Stage 8 of 14Extreme concentrationVery hard to substitute
Chapter 8: Foundries and Fabs
The newest chips need the know-how to build the newest transistors and a fab that costs tens of billions of dollars. Only TSMC, Intel and Samsung have both.
Sharma, Yashvardhan. “Foundries and Fabs.” Chip Supply Chain: How AI Chips Are Made. Updated 29 October 2026. https://chipsupplychain.org/foundries-and-fabs/
A gate-all-around transistor, the switch in a 2 nm-class chip, drawn at true proportions. The three silicon sheets are each 6 nm thick and 30 nm wide. Each sheet is a few dozen atoms from top to bottom, roughly ten thousand times thinner than a human hair. The gold cage is the gate metal wrapping all four sides of each sheet.
In plain terms
A , or fab, builds the , the billions of tiny switches in every , and then wires them together. Each switch works like a valve. Current runs along a narrow strip of silicon called the channel, and a small voltage on a gate above the strip opens or shuts the flow. For decades chipmakers made the strips shorter, so the transistors switched faster and more of them fit on a chip. But once the strip gets short enough, current leaks through even when the gate is off, and the chip burns power doing nothing. The newest kind of transistor, called , cuts this leak by wrapping the gate around all four sides of the strip. A chipmaker needs two things to make the newest chips, and having one does not bring the other. The first is the know-how to build the newest switches, a long sequence of steps that have to run in exact order. In 2026 only TSMC, Intel and Samsung had it. The second is factories that cost tens of billions of dollars each, and in 2022 most of the world’s capacity for the most advanced was in Taiwan. China’s SMIC runs factories, but without machines or the know-how for the newest switches, its most advanced chips are . IBM developed 2 technology, but it no chips until Rapidus finishes a factory to run it, with planned for 2027. No chip company sells enough of one product to fill such a factory on its own, so chip companies share a handful of them, run by that make chips for other companies. TSMC runs the biggest, and all six of its largest plants are in Taiwan.
In short
Since September 2024 the United States has required a license to share the know-how for building , except with countries that impose matching controls1. China has announced $43 billion to develop its own (EUV) , but in September 2025 the judged that EUV remains firmly out of reach2.
SEMI projected in 2026 that China would keep the world’s largest installed capacity in 2026 and 20273. In July 2026 TSMC raised its planned investment in Arizona to $265 billion across 12 facilities4. SIA and BCG projected in May 2024 that the US share of the world’s logic capacity below 10 would rise from almost none in 2022 to 28 percent by 20325.
Nvidia’s Blackwell holds 208 billion switches on two , each die the largest a can print in one shot6. Every switch has to turn on fully, let almost no current through when off, and keep doing it for years. Blackwell is made by TSMC, on a custom process called 4NP6. TSMC calls a plant a GIGAFAB once it can run more than 100,000 , each a disc, twelve inches across7. It operates six of them, and every one is in Taiwan8.
- Who leads
- TWTSMC N2, its first nanosheet node, in volume production since 4Q2512
- USIntel 18A in high-volume production in the United States, as of 202613
- KRSamsung Foundry 3 nm gate-all-around process in initial production since June 202214
- CNSMIC $9.327 billion revenue in 2025, with its lines 93.5 percent full15
- JPRapidus 267.6 billion yen raised in February 2026 to reach 2 nm mass production by 202716
- Where it is made
- TWTaiwan TSMC’s six gigafabs, with N2 at Fab 20 Hsinchu and Fab 22 Kaohsiung, and UMC
- KRSouth Korea Samsung Foundry, Hwaseong and Pyeongtaek
- USUnited States Intel 18A at Fab 52 in Arizona, TSMC Arizona, Samsung Taylor and GlobalFoundries
- JPJapan Rapidus Chitose, Hokkaido, and TSMC Kumamoto
- CNChina SMIC Shanghai and Beijing 300 mm lines, and Hua Hong
- Why substitution is slow
- Since September 2024 the United States has required a license to share the know-how for building , except with countries that impose matching controls1. In March 2026 the Semiconductor Industry Association (SIA) put the cost to build and equip a at as much as $20 to $25 billion17. SemiAnalysis judged in December 2025 that finding the highly skilled engineers a foundry needs remained a major challenge for TSMC in Arizona18.
