference lies in where the chip is manufactured.
That difference shows up in the PDK (Process Design Kit), which Ichiken describes as a collection of design rules for a particular fab. Because the shapes and sizes that can be manufactured differ by fab, once you choose a fab you design to its rules. A fabless company receives a PDK from a foundry such as TSMC, designs to it, and outsources manufacturing. An IDM uses its own PDK and builds the chip itself.
In EDA, the three leaders are Synopsys, Cadence, and Siemens. Synopsys offers a full suite of core design tools and is also a major supplier of interface IP such as USB and PCIe. It has been acquiring Ansys, a physics simulation company. Ichiken's interpretation of the motive: when chips are stacked, as in AI chips, heat builds up, boards warp, and signals have a harder time getting through. Signals at very high frequencies also interfere with each other, and all of these effects influence one another, so circuit design now has to be combined with physical analysis. Cadence is strong in analog and package design and can handle the package substrate together with the chip. Siemens' Calibre, a final pre-production verification tool, is widely used as an industry standard. Other names include Keysight in the US, which is strong in high-frequency design, and companies doing component simulation. For PCB rather than chip design, there are Zuken and Altium, which is owned by Renesas. China is also developing its own EDA tools.
As for IP, designing blocks like CPUs and USB from scratch for every chip would take too much time and manpower and add risk, so companies buy proven blocks and combine them with their own circuits. IP isn't cheap: Ichiken says there is generally an upfront license fee plus a royalty on every chip sold. ARM is the leading IP vendor. It doesn't make chips; it licenses CPU designs and earns from royalties, a business model Ichiken finds interesting. They add that by 2026 ARM appears to be making its own chips for data centers for the first time. Synopsys and Cadence are also major IP vendors. RISC-V, a CPU instruction set anyone can use for free, has been spreading, with companies such as SiFive in the US and Andes in Taiwan active in that area. There are also IP firms specializing in GPUs, memory interfaces, and on-chip interconnect.
Photomasks: Expensive Masters for Every Layer
Once the design is done, manufacturing needs two things: a photomask, a glass master plate carrying the circuit pattern, and a silicon wafer to build on.
At the end of design, the layout data is sent to the mask shop, a handoff the industry calls "tape-out." The mask shop transfers the pattern onto glass to make the photomask. During the exposure step of the front-end process, light passes through the mask onto the wafer to print the circuit, and exposure is repeated many times. Because each layer has a different layout, each layer needs its own mask. According to Ichiken, a state-of-the-art product needs roughly 70 to over 100 masks, and a full mask set for single-digit-nanometer chips costs $10–20 million or more, or over a billion to several billion yen. Because the mask is the master, a single scratch is reproduced on every chip exposed through it. A design error means remaking this costly set, which is why a failed design is so expensive.
Ichiken notes that while mask makers make masks, cutting-edge masks are usually produced in-house by chipmakers such as TSMC, Samsung, and Intel. Merchant mask makers like DNP in Japan sell to outside customers.
The process starts with special glass that doesn't expand or contract. A light-blocking layer is added to make "mask blanks," supplied by companies such as HOYA and AGC. An electron beam writes the circuit onto the blank, and there are equipment makers for e-beam writers. Development, etching, and cleaning steps are not shown on the slide. Inspection follows. For EUV masks, inspection uses an EUV light source, and some companies make those sources in-house; Ushio and Gigaphoton (part of the Komatsu group) also produce EUV sources for mask inspection. EUV is the light used to expose advanced logic, and its short wavelength allows extremely fine circuits. Finally, a very thin film called a pellicle is mounted as a dust cover. Because there is a gap between the pellicle and the mask, dust that lands on the pellicle has little effect on exposure.
Silicon Wafers: From Quartz Rock to 11N Purity
The wafer starts as ultra-high-purity polysilicon, made by companies such as Wacker, Hemlock, and Tokuyama. Polysilicon is melted and grown into a large single crystal, sliced thin, and polished to a mirror finish. Shin-Etsu Chemical is the leading wafer maker, strong in low crystal defect density and high flatness; Ichiken describes it as almost a general materials company, since it also supplies photoresists and mask materials. SUMCO specializes in wafers and focuses on cutting-edge 300mm products. GlobalWafers in Taiwan is another major supplier.
What makes silicon wafers remarkable, Ichiken says, is purity: about eleven nines, known as "11N." Ichiken also gives rough chips-per-wafer figures: around 500 for a smartphone SoC and around 60–80 for a high-end GPU. Defects reduce the usable count further. That is why larger chips cost more, and why chiplets, which build small dies and combine them to act as one chip, have been gaining attention.
The supply chain begins earlier. Quartz or silica rock, made of silicon and oxygen, is heated with carbon to remove the oxygen, yielding metallurgical silicon of about 99% purity, which is far from enough for semiconductors. A further, somewhat tricky process turns it into polysilicon of about 11N. The wafer maker's work then starts: polysilicon is melted in a quartz crucible, a seed crystal is dipped in and slowly pulled up while rotating (the Czochralski method), and a large single-crystal ingot grows. The ingot is trimmed, sliced with thin wire, chamfered to keep the edges from chipping, ground on both sides to even out thickness, chemically etched to remove grinding damage, and polished to a mirror. It is then cleaned, inspected for particles, scratches, and flatness, and packed in special shipping containers. Ichiken stresses that wafer quality depends not only on ingot growth but on the precision of the cutting, grinding, and polishing equipment and materials, and shows a slide of the many companies involved. They find it interesting that even the shipping containers have specialist makers.
In a short digression on non-silicon wafers, Ichiken mentions SiC for EV motor-drive power converters. US company Wolfspeed handles everything from substrates to devices, though its finances in 2025 apparently weren't good. Other examples are InP substrates for optical communications, GaN substrates for high-frequency and optical devices, and GaAs substrates.
The Front-End Process: Building Layers in a Loop
The front-end builds the circuit as many layers on the wafer. It has the most steps and the largest number of equipment, component, and materials companies.
Ichiken shows a chip cross-section. At the bottom is the silicon wafer, on whose surface transistors are built: billions to tens of billions on the latest chips. Making the transistors is the first half, called FEOL (front-end of line), whose main steps are thermal oxidation, lithography, etching, ion implantation, and heat treatment. Stacked above is the copper wiring that connects the transistors: about a dozen layers, getting thicker toward the top. This is BEOL (back-end of line), using deposition, lithography, etching, plating, and planarization. Cleaning and inspection are inserted throughout.
One layer is formed as follows. Clean the wafer surface. Deposit a film. In lithography, coat resist, expose the pattern, and develop it. Then either etch to remove film, which creates shapes, or implant ions to introduce impurities, which changes silicon's properties; the choice depends on the layer. Strip the resist, apply heat treatment (for example after implantation) to restore the atomic arrangement, polish uneven surfaces flat, and inspect for defects and misalignment. Then start again with cleaning for the next layer.
For each step, Ichiken looks at three kinds of companies: equipment makers, whose machines go into foundry and IDM fabs; component makers, which supply lenses, power supplies, valves, wafer stages, and so on to equipment makers; and materials makers, which sell chemicals, gases, photoresists, polishing slurries, and other consumables to fabs.
Cleaning, Deposition, and Lithography
Cleaning removes contaminants using chemical solutions and ultrapure water. SCREEN is the world leader in equipment, with Tokyo Electron, Lam Research, and SEMES also active. Because the tools contain liquids, they need seal materials that withstand chemicals, and they contain wafer-transport robots, which are found in most front-end tools. Chemicals such as hydrogen peroxide and hydrofluoric acid (which dissolves oxide films) are an area where Japan has been strong, and there are makers of ultrapure water systems for rinsing. Since keeping the wafer as clean as possible is essential, Ichiken calls cleaning a very important step.
Deposition lays down films in nanometer increments using CVD and ALD (chemical reactions of gases), sputtering (vaporizing metal and depositing it), and thermal oxidation. Applied Materials, a US company, has the widest lineup, with ASM in the Netherlands and Japanese companies also active. Components include ceramic heaters that warm the wafer, mass flow meters for process gases, and plasma power supplies. Materials, such as sputtering targets for metal films and precursor gases, are also important.
Lithography begins with coating and developing: a thin layer of photoresist is applied, and after exposure, unwanted resist is dissolved with developer to reveal the pattern. Tokyo Electron leads in "coater/developer" equipment, which is used linked to the exposure tool. Materials loom large here, and Japanese companies are particularly strong in photoresists: Tokyo Ohka Kogyo and JSR develop EUV resists, and Shin-Etsu Chemical supplies resists for argon fluoride (ArF) laser exposure, the previous-generation technology. Components include filters that remove tiny particles from the resist and pumps that dispense precise volumes.
