Apple chip maker TSMC plans 5nm chips for 2020, and 3nm in 2022

“Apple’s chip manufacturer TSMC today broke ground on its first 5-nanometer fabrication facility in Taiwan, promising that 5-nanometer chips will be commercially available in 2020, with 3-nanometer chips planned for 2022,” Jeremy Horwitz reports for VentureBeat.

“The tiny new processors will guarantee that future smartphones continue to shrink while offering superior performance and battery life to today’s models,” Horwitz reports. “First shown in physical form by IBM and Samsung last June, the 5-nanometer chip process is capable of squeezing 30 billion transistors — digital on-off switches — into fingernail-sized chips, doubling or tripling the transistor counts of 10-nanometer chips.”

“A 5-nanometer chip might be four times as power-efficient as a 10-nanometer chip while offering the same performance, or four times as fast with the same battery life. Vendors commonly choose something in the middle, such as twice the speed of a prior chip with half the battery drain,” Horwitz reports. “[TSMC] is investing over $17 billion in the factory and $24 billion in 5-nanometer technology as a whole, with plans to fabricate 3-nanometer chips in the same location.”

Read more in the full article here.

MacDailyNews Take: 5nm? 3nm?! Holy shrinkage, Batman!

Apple’s future A-Series chips are going to be technological wonders to behold, that much is sure.

5nm by 2020 and 2nm by 2022 – hey, we’ll drink to that! Interns, TTK. TGIF and cheers, everybody!


  1. Mind blown. I never imagined working transistors at the 3nm level. I just assumed electron tunneling would hinder the use of transistors at that size. Wonders never cease. 30B transistors on a chip!

    Bell Labs announced the first transistor in June 1948, two months before I was born and I’ll have a good chance of being alive to see a count of 30B transistors on a single chip. 70 years from a one transistor component to a 30B transistor component. Simply astounding.

    The human brain is said to have 100B neurons so in theory just four 30B processors in a handheld device could easily exceed that number. Along with advanced AI development that could yield some awesome processing power in the palm of your hand.

      1. I’m puzzled. What is the relevance of some very basic equations to quantum mechanics? You could at least have put some effort in and doug up equations from quantum mechanics …

            1. Just some exerpt from an article in Forbes..
              Put’s things in perspective…. just mind boggling!

              “How big is is one billionth of a meter? It is one millionth of a millimeter, and a millimeter is already very small – about the diameter of the lead in a wooden pencil.

              It helps to put these numbers into perspective. A human hair is around 75 microns (abbreviated 75μm) or 75,000nm (nanometers) in diameter. The relationship between a nanometer and that hair is similar to the relationship between one mile and an inch – one mile is 63,360 inches. A human red blood cell is 6,000-8,000nm across, and the Ebola virus is about 1,500nm long and 50nm wide.

              Semiconductor manufacturers started to use the “nanometer” term around the beginning of this millennium when chip gate lengths had already been shrunk well below a micron. The gate length is the basic measure of how small a chip can be manufactured. One micron is one millionth of a meter, or one thousandth of a millimeter.

              Today’s most advanced NAND memory chips are shipping with transistors whose gate lengths are shorter than 20nm, or less than half the width of the Ebola virus.

              The picture above shows individual silicon atoms overlaid with a 10nm-long arrow. There are about 20 atoms along this line, so one nanometer is about the width of 2 silicon atoms, and the gate length of a 20nm NAND flash chip would be 40 atoms across.”

    1. I remember when they were saying that was going to be a serious limiting problem at around 20nm, certainly Intel chips were suffering at around that density. How have they solved the problem.

      1. When they say 3nm it doesn’t mean everything is that small. Somewhere a small feature is that small. Nevertheless, mind blowing. 3nm is on the order of 10 atoms. One thing they did was develop transistors that feature vertical sheets rather than thin traces. Being thin they can pack many on a chip. Since it is a sheet there are more atoms available which reduces the quantum problems. Truly amazing.

        30B vacuum tubes running one watt each would consume 30GW of power, roughly 30 nuclear power plants worth.

    2. This is remarkable! If it is real. Or maybe TSMC public relations department is full of baloney.

      The author writes paragraphs about the great things that happen when chips get smaller (which is not news). But what is the primary source of this news? And is it credible? Anybody?

      1. he wrote in the article that they broke ground on the production facility. Companies don’t invest 17b in a factory if they wont be able to create the product it is designed to build. jus sayin

    3. Progress is an amazing thing.

      My uncle had a research fellowship at Bell Labs after WWII. Bill Shockley called him into his office one day and announced, “There is the future.” Every transistor that existed on the planet was sitting in an ashtray on Shockley’s desk. Indeed that was the future.

      I have a video from the late 1960s of my uncle conducting a tour of the building that housed his company’s computers (IBM System 360s). He proudly notes the special air-conditioning systems and wiring necessary to support all the data processing for a Fortune 500 company. He also mentions the online storage capacity and computing power of the installation… somewhat less than I have on the iPhone in my pocket.

    4. But each neuron has multiple interconnections. I am not convinced that digital logic can fully emulate human/biological thought processes, even at comparable complexity. It is superior in some ways and inferior in others.

      1. Nick, you carefully & quietly “nicked” the whole “I need a new computer meme.”

        Right now for average every day document work I use a 2009 MBP Core2Duo. Why?

        Well, I can put two documents side by side on its 17 inch screen and compare/edit both. If there were and appreciable slowdown, I would not do it.

        I have a new laptop for 3D solids work, but it obviously has chip flaws and I’m not risking fix failures, so it doesn’t go online.

        New is just “image” for the most part, because next year or after 2 years a MBPro is just, well, “old,” but hardly out of date!

  2. And Apple planned to release HomePod 2017. Maybe TSMC can perform better than Apple. Wouldn’t that be swell? Congratulations on future endeavors TSMC, I’m sure Apple appreciates all that you do for them. Show ’em how it’s done, fellas, Apple could use a little leadership.

Reader Feedback

This site uses Akismet to reduce spam. Learn how your comment data is processed.