China’s SMIC made its most advanced on N+3, a process built without , SemiAnalysis found in June 202619. SMIC cannot buy those scanners, so it prints its tightest layers in on older machines19. China will keep the world’s largest installed fab capacity in 2026 and 2027, SEMI projected in its third-quarter 2026 forecast3.
Intel is the only chipmaker that both develops and manufactures leading-edge in the United States, by its own account in August 202520. The United States had almost no logic capacity below 10 in 2022, and SIA and BCG projected in May 2024 that its share would reach 28 percent by 20325. In July 2026 TSMC raised its planned investment in Arizona to $265 billion across 12 facilities4. The US government agreed in August 2025 to buy a passive 9.9 percent stake in Intel, with no board seat20.
How a transistor works and why it leaks
Why the gate wraps around the channel

- Inside a chip. An AI chip does all its work with transistors, tiny switches built into the silicon. One Nvidia Blackwell GPU holds 208 billion of them. Making the switches is the front end of the factory, and wiring them together comes after.
- On. Each one is a switch. Seen end on, current flows through a strip of silicon, the channel, and the gate sits on top.
- Off, leaking. Switching off, the gate squeezes the flow shut, but only near itself. In a tiny transistor, current still leaks through the part it cannot reach.
- Fin. Stand the channel on edge as a fin, and the gate covers three sides. Much less gets through.
- All around. Split it into thin sheets and wrap the gate around every side, and it shuts far tighter. That lets the switches keep shrinking, so each new chip can pack in more of them.
Current leaks in two main ways, straight through the gate and along the channel from end to end while the gate is off. The second leak grows as the channel gets shorter. Three fixes deal with the two leaks:
- Better insulator. A thicker hafnium-based film, paired with a metal gate, lets fewer electrons slip straight through yet controls the channel as tightly as before. Intel replaced the silicon dioxide insulator with it at 45 in 200721. It said the new film cut leakage through the insulator more than tenfold21.
- More gate. A planar lays the gate on one side of a flat channel, and as the channel shortens the gate loses control of it. A stands the channel on edge as a fin, so the gate covers three sides, and wraps it around all four.
- Thinner channel. The thins the silicon until no current can flow where the gate cannot reach it.
How transistors are built
Making the switches is called the . It builds them in a fixed order, starting with the channel. In a FinFET, the older design, the fab the silicon into thin fins that stand up from the , and the gate goes over each one.
In a gate-all-around transistor, the fab instead grows alternating layers of silicon and silicon- and cuts the stack into narrow bars22. It lays a placeholder, called a dummy gate, across each bar where the real gate will go23. On either side of the placeholder it cuts deep trenches where the source and drain, the two ends of the channel, will go24. Through those trenches it etches the silicon-germanium layers back a little and fills the gaps with an insulator, the inner spacer, which keeps the gate apart from the source and drain22. Then it grows the source and drain as crystal inside the trenches, a step called 24.
Only then does the fab remove the placeholder and etch away the silicon-germanium, leaving thin silicon sheets that hang between the source and drain like shelves22. The inner spacer shields the source and drain from that etch23. The fab coats every sheet with an insulating film a few atoms thick, so the gate can control the channel without touching it. Gate metal then fills the gaps above, below and beside each sheet, covering all four sides25. Last come the contacts, small metal plugs down to the source, the drain and the gate, ready for the wiring layers above.
The steps have to run in exact order, because later layers cover earlier ones. Once gate metal surrounds a sheet, the insulating film can no longer go between the sheet and the metal. Each step is a separate , etch or . An implant step, called ion implantation, fires atoms of boron or phosphorus into chosen patches of silicon so they conduct the way the design needs.
The newest transistor designs
Gate-all-around nanosheets
A nanosheet transistor lays the channel flat again, stacks two to four thin sheets and wraps gate metal around each one. Each fin in a FinFET adds a fixed amount of current, so a designer who needs more current has to add a whole fin. Nanosheets can be drawn at different widths, so current can be set in finer steps.
TSMC’s N2, its first nanosheet , entered in the fourth quarter of 202512. TSMC rates the follow-on N2P at 18 percent more speed at the same power and 36 percent less power at the same speed than its earlier N3E26. It gives N2P 1.2 times the of N3E and 1.15 times its chip density, but its technology page does not define either measure26. These numbers are TSMC’s own claims until someone measures the silicon. Each picks its own baseline, and none publishes a density figure that can be compared with a rival’s.