Exposure, which Ichiken calls a core step, uses a lens to shrink the mask pattern onto the resist. In state-of-the-art EUV, glass absorbs the light, so mirrors are used instead of lenses. Shorter wavelengths allow finer patterns, and finer patterns mean more transistors per chip. ASML in the Netherlands is the only company that can make EUV lithography systems. Nikon and Canon also make exposure tools, but they cannot produce patterns as fine as ASML's EUV.
The EUV light source is made by Cymer, an ASML subsidiary. Ichiken highlights the "droplet generator," which reportedly fires tin droplets a few tens of micrometers across at tens of thousands per second. A CO₂ gas laser with a wavelength of 10.6 micrometers hits the tin. That wavelength is far too long to pattern fine circuits directly. Instead, the tin turns into plasma and emits EUV light at 13.5nm, which is reflected off the photomask onto the wafer's resist through a highly complex optical system. Previous-generation DUV lithography is still widely used, and for its ArF and KrF laser sources, Japan's Gigaphoton is a manufacturer on par with Cymer. The lenses and mirrors in ASML's EUV tools are made by Zeiss in Germany, and the mirror bases use special low-thermal-expansion glass such as Corning's. Materials include the mask blanks and pellicles mentioned earlier and gases for the light source.
Etching, Ion Implantation, Heat Treatment, Planarization, and Inspection
Etching uses plasma gas to remove film in areas the resist doesn't protect. Lam Research leads, with Tokyo Electron, Hitachi High-Tech, and Applied Materials also present. Ichiken explains the electrostatic chuck: because the chamber runs at low pressure, air suction can't hold the wafer, and mechanical clamps apply force to the wafer, which isn't ideal, so static electricity is used instead. Other components include RF power supplies for plasma, vacuum valves, and electrode rings. Materials include fluorine- and chlorine-based gases.
Ion implantation ionizes impurities such as boron and phosphorus and drives them into the wafer to give transistors their switching properties. Equipment makers include Applied Materials, Axcelis, and Sumitomo Heavy Industries; Ichiken remarks on how often Applied Materials comes up. A high vacuum is needed to send ions in a straight line, so vacuum pumps are essential. Dopants are supplied as gases, some highly toxic, so storage containers matter.
Heat treatment restores the atomic arrangement disrupted by implantation so the implanted impurities can function in the transistor. Methods range from slow heating in a furnace to heating for just milliseconds.
Planarization (CMP) polishes the wafer flat using both chemical and mechanical action with a slurry and a pad. Applied Materials is strong in equipment, and Ebara is also a player. This step consumes a lot of materials: slurries from companies such as Fujifilm, Entegris, and Resonac, plus polishing pads and pad conditioners. Ichiken admits they didn't know conditioners existed until they looked it up.
Inspection and metrology checks for particles and defects after each step and measures line widths and overlay misalignment. KLA in the US and Hitachi High-Tech in Japan make CD-SEMs, electron microscopes that measure circuit line widths, and ASML also makes overlay measurement tools. Measuring at nanometer scale requires moving the wafer precisely, so THK's linear guides and position sensors matter. Ichiken notes that because this step only measures, they couldn't find any consumable materials for it.
Cleaning, deposition, lithography, etching, implantation, heat treatment, planarization, and inspection make one loop and one layer. Because a chip has dozens of layers, the loop repeats many times. For state-of-the-art chips, lithography alone is said to be repeated from several dozen to around 100 times, which is why a wafer takes several months to complete the front-end. At the end comes the WAT (Wafer Acceptance Test): probes contact test patterns built onto the wafer and current is applied to confirm the transistors were made as intended. Interim inspections check physical form; WAT is the final electrical check of the front-end.
Shared Components, Factory Infrastructure, and the "Big Five"
Some components appear across many processes. In tools that run at atmospheric pressure (cleaning, coat/develop, CMP) there are transport robots, seal materials, chillers for temperature control, gas flow meters, high-purity valves and fittings that switch gas without contaminating it, and gas panels that bundle the piping. In vacuum or plasma tools (deposition, etching, implantation, EUV lithography, e-beam inspection) there are electrostatic chucks, an area where Japanese ceramics makers are pushing hard; vacuum pumps, vacuum valves separating atmosphere from vacuum, and vacuum gauges; abatement systems that neutralize process gases; RF power supplies; ceramic chamber parts such as walls and showerheads; cleaning and refurbishment services for dirty parts; and quartz parts for heat treatment and deposition.
Factory infrastructure matters too. Wafer containers are carried automatically on ceiling rails, an area where Murata Machinery and Daifuku are strong. The containers are heavy and the volume is huge, so automation is essential. Fabs also need ultrapure water, sealed wafer carriers, and cleanrooms. Ichiken's personal view is that component suppliers are in a good position: whichever equipment maker's tools sell well, they get work.
Summarizing front-end equipment, Ichiken identifies about five major makers. ASML is the only EUV lithography maker. Applied Materials covers deposition, polishing, and implantation and is, in Ichiken's view, probably the broadest lineup. Lam Research leads in etching and excels at digging deep holes for 3D NAND. Tokyo Electron makes the coater/developers used alongside the leading EUV systems. KLA leads in defect inspection. Ichiken adds SCREEN in cleaning, and observes that even the majors have neatly separated areas of expertise.
The Back-End Process: What Gets Built and How It Flows
The back-end cuts the wafer into chips, mounts and connects them, and seals them in resin. Ichiken says people tend to see these as mundane steps, but AI chips have changed them significantly.
In a package cross-section, the chip from the front-end sits in the middle, connected to the package substrate by thin metal wires or bumps (small solder balls). The package substrate fans out the chip's fine terminals to larger terminals for mounting on the product's PCB. Resin encases everything against moisture and impact, and solder balls on the bottom attach it to the board, as on smartphone and PC boards.
Most cutting-edge chips, including smartphone SoCs, PC CPUs, and AI GPUs, use flip-chip packaging instead of wire bonding. The chip is flipped circuit-side down and connected via bumps. Wires can only come from the chip's edges, but bumps can be placed by the thousands across the whole surface, greatly increasing terminal count. Connections are also shorter, which helps high-speed signals. The gap between chip and substrate is filled with underfill resin to protect the bumps from cracking.
Unlike the front-end, the back-end is essentially one-way with no repeated loop, though electrical testing comes first. After WAT, wafer testing applies fine needles to each chip still on the wafer to check whether it works; defective chips are excluded from assembly. Then come back-grinding, dicing (the point where handling shifts from whole wafers to individual chips), die bonding onto a substrate or lead frame, electrical connection by wire bonding or flip-chip, molding in resin, finishing such as marking, and the final test. Ichiken notes that for cutting-edge chips this standard flow is starting to change, which they return to later.
Wafer Testing: Testers, Probe Cards, and Probers
Wafer testing exists because assembling defective chips wastes effort and materials downstream. A probe card, a bundle of fine needles, touches each chip's terminals, and a tester sends signals and reads the responses. The tester and wafer are connected through the probe card. Since a wafer holds hundreds of chips, a prober moves the wafer precisely to position each one. Well-known tester makers are Advantest and Teradyne; prober makers include Tokyo Electron and Tokyo Seimitsu. The two machines are used together as a unit, and the tester is used again in the final test.
Ichiken describes the tester as essentially a collection of measuring instruments: power supplies that apply voltage and measure current, circuits that send signals to each terminal and evaluate the response, and circuits that generate timing at picosecond resolution, replicated for hundreds to thousands of terminals and operated simultaneously. Because testing takes time, the key is how fast it can be finished. Components include probe cards with hundreds of pins and measurement circuits.
Back-End Steps: Grinding, Dicing, Bonding, Connecting, Molding
Back-grinding thins the wafer, which was kept thick through the front-end to prevent breakage but won't fit thin packages as is, especially for stacked chips such as HBM. Protective tape covers the circuit side, and a rotating grinding wheel thins the back. Disco leads the world in equipment and also makes its own grinding wheels. Ichiken explains three roles of back-grind tape. It protects the circuit from grinding debris and the cooling water. It absorbs surface unevenness and holds the wafer flat; since any variation in tape thickness becomes variation in wafer thickness, the tape needs high precision. And it must peel off cleanly: because thinned wafers break easily, a tape whose adhesive weakens under UV light is used so it comes off without residue or force. Lintec and Mitsui Chemicals make such tapes.
Dicing cuts hundreds to thousands of chips apart while still on tape, commonly with a high-speed diamond blade and water spray. Disco and Tokyo Seimitsu are said to account for most blade dicing equipment. Dicing tape, like back-grind tape, comes from Japanese companies such as Lintec.
Die bonding picks chips one by one off the tape and fixes them to a package substrate or lead frame, leaving chips marked defective in wafer testing behind. Besi in the Netherlands and Singaporean companies are strong in equipment. For lead frames and adhesive films, Ichiken again remarks on Japanese strength in materials. High-heat chips such as power semiconductors may use solder or silver sintering for better heat dissipation instead of adhesive.
Electrical connection uses either wire bonding, attaching gold or copper wires thinner than a hair one at a time, or flip-chip, which connects bumps directly and suits CPUs and GPUs because the short wiring carries signals quickly. Equipment includes wire bonders and flip-chip bonders. Tanaka is a bonding wire supplier, and underfill for the gaps between bumps comes from companies such as Namics and Resonac.
Molding encases the chip in resin because exposed chips and wires are vulnerable to moisture, dust, impact, and light. The substrate is clamped in a mold and resin is injected and cured, using methods such as compression molding and transfer molding. TOWA has a large share and is strong in compression molding, where resin is pressed down from above, a method considered suited to encapsulating thin stacked chips such as HBM. The encapsulant is epoxy resin with fillers; Sumitomo Bakelite is considered the world leader, followed by Resonac and Shin-Etsu Chemical. Ichiken again notes the strength of Japanese materials makers.
Finishing includes laser marking and terminal preparation; BGA packages, with solder balls on the back, allow very dense layouts. In the final test, packaged chips are placed in sockets and driven by the tester. Test sockets are a component category, and some companies take on testing as contract work. Companies that handle the whole back-end as a package are OSATs, and ASE is the industry leader. With AI chips integrating multiple dies, though, advanced packaging has become so important that front-end companies such as TSMC and Intel have started doing packaging themselves.
Why the Back-End Is Now in the Spotlight
Ichiken explains that chip performance has traditionally improved in two ways: miniaturization, which packs more transistors into the same area, and making chips larger. Both have hit walls. Each generation of miniaturization requires larger equipment and processes, and costs are rising rapidly. Larger chips are more likely to be ruined by a single defect, so yield falls.
Hence chiplets: build small dies by function, select only the good ones, and reconnect them at high density side by side or stacked inside the package. This makes the chip effectively larger. In AI GPUs, HBM is placed right next to the GPU so large amounts of data move over short distances.
Ichiken then walks through the package substrate. The center core is glass cloth impregnated with resin, which resists heat-induced warping; Nittobo's T-glass is commonly used. Above and below are build-up layers of alternating insulating film and copper wiring; the insulating film is Ajinomoto's ABF, and the wiring uses Mitsui Mining & Smelting's ultra-thin copper foil. Spacing widens layer by layer to connect the chip's fine terminals to the large solder balls below. The outermost green layer is solder resist, protecting the wiring. Ichiken says that the larger the AI chip, the more substrate materials and build-up quality matter. Ibiden is one manufacturer of these substrates. Ichiken acknowledges this section was rushed and incomplete and plans a separate, thorough video on packaging.
On how chips are connected inside the package, Ichiken says AI chip performance depends not just on a single chip but on how closely multiple chips can be connected and how many wiring paths can be added. Flip-chip is the standard for high-performance chips. Many AI GPUs go further: GPU and HBM sit side by side on a silicon interposer. This is needed because the green organic substrate can't achieve silicon's precision in wiring fineness and via spacing, which makes connecting GPU and HBM directly on it difficult, especially given the large number of connections. Precision-machined silicon links GPU and memory, and the interposer then connects to the organic substrate and ultimately to the PCB. Because the back-end now helps determine the chip's performance itself, Ichiken concludes, more sophisticated technologies closer to front-end processing are moving into it.
PCB Assembly and Why MLCCs Matter for AI
Finished chips only fulfill their function once mounted with other components on product boards. Mounting machines (placement machines) are an area where Japanese makers are strong: FUJI, Yamaha Motor, Panasonic Connect, and JUKI. Japanese companies are also competitive in electronic components around the chip.
Ichiken singles out MLCCs (multilayer ceramic capacitors), thousands of which sit on AI GPU boards. When an AI GPU starts computing, billions of transistors switch at once, and currents of several hundred amperes flow over nanosecond timescales, causing voltage drops in the wiring to the chip. Meanwhile the chip runs at around 1V and tolerates fluctuations of only a few tens of millivolts; beyond that it won't work properly. The power supply responds only in milliseconds, so MLCCs are placed near the GPU to deliver charge instantly, suppressing voltage fluctuation and keeping power supply impedance low. Ichiken considers this important enough to cover in another video.
Conclusion
Ichiken's takeaway is that making a single chip is divided among many countries and companies, each handling its own part of the process, and that changing even one step can affect the steps that follow. They present the slides as a summary of what they learned through their research and invite viewers to add further information in the comments for others' benefit. Package substrates and MLCCs remain open topics they intend to cover in future videos.
This video is brought to you by DigiKey. Hello. This video is an explanatory guide designed to help you get a broad overview of the semiconductor industry. I originally created this document while studying the semiconductor industry. This is a version of the materials I had been compiling that I’ve reworked for use in a video.
What I was feeling while studying was that there are quite a few materials and books that explain the semiconductor manufacturing process, but there weren't that many materials available to browse through showing what kinds of companies are involved in each stage of the process. So, this time, we’ll be covering semiconductor design, materials, and manufacturing equipment, front-end, back-end, and testing. Following the process involved in making a single chip, I have organized and summarized what kinds of players are involved in each stage and the various devices and materials involved.
Although we often use the term “semiconductor” as a general term, there are actually many different types. This time, the focus is more on cutting-edge logic semiconductors. In other words, we're focusing our discussion on the chips that handle calculations in smartphones, computers, and AI servers. Memory and power semiconductors will also be mentioned to some extent, but the processes and companies featured here are selected based on the workflow for developing cutting-edge logic.
I prepared these materials with the goal of making it clear which stage of the process a company is involved in when I see the names of semiconductor companies in the news or elsewhere.
Before we begin the commentary, let me introduce today’s sponsor. Once again this year, the event is being hosted by CQ Publishing, the publisher of Transistor Technology. The Super Electronics Expo 2026 will be held in Akihabara. It was a huge success last year, and this year's event will be held on an even larger scale. Of course, Ichiken will also be there at the DigiKey booth, and I will be giving presentations on electric karts racing at Tsukuba Circuit and domestically manufactured chargers.
Also, at the Super Electronics Expo, we'll be holding hands-on workshops based on a popular series on the DigiKey channel, a series on building your own CPU for high school students. In response to requests from people who want to try making it themselves and actually learn, we'll explain CPU design and manufacturing in an easy-to-understand way while demonstrating operation on an FPGA. You can also ask the instructor questions in real time, which you couldn't do in the video, so if there's anything you don't understand, please feel free to ask.
After registering to attend the Super Electronics Expo on November 3, please be sure to register for the DigiKey workshop as well. There is a link in the description.
First, let's take a look at the process of manufacturing semiconductors. The three items on the left are the ones you need to gather before you start making this: circuit design, the photomask used to create the circuit pattern based on that design data, and the wafer that serves as the base for the chip.
This central section is where the chips are actually manufactured. After building the circuit in the front-end process, while the chips are still on the wafer, select the good ones during testing. Cut them into pieces during the back-end process and wrap them in packaging. We'll test them one last time before shipping them. There's a part here that says “before cutting,” and that refers to dicing. Test them before dicing, and then once they're ready for use as electronic components—that is, after they're completed. We'll run two tests.
The companies that manufacture the chips are responsible for this entire process: a foundry that manufactures products based on designs from other companies, an IDM that handles everything in-house, from design to manufacturing, and an OSAT, which handles assembly and inspection. We'll take a closer look at each of these later. The chips produced here are mounted onto the product during the PCB assembly process.
Furthermore, while people tend to focus on manufacturing equipment below that, there are also companies that specialize in materials, components inside equipment, and factory infrastructure, among others. All of these elements support semiconductor manufacturing.
Next is this one. Including the companies I'm introducing today, I've put together a list here of the general types of companies that exist. In order from left to right: design, wafers, masks, front-end processes, back-end processes, testing, and board assembly. The red sections represent Japanese companies. When you look at them side by side like this, you can see that Japanese companies are common especially in manufacturing equipment for the front-end and back-end processes, the parts used in it, and the materials section.
Now for the bottom row. These are companies that manufacture chips, handling everything from front-end processes to testing. In other words, foundries, memory manufacturers, and IDMs. At the very bottom are companies that specialize in the equipment, components, and factory infrastructure that support the entire process from below.
Next, I'll break down the overall flow a little more, but first, take a look at this. For those who want to know a little more, here are two helpful resources. I'm not sure what the author's name is, though. These are slides on the semiconductor industry compiled by “🐇の半導体勉強部屋.” From companies that do design and manufacturing to EDA design software, IP for circuit components, wafer manufacturing equipment, materials, gases, chemicals, and trading companies, companies in the industry are listed by sector. It's a very clear way to get a grasp of the big picture of the industry.
The second is a page from semiconductor-industry.com explaining the manufacturing process involved in making semiconductors. A current process engineer named ずーぼ explains the entire front-end and back-end process workflow step by step, giving a better overview. Since both are extremely useful resources, if you're interested, please check them out via the links in the description.