Intel and Samsung also make gate-all-around transistors. Intel’s version, called RibbonFET, is built into its 18A process, which Intel says is in high-volume production in the United States13. Samsung got there first, starting initial production of its 3 nm gate-all-around process in June 202214. But SemiAnalysis reported in February 2023 that, because of problems, high-volume on Samsung’s 3 nm gate-all-around technology were expected only in 202427.
SemiAnalysis reported in 2024 that the cell, which holds one bit in the fast memory on a , had barely shrunk since the 5 nm node28. TSMC said in 2025 that its 2 nm SRAM uses a 0.021 square-micrometer cell and gets 1.1 times the density of its previous node by tuning design and process together29. See Memory and HBM.
Backside power delivery
In a conventional chip, power and signal share one stack of copper wiring above the transistors, where wide power rails crowd out the signal wires. With , the fab grinds the wafer thin and builds a second network for power underneath it.
Intel’s version, PowerVia, is part of 18A, and Intel calls it an industry first13. It cut routed area, the space a block of circuits takes once its wiring is laid out, by 11 percent and dynamic voltage droop tenfold30. Droop is the sag in supply voltage when a block of logic switches on at once. TSMC’s Super Power Rail connects the backside network directly to each transistor’s source and drain28. Against N2P, TSMC rates A16, which uses Super Power Rail, at 8 to 10 percent more speed, 15 to 20 percent less power and up to 1.10 times the chip density26.
TSMC plans to start A14 production in 2028, with more than 20 percent higher logic density than N231. It said in April 2026 that two more processes, A13 and A12, are scheduled to enter production in 202932.
TSMC plans four logic nodes beyond N2 for 2028 and 2029
TSMC’s logic roadmap beyond N2, from its own announcements
- N2U, 3–4% faster or 8–10% lower power than N2P
- A14, the base node for A13 and A12
- A13, a direct shrink of A14 for 6% area savings
- A12, the A14 platform with Super Power Rail backside power
Sharma, Yashvardhan. “Figure 08.2: TSMC’s logic roadmap beyond N2, from its own announcements.” Chip Supply Chain: How AI Chips Are Made. Updated 29 October 2026. https://chipsupplychain.org/foundries-and-fabs/#fig-foundries-and-fabs-node-roadmap. Data: TSMC, 2026 North America Technology Symposium, 23 April 2026.
CFET and what comes after
A basic logic circuit needs two transistors that switch on opposite signals, and in production chips the two sit side by side. Complementary field-effect transistors, or CFETs, stack the two on top of each other33. The research institute imec sees them as an attractive design for logic nodes beyond 1 nm33.
The wiring above the transistors
Wiring the transistors together is called the , and it has become as hard to make as the transistor. Copper, the usual wiring metal, seeps into the insulator around it, so every wire needs a barrier, a liner and a cap34. As the most critical wiring levels drop below a 20 nm , those extra layers take a large share of the wire’s width34. Pitch is the distance from the center of one wire to the center of the next. Two metals are moving into those narrow levels.
- Ruthenium shows promise for use without a barrier, which would keep the wire’s resistance low34. imec has demonstrated ruthenium lines at a 16 nm pitch35.
- Molybdenum in a thin film has up to 30 percent lower resistivity than tungsten, the metal it is replacing in contacts36. It needs no barrier either, so chipmakers are putting it into some contacts and local wiring layers36.
What foundries do and how a fab runs
A foundry makes chips designed by other companies, running all their designs on the same tools. TSMC and UMC, both based in Taiwan, make chips only for other companies. TSMC, founded in 1987, pioneered this pure-play model37.
Inside a fab, a chip is built up in layers, and a similar loop of steps repeats for each one. Wafers travel in batches, called , inside a sealed pod called a FOUP, short for front-opening unified pod. Entegris, a US maker of these pods, offers a standard version that holds 25 wafers38. An overhead track carries the pod from tool to tool.
At each tool a robot lifts one wafer out and the tool does its one job, such as depositing a film, printing a pattern, etching it, cleaning it or measuring it. Then the robot puts the wafer back, and the pod moves on to the next tool, often across the building. When a layer is finished, the loop starts again with the next one.
Turning silicon into a finished, packaged chip takes 1,000 to 1,500 steps, depending on how complex the chip is, the Semiconductor Industry Association said in 202539. In 2021 the association put the time to make a finished wafer at about 12 weeks on average, and up to 14 to 20 weeks for advanced processes40. The pod keeps the wafer clean for all of that time, because the wafer never meets the room’s air.
How a fab builds a chip, layer by layer

- The fab. Sealed pods carry wafers from tool to tool on a track near the ceiling.
- Deposit. The first tool lays a thin film on the wafer.
- Print. The next prints a pattern on it with light.
- Etch. An etch tool cuts the pattern into the film.
- Clean, measure. The wafer is cleaned, then measured to check the layer came out right.
- Next layer. Then the loop starts again for the next layer. The transistors come first, then the wiring above them.
- Finished. Making one chip takes 1,000 to 1,500 steps, the Semiconductor Industry Association estimated in 2025.
Each tool is set by a recipe, the list of gases, temperatures and times for one step. A foundry offers every customer the same list of recipes, so every design in the building can run on the same shared tools. The tools run day and night, because an idle tool earns nothing. The finished wafer is tested (see Test and assembly), then cut and packaged.
Chips at 7 nm and below, meaning the newest processes, were 79 percent of TSMC’s wafer revenue in the third quarter of 202641. That share split into 33 percent at 3 nm, 28 at 5 nm, 10 at 7 nm and 8 at 2 nm41. TSMC’s revenue for the quarter was $46.72 billion, at a 66.4 percent 41.
Intel and Samsung design and sell their own chips and rent out capacity too. Renting out capacity is harder for them, because customers dislike handing their designs to a competitor and each company’s own products take its best capacity. Intel Foundry reported $5.8 billion of segment revenue in the second quarter of 2026 and an of $2.089 billion42. Only $293 million of that revenue came from outside customers9.
The cost and capacity of the newest fabs
A new fab costs tens of billions of dollars. In 2023 the Boston Consulting Group (BCG) estimated that a new fab for chips at 2 nm and below, completed in 2026, would cost $35 to $43 billion to build, equip and run over ten years10. Building and equipping alone takes as much as $20 to $25 billion of capital for a leading-edge fab, the Semiconductor Industry Association (SIA) told the US Senate in March 202617.
Most of TSMC’s capacity is in its six GIGAFABs. Together, the plants TSMC and its subsidiaries run had capacity for more than 17 million wafers a year in 2025, counted as if each were the standard 12-inch size8. The six largest plants had a combined capacity of more than 13 million7. The other plants are five smaller fabs in Taiwan, two in China, two in the United States and one in Japan8. In July 2026 TSMC announced another $100 billion for four more fabs at 2 nm or below, taking its planned investment in Arizona to $265 billion across 12 facilities4.
In 2025 the biggest AI chip designers used most of the world’s supply of two inputs but only a small share of its advanced chipmaking. Epoch AI estimated that the four largest used around 90 percent of the world’s packaging capacity and memory supply43. CoWoS is TSMC’s packaging for AI chips, and HBM is the stacked memory that sits beside them. The same four used only 12 percent of advanced logic production43. See Advanced packaging.
AI have usually trailed the leading edge. A die as large as theirs is more likely to hold a defect44. So the share of dies that work, called the yield, gets high enough only once a process has run long enough to cut its defect rate.
Nvidia’s Blackwell is made on an older TSMC process than AMD’s MI455X. Epoch AI estimated in December 2025 that a 300 mm wafer on Blackwell’s 4NP process costs $17,00011. AMD’s MI455X, launched on 23 July 2026, is made on TSMC’s 2 nm and 3 nm processes45. It spreads its compute across eight separate dies46.
TSMC, Intel, Samsung, SMIC and Rapidus
TSMC, Intel and Samsung build gate-all-around nodes with (EUV) . China’s SMIC builds nodes without it19. SemiAnalysis reported in September 2026 that Intel’s Panther Lake, Samsung’s Exynos 2600 and Apple’s A20 Pro, which is made on TSMC’s N2, all use gate-all-around transistors47.