From here on, let's first take a look at who is doing what within the industry. There are two main methods for manufacturing semiconductors. One is a division-of-labor model where design, manufacturing, assembly, and inspection are each assigned separately. A fabless company is one that focuses solely on design. NVIDIA and Apple—and recently Google, Amazon, and Microsoft—even companies whose core business isn't semiconductors design chips for their own use.
Next, foundries are the companies that handle the actual manufacturing of semiconductors. Specifically, this refers to what is known as the front-end process in semiconductor manufacturing. I'll explain the front-end steps later. The most famous one is TSMC. In addition, Rapidus, which aims to mass-produce cutting-edge logic in Japan, is also a foundry. Then there's Samsung, among others. Compared with fabless companies like NVIDIA and Apple, foundries aren't exactly household names, although you've probably heard about them at least once on the news.
Here are the back-end processes and testing sections. Companies responsible for assembly and inspection are called OSATs. The model described above is the horizontal division of labor; the other business model is the vertically integrated model. In addition to Intel, Texas Instruments, and Renesas, power semiconductor manufacturers Infineon, ROHM, Mitsubishi Electric, and Fuji Electric, and companies such as Sony, which manufactures image sensors, fall into this category.
Then there are memory manufacturers like Samsung, SK Hynix, Micron, and Kioxia as well. Since they design the products in-house and manufacture them at their own factories, I think this is similar to an IDM. Also, while Samsung and Intel are IDMs that manufacture their own products, they also operate foundry businesses that manufacture products for other companies.
In this diagram, it's divided into two parts, but in reality, there are many companies that fall somewhere between these two extremes. An IDM manufactures products at its own factories, but it may outsource the manufacturing of only its cutting-edge products to companies like TSMC. This type of structure is called “fab-lite.” Actually, the companies listed here aren't exactly split into two clear groups either. Infineon also outsources some microcontrollers and other products to foundries, and Sony also outsources the production of some parts of its image sensors. Most companies decide for themselves how much of the product they will manufacture in-house and where to delegate responsibility based on each product. This slide broadly categorizes them based on which side they lean toward.
Among the division of roles we just looked at, first, let's take a look at some of the leading fabless companies that design chips. A fabless company does not own its own factories. It sells the chips it designs under its own brand and outsources manufacturing to foundries such as TSMC. There are all kinds of fabless companies, but it seems everyone has their own areas of expertise. If it's NVIDIA, it's GPUs. If it's AMD, it's CPUs. If it's Qualcomm, it's Snapdragon, smartphone SoCs and things like that. There are also companies like Broadcom, designing Google's TPUs—computing chips specifically for AI.
Next, let's take a look at some of the key players among the foundries that manufacture chips. I think the name you hear most often among foundries is TSMC. Especially when it comes to cutting-edge logic semiconductors, there are products that seem like only TSMC can make them. Then there's Samsung, Intel Foundry, SMIC, and Rapidus. Rapidus will also manufacture semiconductors equivalent to 2 nm in Hokkaido. I have high hopes for this because they're really giving it their all. There are also other companies like these that operate as foundries.
The second is an IDM that handles everything in-house, from design to manufacturing. It is a vertically integrated manufacturer. Intel, a leading example, manufactures CPUs for PCs and servers at its own factories, but as shown on the previous slide, it has recently begun venturing into contract manufacturing as well. Sony is the world leader in camera image sensors. Renesas is also a major manufacturer of microcontrollers for automotive applications, and Infineon is a leading manufacturer of power semiconductors. In addition to standard silicon power semiconductors, it also manufactures power semiconductors using materials such as silicon carbide, SiC, and gallium nitride, GaN, which can withstand high voltages and support fast switching.
Texas Instruments is a leading manufacturer of analog ICs that handle power, audio, and sensor signals. Japanese power semiconductor manufacturers such as Mitsubishi Electric, Fuji Electric, and ROHM also each have their own areas of expertise. Then there are companies like the ones listed below. They're not strictly IDMs—they also have some elements of fab-lite—but this time, I went with IDM.
Finally, here are memory manufacturers, prime examples of IDMs. Memory manufacturers also design their products in-house and manufacture them at their own factories. There are two main types of memory. DRAM is a place where data being processed is temporarily stored, and when the power is turned off, that data will be lost. HBM, which has been making waves in the AI world, is a type of high-bandwidth memory, and it's also a type of DRAM. By stacking multiple DRAM chips on top of each other, it's designed to allow the exchange of large amounts of data at once. NAND is a storage medium for photos and large amounts of data, among other things, and the data doesn't disappear even when the power is turned off.
As for what kinds of companies there are, the largest companies are concentrated in South Korea: Samsung and SK Hynix. Both companies seem like they'll be among the winners in the AI era. That's how it turned out. Micron is another well-known company. There's also Kioxia in Japan, which manufactures NAND memory. To increase speed and capacity, companies develop and manufacture DRAM and NAND products made by stacking multiple layers like this. In addition, Chinese companies such as CXMT and YMTC have been gaining ground recently.
Now that we understand the division of roles within the industry, from here on, we'll take a look at each step of the chip manufacturing process: design, wafers, masks, front-end process, back-end process, testing, and PCB assembly. Let's take a look at what kinds of companies are active in each stage of the process and what their strengths are.
Let's start with the design phase, which is the first step. Before looking at design firms, let’s start by clarifying who actually uses EDA and IP. Terms like “EDA” and “IP” suddenly came up. First of all, EDA refers to design software, and IP refers to circuit components. As we saw at the beginning, fabless companies and IDMs design chips. The chips both design contain tens of billions of transistors, and of course, it’s impossible to design them by hand. So they use design software called EDA.
Also, for circuits where the operations are predetermined, like the inner workings of a CPU or a USB interface, they buy ready-made circuit components called IP and combine those elements to create them on the chip. EDA and IP are used by fabless companies like NVIDIA and Apple, which have no manufacturing facilities, and the same is true for IDMs like Intel and Renesas, which operate their own factories.
The difference lies in where and by whom the semiconductors are manufactured. That's what the PDK is used for. It's called a Process Design Kit. It's sort of like a collection of design rules for each factory. This is because the shapes and sizes of semiconductors that can be manufactured vary depending on the semiconductor factory. Once you've decided which factory to use, you design the product in accordance with that factory's rules. Fabless companies outsource manufacturing. For example, after receiving a PDK from a foundry such as TSMC, they design the product and outsource manufacturing to a foundry. In the case of an IDM, designs are created using the company's own PDK, and they make it themselves.
Now let's take a look at the companies that develop EDA and IP. Here they are. In the EDA industry, Synopsys, Cadence, and Siemens are the three leading companies. Synopsys offers a full suite of core design tools and is also a major player in connection-related IP, such as USB for connecting chips to each other or PCIe for high-speed connections inside a computer.
Recently, they have been acquiring Ansys, a physics simulation company. If you stack multiple chips on top of each other, like with AI chips, heat builds up, causing the circuit board to warp or making it harder for signals to get through. Since the signals themselves operate at very high frequencies, they can interfere with one another. Moreover, since they all influence one another, it has become necessary to combine circuit design with physical analysis. I think that's the goal of the acquisition.
Cadence excels in analog and package design, and you can design the chip, package, and substrate as a single unit. Siemens has a product called Calibre, a tool for performing a final check before production. It has become a tool that is widely used as an industry standard. Other examples include Keysight from the U.S., which excels in high-frequency applications, or
Companies that perform component simulations, or examples such as Zuken for PCB design and Renesas' Altium. This is less about chip design and more about the design of the printed circuit board. And China is also working on developing its own EDA software.
Next is IP. Components that perform specific functions, such as the CPU and USB—if I had to design everything from scratch every time I made a chip, we wouldn't have enough time, and it would take a lot of manpower. And there is also a risk of problems occurring. So, we buy a circuit that has already been proven to work as a component and combine it with our company's circuitry.
But IPs aren't exactly cheap, are they? There's an initial contract fee, and in addition, a usage fee is charged for every chip sold. That's generally the case.
Arm is the leading IP vendor. Arm doesn't manufacture chips itself; instead, it licenses its CPU designs. And the idea is to make money from those usage fees. It has grown thanks to a pretty interesting business model. However, by 2026, it seems they're also taking the step of manufacturing their own chips for data centers for the first time.
Although it isn't mentioned in the IP vendor section, as I mentioned earlier, Synopsys and Cadence are also major IP vendors. As for other IP vendors, for example, RISC-V, a CPU instruction set which anyone can use for free, has also been expanding recently. Companies such as SiFive in the U.S. and Andes in Taiwan are involved in this field. Then there are things like GPUs and memory connections, connections in the chip's wiring, and so on. There are several IP firms specializing in each of these fields.
Now that the design is complete, we'll move on to the real world. To make a chip, you first need two things. One is a master plate on which the designed circuit's pattern is drawn on a glass plate. It's called a photomask. You could say this design is intended for making masks. Another key component is the silicon wafer, which serves as the foundation for building the circuit. Once these two are in place, the next step in the process involves using a mask and a wafer to make a semiconductor chip.
First, let's take a look at this photomask. Before we talk about photomasks, let's take a look at the relationship between design and masks. At the end of the design process, the pattern data for the completed circuit is delivered to the mask manufacturing plant. In the semiconductor industry, this is called a "tape-out." At the mask factory, they transcribe the circuit pattern from this data onto a glass plate to make a photomask.
There's a process called "exposure" within the front-end process. During that exposure process, light is shone through this mask onto the wafer to print the circuit pattern. The printing process isn't just done once; we keep repeating this hundreds and thousands of times.
So, do I only need to make one mask here? That's the question I have, though. That's not the case, is it? In the front-end process, we repeat the exposure multiple times, but since the circuit layout differs from layer to layer, each layer requires a different type of mask. As I'll explain later, when manufacturing semiconductors, we keep stacking the circuits one on top of the other. Therefore, a mask is required for each type of circuit being built.
Specifically, how many masks do you need? With state-of-the-art chips, a single product requires between 70 and over 100 masks. This set of masks is incredibly expensive. For cutting-edge semiconductors in the single-digit-nm generation, a single set of masks costs between 10 million and 20 million dollars or more. It ranges from over a billion yen to several billion yen.
And since this is the original mask, if there's even a single scratch on the mask, that part of every chip exposed through it ends up with the exact same scratch. Furthermore, if the mask is designed incorrectly, we'll end up having to remake that extremely expensive mask. So, if a semiconductor design fails, the masks have to be remade, and that would end up being extremely expensive.
Now, let's take a look at how photomasks are made. The first thing to keep in mind here is who is making the masks. It's true that mask manufacturers are the ones who make masks, but when it comes to cutting-edge masks, it is common for semiconductor manufacturers like TSMC, Samsung, and Intel to produce these in-house. As for other masks, there's DNP in Japan, for example, a mask manufacturer that sells to outside customers.
Let's take a look at the mask-making process step by step. First, here is a glass plate. It's made of a special type of glass used for masks, so it doesn't stretch or shrink. These blank plates, which have a layer that blocks light, are called "mask blanks." These materials are supplied by manufacturers such as HOYA and AGC.
And on these mask blanks, we use an electron beam to draw the circuit. These are the kinds of companies that sell devices that use an electron beam to write on surfaces. It says to check the circuit right after drawing it, but in the actual process, this includes processes such as development, etching, and cleaning. It's not written here.
In the inspection process, for example, in the case of an EUV photomask, you would use an EUV light source as well to perform the inspection. Some companies even manufacture those light sources in-house, and companies like Ushio in Japan and Gigaphoton, which is part of the Komatsu Group, also manufacture EUV light sources for mask inspection.
The term "EUV" just came up a moment ago. I'll explain this later, but this refers to the light source used when exposing advanced logic semiconductors. One of its features is that its short wavelength allows for the creation of extremely fine circuits.
Finally, it says "dust cover" here. This is the process of applying an extremely thin film called a pellicle. This is designed to keep dust from getting on the mask. Even if dust were to get on the pellicle, since there is a distance between the mask and the pellicle, this minimizes the impact of dust on exposure.
Next up is the wafer. This is the part that forms the base of the chip. The raw material is ultra-high-purity polysilicon. In Japanese, this is called "polycrystalline silicon." This is silicon consisting of a large number of small crystals clustered together. Companies such as Wacker, Hemlock, and Japan's Tokuyama manufacture these products. Melt this polysilicon and grow it into a single large crystal, then slice it thinly and polish it to a mirror-like finish to make a silicon wafer.
Shin-Etsu Chemical is the leading manufacturer of silicon wafers. Its strengths lie in its low crystal defect density and high flatness. What's more, Shin-Etsu Chemical even has materials for photoresists and masks, so it's kind of like a general-purpose materials manufacturer. Next up is SUMCO. This is a company that specializes in wafers and focuses on cutting-edge 300 mm products. And then there's GlobalWafers in Taiwan and companies like that.
What makes silicon wafers so amazing really comes down to the high purity. It says "99.9999..." and so on, but it is known as "11N" and has an extremely high purity—equivalent to about 11 nines in a row. That means you can't make semiconductors unless the purity is that high.
And here's another one. I know this is going to be a bit about pricing, but that's the number of chips that can be obtained from a single wafer. For example, for smartphone SoCs, it's around 500 units or so. With a high-end GPU, it's around 60 or 80. Generally speaking, the numbers fall within this range. Plus, as we go along, we end up with some defective products, so the actual number of usable ones will be even fewer than what's listed here. This is why the larger the chip, the higher the price.
Since higher prices mean higher costs, to avoid this, lately there's been the use of something called "chiplets": making small semiconductors, putting all those pieces together, and using them as a single chip. There are technologies like that as well.
On the previous page, it seemed like polysilicon just appeared out of nowhere, but in fact, we start with materials such as quartz and silica rock, rocks made of silicon and oxygen. This is heated at a high temperature along with carbon to remove the oxygen and produce metallic silicon with a purity of approximately 99%. However, that level of purity is still far from sufficient for semiconductors. The manufacturing process is a little tricky, but there is a process for converting that metallic silicon into polysilicon. The polysilicon itself is of extremely high purity, about 11N.
This is where the wafer manufacturer's work begins. Melt the polysilicon in a quartz crucible, dip in the seed crystal, then slowly pull it up while rotating it. This is known as the Czochralski method, and we grow a large ingot of single-crystal material. Next, trim the outer edge to shape it and use a thin wire to cut the ingot into slices.
If you just cut the disc, the edges are prone to chipping, and the thickness varies. Round off the edges by chamfering them, then shave it down from the top and bottom to even out the thickness. There are some surface scratches from grinding, but you can dissolve them with a chemical solution and remove them. Then, using an abrasive, we polish it until it's as smooth as a mirror. Finally, clean the part and inspect the surface for fine particles, scratches, flatness, and so on. The procedure is to pack the items in special containers before shipping them.
Wafer quality depends not only on how the ingots are made, but also on things like cutting, grinding, and polishing. It depends on the precision of such devices and the materials used. This is the page showing what kinds of companies are involved in actually making wafers. I won't explain everything, but as you can see, a great many companies are involved in the production of wafers. If you're curious, pause the video and take a look. What's interesting is that there are even containers for shipping. This part is pretty interesting, isn't it?
I'll digress a bit here, but let's also take a look at some specialty wafers other than silicon. Silicon carbide, SiC, is used in power converters that drive the motors of electric vehicles, EVs. The American company Wolfspeed handles the entire manufacturing process, from substrates to devices. It just seems that the company's financial situation in 2025 wasn't very good. And then there are things like InP substrates for optical communications, GaN substrates for high-frequency devices and optical devices, and GaAs substrates, among other things.
Next, now that we have the wafer and mask ready, we're finally moving on to the front-end process. This is the process of building the circuit in multiple layers on top of a wafer. It has the most steps, and companies that manufacture equipment, components, and materials are also the most numerous.
First, let's take a look at what is being produced in the front-end process. We're going to be making a chip, and this is actually what we're going to build. This is a cross-sectional view of the chip. This gray part at the very bottom is the silicon wafer. We'll build transistors on that surface. With the latest silicon chips, several billion to tens of billions, or even more than that, are being made.
The first half of the front-end process involves manufacturing these transistors. In the industry, it's called FEOL, Front-End of Line. Thermal oxidation, lithography, etching, ion implantation, and heat treatment are the main processes.
And piled on top of that is this copper wiring layer that connects the transistors to each other. There are about a dozen layers, and the wiring gets thicker as you go up. BEOL is used to create this layer. This section covers film deposition, lithography, etching, plating, and planarization as the main processes. Cleaning and inspection are incorporated into the process at various intervals throughout both stages. We'll combine these two elements using the process I'm about to explain.
In the front-end process, the eight steps are repeated numerous times in this order, and we build the circuit layer by layer. First, let's follow the process of how a single layer is formed. First, clean the wafer's surface, and then a film is deposited onto the wafer surface during the film deposition process. Next is lithography. Here, apply the resist, print the circuit pattern during exposure, and reveal its shape during development.
Using that shape, we either remove the film through etching or introduce impurities through ion implantation. Etching is a process used to create shapes, and ion implantation is a process that alters the properties of silicon. Since each has a different purpose, we choose one or the other depending on what we're making and which layer we're working on. Once you're done, remove the resist that is no longer needed.
For example, after ion implantation, heat treatment is also used to align the atomic structure. A layer whose surface has become uneven is polished until it becomes flat through the planarization process. Then, inspect it to make sure there are no defects or misalignments, and move on to making the next layer. And then it starts all over again with the cleaning.
From here on, let's take a look at the different types of players for each step. Before looking at companies by production stage, keep in mind that there are about three different perspectives on the companies involved in a single production process.