In August 2025 Intel called itself the only semiconductor company that both develops and manufactures leading-edge logic chips in the United States20. That month the US government agreed to buy 433.3 million newly issued Intel shares at $20.47 each, or $8.9 billion in all, for a 9.9 percent stake20. The stake is passive, with no board seat20. The deal also gave the government a five-year warrant to buy a further 5 percent of Intel at $20 a share20. It can use the warrant only if Intel’s ownership of its foundry business falls below 51 percent20.
In 2026 Intel began high-volume manufacturing of some of its Panther Lake processors using ASML’s extreme ultraviolet scanners42. But it is building fewer fabs than it planned. In July 2025 Intel said it would no longer go ahead with planned projects in Germany and Poland and would slow construction in Ohio48.
In December 2024 the US Commerce Department awarded Samsung up to $4.745 billion of funding, to support more than $37 billion of planned investment in Central Texas49. That plan covers two new leading-edge logic fabs and a research fab in Taylor, and an expansion of Samsung’s existing plant in Austin49.
China’s SMIC has the factories without the know-how for the newest transistors. Its revenue rose 16.2 percent to $9.327 billion in 2025, at a 21 percent gross margin15. Its lines ran 93.5 percent full, on monthly capacity above a million wafers counted in the smaller 8-inch size15.
A node’s name, such as 7 nm, does not measure any feature on the chip. SMIC’s N+3, its third-generation 7 nm process, sets its narrowest wires 32.5 nm apart, tighter than the 36 nm in Intel’s Panther Lake chips on 18A19. But the 18A process itself allows wires 32 nm apart19. SemiAnalysis called the comparison an incomplete, cherry-picked metric19. N+3 packs 113.4 million transistors per square millimeter, against 107.7 for TSMC’s N619.
Two research groups disagree about where Huawei’s Ascend AI chips are made. Georgetown University’s Center for Security and Emerging Technology (CSET) reported in June 2024 that the Ascend 910B was believed to be made on SMIC’s N+2 process50. But the research firm SemiAnalysis judged in April 2025 that most Ascend 910B and 910C chips were made on TSMC’s 7 nm process51. It said Huawei got around the sanctions that barred it from TSMC by buying about $500 million of 7 nm wafers through another company, Sophgo51.
IBM shows the opposite split. Its 2 nm technology yields no chips until Rapidus, a state-backed newcomer in Japan, finishes a factory to run it. The two firms agreed in December 2022 to develop that technology further for production at Rapidus’s plant in Japan52.
In July 2025 Rapidus said it had started making 2 nm prototypes at the plant, in Chitose, Hokkaido, and would begin mass production in 202753. In February 2026 it raised 267.6 billion yen from the Japanese government and 32 companies to reach 2 nm mass production by 202716. By June 2026 Rapidus had raised 424.95 billion yen in equity54.
Why leading-edge chipmaking is so concentrated
A firm that wants to make the newest chips has to develop its own process, because TSMC, Intel and Samsung do not sell their leading-edge process flows. A process flow sets the steps, their order and the tools for each. In September 2024 the United States began requiring a license to share the know-how for building gate-all-around transistors, except with countries that impose matching controls1.
China has money for chipmaking but cannot buy EUV scanners. It launched its third in May 2024 with 344 billion yuan, about $47 billion55. China’s 7 nm chips have been made on machines, a generation behind EUV2. China had stockpiled those machines from ASML, the Dutch scanner maker, before export restrictions took effect2. SMIC gets fine features out of them by building its tightest layers in several patterning steps (see figure 04.2)19.
In July 2026 Reuters reported that a Shanghai state firm had begun making home-grown scanners56. Reuters’s own source said the firm’s machine still needed testing and was far from matching ASML’s models56.
Domestic EUV is further off. China has announced $43 billion to develop it2. In September 2025 the (CSIS) judged that EUV remains firmly out of reach despite massive state investment2. See Lithography.
The factories are concentrated too. In 2022 South Korea and Taiwan held almost all of the world’s logic capacity below 10 nm, and the United States had almost none57. Taiwan alone held 69 percent5. In May 2024 SIA and BCG projected that the US share of that capacity would reach 28 percent by 20325.
China leads the world in capacity for logic chips made at 28 nm or older, but little of its capacity is advanced58. Most of the world’s wafers come off such , which make power devices, analog chips and microcontrollers. Across all nodes, old and new, the industry association SEMI forecast in June 2024 that China would have installed capacity of 10.1 million wafers a month in 2025, the most of any region59. It forecast 5.8 million for Taiwan, out of a world total of 33.7 million, all counted in the 8-inch size59. In the third-quarter 2026 update of its World Fab Forecast, SEMI projected that China would keep the largest installed capacity in 2026 and 20273.
China led SEMI’s 2025 capacity forecast at 10.1 million 8-inch-equivalent wafers monthly
SEMI’s June 2024 forecast of installed wafer fab capacity by region in 2025, all nodesmillion 8-inch-equivalent wafers per month
| China | 10.1 million 8-inch-equivalent wafers per month |
|---|---|
| Taiwan | 5.8 million 8-inch-equivalent wafers per month |
| South Korea | 5.4 million 8-inch-equivalent wafers per month |
| Japan | 4.7 million 8-inch-equivalent wafers per month |
| Americas | 3.2 million 8-inch-equivalent wafers per month |
| Europe and Mideast | 2.7 million 8-inch-equivalent wafers per month |
| SE Asia | 1.8 million 8-inch-equivalent wafers per month |
Sharma, Yashvardhan. “Figure 08.5: SEMI’s June 2024 forecast of installed wafer fab capacity by region in 2025, all nodes.” Chip Supply Chain: How AI Chips Are Made. Updated 29 October 2026. https://chipsupplychain.org/foundries-and-fabs/#fig-foundries-and-fabs-capacity-by-region. Data: SEMI World Fab Forecast, June 2024.
Chip production in Taiwan faces three risks:
- Seismic. Taiwan sits on an active plate boundary, and after an earthquake in January 2025 TSMC had to scrap some wafers in process60.
- Power. TSMC says power outages, dips or surges caused by problems at its electricity supplier have at times interrupted its operations61.
- Political. In December 2025 SemiAnalysis put Taiwan’s share of the world’s advanced-node capacity at over 90 percent, though it did not say which node sizes count as advanced18. So a blockade or conflict would cut most of that output off from the world market at once. See Who Controls the Chain.
Building capacity outside Taiwan is slow, and TSMC has yet to build a gigafab anywhere else8. SemiAnalysis wrote in December 2025 that a foundry needs highly skilled engineers to get the most working chips from each wafer18. It judged that finding enough workers remained a major challenge for TSMC in Arizona18.
Flashcards
Recall each answer, then turn the card.
What does a chip factory do?
It builds the transistors, billions of tiny switches, and wires them together into chips.
Why do the newest transistors wrap the gate around the channel?
What two things does a chipmaker need to make the newest chips?
The know-how to build the newest transistors, and factories that cost tens of billions of dollars each.
What is a foundry, and which is the biggest?
A company that makes chips designed by other companies. TSMC runs the biggest, and its six largest plants are all in Taiwan.
How does China’s SMIC make advanced chips without EUV?
It makes its 7 nm-class chips on older deep ultraviolet machines, a generation behind EUV.
Five things to remember
- The newest chips need both the know-how to build the newest transistors and factories that cost tens of billions of dollars, and in 2026 only TSMC, Intel and Samsung had both.
- The newest transistors wrap the gate around all four sides of the channel to cut the current that leaks when the switch is off.
- A foundry makes chips for other companies, because no chip company sells enough of one product to fill a fab alone.
- All six of TSMC’s largest plants are in Taiwan, which held 69 percent of the world’s logic capacity below 10 nm in 2022.
- China’s SMIC makes 7 nm-class chips without EUV by building its tightest layers in several patterning steps.
Sources (61)
Source tiers A company filings, governments, standards bodies, universities and research institutions. B SemiAnalysis, Reuters, the Financial Times and Nikkei Asia.