The first is an equipment manufacturer. This is a manufacturer that produces machines that perform the process itself. This equipment will be housed in semiconductor factories owned by foundries or IDMs. The second is a parts manufacturer. These manufacture parts used in those types of devices, such as lenses, power sources, valves, and platforms for holding wafers, and supply them to equipment manufacturers. The third is a materials manufacturer. These supply semiconductor factories with materials used in manufacturing processes, such as chemicals, gases, photoresists, and polishing fluids.
First up is the cleaning process. I'm sorry the text is a little small, but I've divided this into three categories: equipment, the components used within it, and materials. In the first cleaning step, we use a cleaning solution and ultrapure water to remove contaminants from the wafer. In terms of equipment, SCREEN is the world's best, and Tokyo Electron, Lam Research, and SEMES also work in this area.
There really is liquid inside the device, containing the chemical solution. There are sealing materials that won't be damaged even when exposed to chemicals, and robots inside that transport wafers, among other things. It does say "transport robot," but these are found inside most machines, even in front-end processes other than the cleaning step.
And then there are the parts that involve solutions or chemicals. This is an area where Japan has actually been quite strong. There are things like hydrogen peroxide and hydrofluoric acid, which dissolves oxide films. Next, we produce ultrapure water to rinse off any chemical solutions adhering to the wafer. There are machines like that, too.
After all, to make semiconductors, it's important to keep the wafer as clean as possible. This cleaning process is a very important step.
The second step is the film deposition process. These include CVD and ALD, which form films through chemical reactions involving gases; sputtering, which involves vaporizing metal and depositing it onto a surface; and thermal oxidation, which involves oxidizing the surface. We apply the film in increments of nanometers.
For film deposition equipment, Applied Materials is an American company with the widest range of offerings. Then there are companies like ASM in the Netherlands and Japanese companies that handle those kinds of things as well.
Those are the parts used inside the equipment. This could be a heater used to warm the wafer, for example—a ceramic heater. Also, when we're forming the film, we pass gas through it, so there are components for measuring the flow rate of that gas. Since we'll be using plasma, there's a power supply for the plasma system as well.
Then there are the materials for making the film, too. That aspect is also quite important. Things like creating metal films using sputtering, or using feed gas to create a film through chemical reactions. There are companies that supply raw materials used in those places.
Next up is lithography. First, let's take a look at the coating and development processes. In fact, the exposure process also takes place during this time. Coating and development involve applying a thin layer of photosensitive resist onto the wafer, then, after exposure, dissolving the unwanted resist with developer. This is the process of developing the circuit layout.
Tokyo Electron is the world leader in equipment. This coating and development equipment is called a coater-developer, and it is used in conjunction with an exposure unit.
Regarding this section on the coater-developer, the materials actually have quite a strong presence. Japanese companies are particularly strong in the field of photoresists. Companies like Tokyo Ohka Kogyo and JSR are developing resists for EUV, and Shin-Etsu Chemical also supplies photoresists for exposure using argon fluoride laser light, a technology from the previous generation.
In the component section, there are filters that remove minute foreign particles mixed into the resist. Another key component is a pump that accurately dispenses a fixed volume.
Next is a process that is considered one of the core steps in lithography. Let's take a look at exposure. This involves using a lens to reduce the size of the circuit etched onto the photomask and exposing the resist. With state-of-the-art EUV, light is absorbed by glass, so a mirror is used instead of a lens.
As stated here, the shorter the wavelength, the finer the detail you can achieve. The more detailed the design, the more transistors can be packed onto the chip. Since that's the nature of it, making things in small, detailed patterns becomes important.
Looking at the equipment, ASML, based in the Netherlands, is the only company capable of manufacturing EUV lithography systems. The ability to produce EUV allows us to design circuits with extremely high precision. In addition, Nikon and Canon of Japan also manufacture exposure systems. It's just that the lithography systems Nikon and Canon offer cannot produce circuits as fine as those produced by ASML's EUV lithography systems.
Let's take a look at the parts. The light source, which is the heart of the system, is manufactured by Cymer, a subsidiary of ASML. This is actually pretty interesting. If you watch other videos, you'll see what I mean. There's a part called a droplet generator that ejects the tin droplets described here. What exactly does that part do? It's a component that reportedly fires tin particles with a diameter of several tens of micrometers at a rate of tens of thousands per second.
By the way, the laser used to irradiate the tin is a CO₂ gas laser. Its wavelength is 10.6 micrometers, which is very long compared to EUV. Even if you shine that light on the wafer, you can't expose the fine circuit patterns. When a CO₂ gas laser is directed at a tin droplet, it turns into plasma. When it becomes plasma, it emits EUV light with a wavelength of 13.5 nm, which is very short.
The light emitted from there is directed onto the photomask. By reflecting that light, we're basically applying it to the resist on the wafer. The interior features a highly complex optical system.
Exposure using light known as deep ultraviolet, or DUV, from the generation before EUV is also still widely used today. In the case of ArF and KrF laser sources, which serve as these light sources, Japan's Gigaphoton is another manufacturer on par with Cymer.
The lenses and mirrors inside ASML's EUV lithography systems are made by Zeiss in Germany. The base of the mirror uses materials such as Corning's special glass that does not expand when heated. As for the raw materials, there are the mask substrates and pellicles mentioned earlier when we were making masks. There are also things like gas for the light source.
Next up is etching. This involves using plasma gas to etch away the film in areas not protected by the resist, etching the shape of a circuit. Lam Research is the leading manufacturer of such equipment, along with Tokyo Electron, Hitachi High-Tech, and Applied Materials.
There are various parts inside this equipment as well. For example, there's something called an electrostatic chuck, which uses static electricity to firmly hold onto a wafer. Since this equipment operates under relatively low atmospheric pressure, it's basically impossible to use air to hold a wafer in place. There's also the option of securing the wafer with claws, for example. After all, mechanical force is applied to the wafer, so it's not that great. In the end, we've ended up using an electrostatic chuck.
Then there's the high-frequency power supply used to generate plasma, vacuum valves, and electrode rings, among other things. There are also fluorine-based and chlorine-based gaseous materials.
Next is ion implantation. Impurities such as boron and phosphorus are converted into ions and implanted to create the properties of a transistor—that is, a switch—on a wafer. Major equipment manufacturers include Applied Materials, Axcelis, and Japan's Sumitomo Heavy Industries. Applied Materials comes up quite a bit, doesn't it?
We'll be embedding ions into the substrate, but a high vacuum is required to propel those ions in a straight line. That's why there's a vacuum pump inside. Regarding the raw materials, impurities are supplied in the form of gas. Since some of these gases can be quite toxic, containers for storing that gas are also important.
Next up is heat treatment. This process involves heating the atomic arrangement, which has been disrupted by ion implantation, to restore it to its original state. By activating the implanted impurities, it can function as a transistor. There are various heating methods. Some involve slow heating inside a furnace, while others heat the material in just a few milliseconds.
Next is the planarization process. This is also known as CMP. Using a polishing solution and a polishing pad, the wafer's surface is polished until it is flat through both chemical and mechanical action. Applied Materials is a strong player in the equipment sector, and Ebara Corporation is also included, for example.
Another characteristic of this process is that it uses a lot of materials. The polishing solution is also called a polishing slurry. Fujifilm, Entegris, and Resonac make it. Then there are polishing pads and conditioners. I didn't even realize there was such a thing as a conditioner. This is something you might not realize unless you actually look it up.
The next step is inspection and measurement. We check for foreign objects or defects after each process and measure the line width and misalignment of overlapping circuit patterns. There are companies such as KLA in the U.S. and Hitachi High-Tech in Japan. An electron microscope called a CD-SEM is used to measure the line width of the circuit. ASML also has equipment for measuring alignment errors.
To measure distances in nanometers, we need a mechanism to move the wafer precisely. THK is used for the linear guides in that area. Sensors for measuring position are also important. As far as I can tell from my research, since this process is just a matter of measuring, I couldn't find any consumable materials.
Let's list the steps we've covered so far once again: cleaning, film deposition, lithography, etching, ion implantation, heat treatment, planarization, and inspection. One full cycle creates one layer of the circuit. Then we go back to cleaning and start building the next layer.
As we've seen so far, once the entire cycle from cleaning to inspection is complete, circuits are formed on a wafer. Since the chip's circuitry is made up of dozens of layers stacked on top of each other, we repeat the same process over and over again. When it comes to state-of-the-art chips, even lithography alone, which is used to form circuit patterns, is said to be repeated anywhere from several dozen times to around 100 times. For this reason, it takes several months for a single wafer to complete the front-end process.
Once all the layers are complete, we perform a final electrical test. This is called WAT, or Wafer Acceptance Test. In other words, it's a wafer acceptance inspection. We've already created test patterns on the chip as well. We place a needle there and pass an electric current through it. We verify the quality of the manufacturing process to ensure that the transistors are being produced as intended. Whereas interim inspections focus on the form, the tests performed here are the final electrical check of the front-end process.