- ACommerce Control List Additions and Revisions; Implementation of Controls on Advanced Technologies Consistent With Controls Implemented by International Partners
- ABreakthroughs or Boasts? Assessing Recent Chinese Lithography Advancements
- AWorld Fab Forecast
- ATSMC Announces Additional $100 Billion Investment In Arizona
- AEmerging Resilience in the Semiconductor Supply Chain
- ANVIDIA Blackwell Architecture
- AGIGAFAB® Facilities
- AFab Capacity
- AIntel, quarterly report on Form 10-Q for the quarter ended 27 June 2026
- ANavigating the Costly Economics of Chip Making
- ANVIDIA’s B200 costs around $6,400 to produce
- A2nm Technology
- AIntel 18A: See Our Biggest Process Innovation
- ASamsung Begins Chip Production Using 3nm Process Technology With GAA Architecture
- ASMIC Announces 2025 Annual Results
- ARapidus Secures 267.6 Billion Yen in Funding from Japan Government and Private Sector Companies
- ATestimony of the Semiconductor Industry Association (SIA) before the Senate Committee on Environment & Public Works
- BTSMC Overseas Fabs – A Success?
- BIs SMIC N+3’s Metal Pitch Smaller than Intel 18A’s?
- AIntel and Trump Administration Reach Historic Agreement to Accelerate American Technology and Manufacturing Leadership
- AIntel’s Transistor Technology Breakthrough Represents Biggest Change to Computer Chips in 40 Years
- AEntering the nanosheet transistor era
- ADevelopment of SiGe Indentation Process Control for Gate-All-Around FET Technology Enablement
- AApplied Materials Unveils Next-Gen Chipmaking Products to Supercharge AI Performance
- AApplied Materials Technologies Enable 2D Scaling with EUV and 3D Gate-All-Around Transistors
- AHPC Platform – Advanced Technologies
- BThe Future of the Transistor
- BClash of the Foundries: Gate All Around + Backside Power at 2nm
- AA 38.1Mb/mm2 SRAM in a 2nm-CMOS-Nanosheet Technology for High-Density and Energy-Efficient Compute
- AIntel Foundry Details Process Milestones and Future Innovation at VLSI Symposium
- ATSMC Unveils Next-Generation A14 Process at North America Technology Symposium
- ATSMC Debuts A13 Technology at 2026 North America Technology Symposium
- ACFET (complementary FET)
- ASemi-damascene metallization
- A16nm Ru lines using semi-damascene integration approach
- AMolybdenum’s Role in Ultra-Fast Computing: The Metal Behind the Speed
- ATSMC, company profile page
- AA300 FOUPs: Front opening unified pod platform with super microenvironment control
- ATestimony of David Isaacs, Semiconductor Industry Association, hearing on “Trade in Critical Supply Chains”
- AChipmakers Are Ramping Up Production to Address Semiconductor Shortage. Here’s Why that Takes Time
- ATSMC Reports Third Quarter EPS of NT$29.43
- AIntel Reports Second-Quarter 2026 Financial Results
- AAdvanced packaging and HBM, not logic dies, were the bottlenecks on AI chip production in 2025
- ABehrooz Parhami, Dependable Computing: A Multilevel Approach, part 2
- AAMD Instinct™ MI455X GPUs
- AAMD Instinct MI455X GPU
- BIntel Panther Lake Teardown
- AIntel Reports Second-Quarter 2025 Financial Results
- ABiden-Harris Administration Announces CHIPS Incentives Award with Samsung Electronics to Solidify U.S. Leadership in Leading-Edge Semiconductor Production
- APushing the Limits: Huawei’s AI Chip Tests U.S. Export Controls
- BHuawei AI CloudMatrix 384
- AIBM and Rapidus Form Strategic Partnership to Build Advanced Semiconductor Technology and Ecosystem in Japan
- ARapidus Achieves Significant Milestone at its State-of-the-Art Foundry with Prototyping of Leading-Edge 2nm GAA Transistors
- ARapidus Completes 150 Billion Yen Funding Round from Japan Government
- BChina launches $47bn chip fund to counter U.S. restrictions
- BChina starts production of home-grown immersion DUV chipmaking tools, source says
- AEmerging Resilience in the Semiconductor Supply Chain
- AEmerging Resilience in the Semiconductor Supply Chain
- AGlobal Semiconductor Fab Capacity Projected to Expand 6% in 2024 and 7% in 2025, SEMI Reports
- ATSMC January 2025 Revenue Report and the Statement on the Impact of Earthquake
- ATSMC, annual report on Form 20-F for the fiscal year ended 31 December 2025