Up to this point, we also took a look at the parts used inside the equipment for each step. There are also common parts that are used across multiple processes. First, we'll introduce components found in equipment for processes performed in the atmosphere, including cleaning, coating, development, and CMP.
First, let's look at transport robots. These transport the wafers inside the equipment. Then there are sealing materials, which seal joints against vacuum and chemical solutions. There are also chillers for temperature control.
There are flow meters to accurately measure the gas flow rate, along with high-purity valves and fittings to stop or connect the gas without contaminating it, keeping it pure. There's also something called a gas panel that bundles those gas pipes together.
Now, regarding the second group of shared parts, these are for processes that use vacuum or plasma: film deposition, etching, ion implantation, EUV lithography, and inspections that utilize electron beams.
First, the electrostatic chuck. It uses static electricity to hold the wafer. Since you can't use air to create suction in a vacuum, you need an electrostatic chuck. This is an area where Japanese ceramic manufacturers are really making a strong effort.
Next is the vacuum pump, which creates a vacuum inside the equipment. There are also vacuum valves that separate the atmospheric-pressure side from the vacuum side. You'll also need a vacuum gauge to measure the vacuum level.
Gas is required for film deposition and ion implantation, so there are treatment facilities that neutralize that gas. There are also high-frequency power supplies for generating plasma, ceramic parts such as the walls and showerheads inside the unit—the components of the chamber—and cleaning and refurbishing services to wash dirty parts. There are also quartz components used in heat treatment and film deposition. There are all sorts of things like that.
There are also companies that play a key role in the infrastructure that keeps the factories themselves running. There's a system that automatically transports the containers holding wafers along a rail on the ceiling. Companies like Murata Machinery and Daifuku are particularly strong in this area. The container itself is also pretty heavy, and given the sheer volume involved, automated transport is essential.
In addition, we'll need ultrapure water for cleaning, sealed containers for transporting wafers, and facilities like cleanrooms for semiconductor manufacturing. These parts manufacturers have work to do no matter which manufacturer's equipment sells well. Personally, I think they're in a pretty good position.
To summarize the equipment manufacturers for the front-end processes we've looked at so far, I'd say there are about five major equipment manufacturers. ASML is the only company in the world capable of manufacturing EUV lithography systems. Applied Materials covers everything from film deposition to polishing and ion implantation. I think they're probably the most comprehensive manufacturer with the widest selection.
Lam Research is the leading manufacturer of etching equipment and excels in technologies for digging deep holes in 3D NAND. Tokyo Electron manufactures the world's leading equipment for applying and developing photoresist, used in conjunction with EUV lithography systems. Furthermore, KLA is the leading manufacturer of inspection equipment for detecting defects on wafers.
I forgot to mention SCREEN, the cleaning equipment manufacturer. Even among major manufacturers, their areas of expertise vary. You can see that it's separated pretty neatly.
Now we're finally moving on to the next stage. Wafers on which circuits have been formed in the front-end process proceed to the back-end process. In this process, the wafer is cut into individual chips, which are then mounted on a substrate and connected. It's like wrapping it all up with resin at the end. People tend to think of these as mundane processes, but with the advent of AI chips, this process has undergone significant changes.
First, let's take a look at what we'll be making in the later stages of the process. Once the circuit has been completed on the wafer in the previous process, it is shaped into components that can be incorporated into electronic devices. This is what the package looks like with all the parts inside. We're looking at a cross-section of that package.
The chip in the middle is the one we made in the previous process. The chip is connected to the package substrate via thin metal wires or bumps, which are small solder balls. The package substrate's function is to fan out the chip's fine terminals to larger terminals that mount onto the product's printed circuit board.
The entire component is encased in resin to protect the chip from moisture and impact. At the very bottom are the solder balls. These are used to solder components onto the printed circuit boards of smartphones and computers.
In other words, the back-end process involves cutting chips from the wafer, connecting them, and sealing them with resin, as described above. This is just a bit of supplementary information.
What I showed you in the cross-sectional view a moment ago looked like this: it was made up of fine metal wires connecting the chip, but cutting-edge semiconductors, such as smartphone and computer CPUs and the latest GPUs for AI, almost all use this format. It's a flip-chip design. To do this, flip the chip over so that the circuit side is facing down. It is designed to be mounted on a circuit board.
The components that connect this package substrate to the chip are called bumps. They're small solder bumps. Wires can only be routed from the outer edge of the chip, but with bumps, you can arrange several thousand of them across the entire surface, so you can significantly increase the number of terminals. Also, since the connections don't have to go through wires, the wiring is shorter, making it easier to transmit high-speed signals. The gap between the chip and the circuit board is filled with a resin called underfill. Measures have also been taken to prevent the bumps from cracking and to protect them.
Let's take a look at the back-end processing workflow here. In the front-end process, we repeated the same steps over and over again, but the subsequent steps are not repeated; the process is essentially one-way. However, we need to perform electrical testing before moving on to the next stage.
The first is the wafer acceptance inspection. This is the step performed at the very end of the front-end process. This was also explained in the previous slide. We do this to verify whether the transistor was built as intended. After that, we'll also perform wafer testing. This involves applying a thin needle to each chip while it is still in wafer form to check whether the chip is capable of operating when an electric current is applied. Defective chips will not be used in subsequent assembly.
After that, the back-end process begins. The first step is back-side grinding: grinding the back of a wafer to make it thinner. Next is the dicing process, which cuts the wafer into individual chips. At this point, we shift from working with entire wafers to working with individual chips. Next, place the chip onto the package substrate or lead frame. This is die bonding.
Next is the process of electrically connecting the chip to the circuit board: wire bonding using thin metal wires, or flip-chip bonding, where the chip is flipped over and connected directly via bumps. Then comes molding, the process of encasing the chip in resin. Next is the finishing process, such as engraving, and then the final test. When it comes to cutting-edge semiconductors, the way things are done in this general area is starting to change a bit. I'll introduce a few more things toward the end.
First, let's take a look at the wafer testing that takes place before the wafer enters the back-end process. Before wafers with circuits completed in the front-end process are diced, each chip undergoes an electrical test. A probe card—a bundle of fine needles—is placed against the chip's terminals. We then send a signal from the tester to that point to determine whether it operates correctly. Chips identified as defective will not proceed to the subsequent assembly process. This is because assembling defective products results in wasted effort and materials in subsequent processes.
There's a device that moves the wafer precisely and applies needles to each chip one by one. That's called a prober. Testing this wafer involves a somewhat complex setup. What does a tester do? It outputs a test signal and reads the returned signal. However, a probe card is required to connect the tester to the wafer. There are pins protruding from the probe card, and you press it against the chip to establish an electrical connection. Of course, the probe card and the tester are connected. So now the tester and the wafer are connected via a probe card.
Furthermore, there are hundreds of chips on a single wafer. A prober is the device that moves the wafer to the position of each of those chips. Therefore, when it comes to wafer testing, Advantest and Teradyne are pretty well-known tester companies, aren't they? Companies that make probers include Tokyo Electron and Tokyo Seimitsu. These two devices are combined and used as a single unit.
By the way, this tester is also used in the final test. This tester is essentially a collection of measuring instruments: a power supply that applies a voltage to a chip and measures the current, a circuit that sends a signal to each terminal one by one and evaluates the returned signal, and circuits for generating signal timing in picoseconds, among others. These are arranged in rows corresponding to the number of terminals—from a few hundred to a few thousand—and operated at the same time. Testing is a time-consuming process, so the key here is how quickly we can finish it.
The components include probe cards and measurement circuits. A probe card has several hundred pins and is used to contact circuits while they are still in wafer form.
From here, we'll move on to the next stage of the process. The first back-end processing step is back-side grinding. In the front-end process, we handled the wafers while they were still thick to prevent them from breaking, but as they are, they won't fit in a thin package. This is especially true when stacking multiple chips, such as HBM or memory.
First, let's talk about how to grind it down. Apply protective tape to the surface where the circuit was made to protect the circuit itself. Then, on the other side, we use a rotating grinding wheel to grind it down until it becomes thin. Japan's DISCO leads the world in this equipment. DISCO also manufactures its own grinding wheels.
An essential material here is back-grind tape. Back-grind tape serves several purposes. First, when grinding, shavings come off, and water is sprayed on it to cool it down. Its role is to protect the circuitry from those shavings and water. Second, it absorbs the unevenness of the circuit surface and supports the wafer so that it lies flat. If the tape is uneven in thickness, that variation will directly result in thickness variations in the wafer after it has been ground, so high precision is also required for the thickness of the tape.
Third, it peels off cleanly after you're done grinding. Since thinned wafers are prone to breaking, a type of tape whose adhesive weakens when exposed to ultraviolet light is used, allowing it to be peeled off without leaving any adhesive residue and without using force. This tape is made by Lintec or Mitsui Chemicals, among others. Then there's the grinding wheel used for grinding. Those aspects are also important.
Next is the dicing process. A single wafer contains hundreds to thousands of chips. Cut this into individual chips while leaving the tape on. A common method involves rotating a diamond blade at high speed to cut it while sprinkling water on it. In this sector as well, DISCO is the global leader. Along with Tokyo Seimitsu, it is believed to account for the majority of blade-cutting devices. The dicing tape applied when cutting the chips, just like back-grind tape, is made by companies such as Lintec. It seems that Japanese companies are strong in that area.
The next step is die bonding. Pick up the chips one by one from the tape. They are placed and secured on a base such as a package substrate or lead frame. Chips identified as defective during wafer testing are not picked up here; they are left as they are. For the equipment, Besi from the Netherlands and Singaporean companies are quite strong in this area. Then there are materials like lead frames and adhesive films. Looking at those materials really makes me think that Japanese manufacturers have a strong edge in materials.
Right now, I'm holding this in place with glue, but for chips that generate a lot of heat, such as power semiconductors, we also use solder joints to ensure good heat dissipation, or silver sintering, among other things.
Next is the process of electrically connecting the chip to the circuit board. There are two main methods. One method is wire bonding. We attach gold and copper wires, each thinner than a strand of hair, one by one. Another is flip-chip bonding. Turn the chip over and connect it directly to the circuit board using small solder beads on the surface called bumps. Because the wiring is short, it can transmit signals quickly, so it is commonly used for connecting CPUs and GPUs.
Examples of such equipment include wire bonders and devices that connect chips using bumps. In terms of materials, Tanaka Electronics supplies bonding wire. For filling the gaps between the bumps, you'll see companies like Namics and Resonac.
The next step is the molding process. If the chips and wires are exposed, they're susceptible to moisture, dust, impacts, and light, so they are completely covered and wrapped in resin to protect them. Basically, the chip mounted on the circuit board is clamped between the mold, and resin is introduced and allowed to harden. There are methods like compression molding and transfer molding.
Looking at the equipment, a company called TOWA has a fairly large market share. It is particularly strong in methods that involve pressing resin down from above, which are considered suitable for encapsulating thin, stacked chips, such as those used in HBM. Then there's the encapsulant. This is epoxy resin with something mixed into it, used as a sealing material. Sumitomo Bakelite is considered the world leader. Other companies include Resonac and Shin-Etsu Chemical. When I see something like this, I can tell that the materials sector is pretty strong.
Next comes finishing. We also do things like laser engraving and preparing the terminals for attachment to the printed circuit board. A commonly used package is called BGA. It's one of those with little beads of solder lined up on the back. Since we can achieve a very high-density layout, we finish those terminals and so on.
Next up is the final test of the finished product. Insert the packaged chip into the socket, and apply a signal from the tester to verify that it operates correctly. We use equipment for inspecting finished products like these, along with test socket components into which the chips are inserted. There are also companies that undertake this testing on a contract basis.
Let's take a moment to review the back-end process flow once again. After selecting good chips during wafer testing, grind down the back side to make it thinner, cut it into individual pieces, place them on the circuit board, and connect them electrically. We encase them in resin and finish the engraving and terminals. Then, after the final test is over, it's time for shipment.
Now that we've looked at this series of steps, there are companies that handle all of these back-end processes as a package. These are OSATs, companies that specialize in providing back-end manufacturing and testing services. From what I've found, ASE is the industry leader. There are companies that do this kind of assembly, but recently there has been a trend toward integrating multiple chips into a single unit, such as with AI chips. Advanced packaging is becoming increasingly important, and front-end companies such as TSMC and Intel have also started handling the packaging themselves.
We've looked at the back-end processing steps so far. This is a bit of a digression, but let's take a look at why back-end processes are attracting so much attention. Until now, chip performance has been improved primarily in two ways. First, through miniaturization, we're making transistors smaller and packing as many transistors as possible into the same area. Another approach is to make the chip itself larger, increasing the number of transistors that can be mounted.
But both of them have hit a wall. As miniaturization progresses with each generation, the equipment and processes also become larger, and costs are rising rapidly. Regarding the size of the chip, the larger the chip, the higher the probability that a single defect will cause the entire chip to fail, and the percentage of good products will decline.
That's why the concept of chiplets is often used in this context. This involves manufacturing small chips, each with a specific function, and then selecting only the good ones. Arrange them side by side or stack them vertically inside the package, and reconnect them at a high density. Doing this effectively makes the chip bigger. With HBM used alongside GPUs for AI, stacking the memory right next to the GPU also allows large amounts of data to be transferred over short distances.
From here on, we'll take a look at the package substrates that form the foundation of the system and how they make connections. Here's what's inside the package substrate. The terminals on the chip are very fine, and the terminals on the product's printed circuit board are large. The package substrate serves as the bridge between them.
The part in the center is called the core. It is made by impregnating glass cloth with resin. It provides support to prevent warping due to heat. Nittobo's T-glass is commonly used here. Layers of insulating film and copper wiring are alternately stacked on top of and beneath it. This is called the build-up layer, and this insulating film is Ajinomoto's ABF. The spacing between the copper traces is gradually increased layer by layer, connecting the fine terminals on the chip above to the large solder-ball terminals below. Mitsui Metal's ultra-thin copper foil is used in this wiring. The outermost green layer is the solder resist, which protects the wiring.
The larger the chip for AI, the more important the material of these circuit boards becomes. The quality of the copper wiring and build-up layer is also crucial. That's about it for the circuit board materials. As for the manufacturers that actually produce those circuit boards, there are companies like Ibiden. Regarding package substrates, this introduction ended up being a bit rushed, and there are a few parts where the explanation is incomplete. I'd like to take a thorough look at this part of the package in another video.
Next, let's look at this slide. This shows how the chips are connected inside the package. When it comes to AI chips, it's not just about the performance of a single chip. How closely and compactly multiple chips can be connected, and how many wiring buses there are, determine performance.
First up is flip-chip bonding. This method involves flipping the chip over and connecting it directly to the circuit board via bumps. This is a fairly standard packaging method for high-performance chips.
Next, take a look at this. On top of an interposer, which is a silicon-based intermediate substrate, we arrange extremely high-performance semiconductors, such as GPUs and HBM, side by side. Many GPUs designed for AI take this form.
The reason an interposer is needed is that this green circuit board is made of an organic substrate. Things like the fineness of the wiring and the spacing between vias cannot be made with the same level of precision as with silicon. In other words, it can actually be quite difficult to connect the GPU and HBM on this organic substrate. Plus, there are quite a lot of connections to make, which makes it even harder. Therefore, through precision machining of silicon,
Connect the GPU to the memory. Next, connect the wires from the interposer to this organic substrate. The idea is that they will ultimately be connected to the printed circuit board. That's what we do.
As such, the back-end process is changing to determine the chip's performance itself, and more sophisticated technologies closer to the front-end process are making their way into the field.
And finally, it's time for PCB assembly. The finished chips are mounted on the power supply boards of smartphones, computers, and other devices. It is only when mounted alongside other electronic components that a chip finally fulfills its function. Here, too, there are placement machines that load components, and Japanese companies have a competitive edge in electronic components related to chips.
Here are the specifics: To use the finished semiconductor, we need to mount semiconductors onto the printed circuit board like this. When mounting semiconductors and electronic components onto that printed circuit board, we'll use a placement machine. This is an area where Japanese manufacturers, such as FUJI, Yamaha Motor, Panasonic Connect, and JUKI, are quite strong.
Next are the passive components surrounding the semiconductor chip, MLCCs in particular. Thousands of MLCCs are mounted on the circuit boards of GPUs used for AI. When a GPU for AI starts up, billions of transistors turn on and off all at once. When the computation begins, a current of several hundred amperes flows for a duration of nanoseconds. This will cause a voltage drop in the wiring leading up to the chip.
On the other hand, the voltage required by this chip is around 1 V, and voltage fluctuations are limited to only a few tens of mV. If the voltage fluctuates, the GPU won't work properly, so the key is how to minimize those voltage fluctuations.
The power supply doesn't have a very fast response time, does it? It's in milliseconds. So, place the capacitor near the GPU to instantly supply charge from the MLCC and minimize voltage fluctuations. Also, minimize the power supply impedance. That is why MLCCs are so important. This part is also pretty important, so I'd like to make another video.
So, here's a summary. I learned that, in the process of making a single chip, many countries and companies are producing it by dividing up the production steps, each in their own way. Even changing just one step in this process can affect the subsequent steps. I suppose this might be a bit of a challenge.
I've summarized what I found out while doing my research here, so if you found this video helpful, please give us a thumbs-up and subscribe to our channel! If you could leave any additional information in the comments, that would be great. I think other people who watch this video will find it helpful, so I look forward to working with you. Thank you for watching the video all the way to the end.
Article published
