This Is the End of the Silicon Chip, Here’s What’s Next
Articles,  Blog

This Is the End of the Silicon Chip, Here’s What’s Next

Electronics keep getting smaller, and it’s
all thanks to electric switches called transistors. When these little fellas replaced vacuum tubes,
computers went from the size of entire rooms to merely the corner of a room. In the 60-odd years since, transistors have
kept shrinking, scaling down and powering up our electronics with them. But many think we are hitting a choke point…
what if we can’t make anything smaller?! Transistors have 3 terminals, a source, a
drain, and a gate. Current flows from the source and, if the
gate allows the electrons to pass, out the drain. You know how computer code is ones and zeros
at its most fundamental level? Well, this is where that physically happens. Current flowing through an open gate represents
a one, a closed gate and no current represents a zero. Pack thousands of these transistors together
and they can do calculations and act as a computer’s brain. Early computers had thousands of transistors,
but one way to build a better brain is to squeeze more transistors onto a chip, and
today’s chips could have billions. That means scaling all the parts of a transistor
down. Not only do smaller transistors allow for
a higher density, but it also means they can switch from on to off faster, so small is
good! Right now, commercially available chips typically
have transistors with gates about 14 to 20 nm across, depending on the chip. The problem is, as gates get thinner, quantum
mechanics (the physics that govern tiny parts of atoms) start to come into play. For example, if the gates are too thin then
they won’t be able to stop electrons because the electrons will tunnel through. Not in a literal sense, they don’t bore
through the gate like an escaped convict. They tunnel in the quirky quantum sense, where
essentially electrons disappear on one side of the gate and reappear on the other. Eat your heart out, Andy Dufresne. If the gate is supposed to be closed to current
– meaning a zero – and it ain’t, that’s a big problem. Researchers predict that the lower limit for
silicon gate is 5 nm, and by 2021 it won’t be economically efficient to keep shrinking
transistors. So, where do we go when the laws of physics
stop the march of technological progress? Well, the good news is, there are other ways
of improving performance. Machine learning could help develop more efficient
algorithms to use with current transistors. Or, we could switch to light based computers
with optical gates. That, could actually boost performance 20
fold, though the hardware is a bit larger. Or, we could stop using silicon. Researchers have managed to make a transistor
out of molybdenum disulfide with a carbon nanotube gate thats just one nanometer across. It gets around electron tunneling because
electrons don’t flow as fast through the molybdenum disulfide as they do through silicon. But, mass production would probably be expensive,
it was hard enough to make the proof of concept. The point is, progress can happen even if
we can’t shrink. Keep in mind when transistors were first used
in a computer they were 20 times more expensive than vacuum tubes. Even though transistors were much costlier,
computer scientists knew they had more of an upside in the long run, and look where
that led us! …Watching cat videos on your phone while
you avoid eye contact on your commute to work. Thanks nerds! Make these 0’s and 1’s on your computer
work hard by making your very own website with Special thanks to Domain dot com for sponsoring
this episode of Seeker! Domain dot com is awesome, affordable, reliable,
and have all the tools you need to build a new website. Start sharing your ideas with the world on a professional website. can fulfill all your website needs. They offer dot com and dot net domain names,
and intuitive website builders. They have over three hundred domain extensions
to fit your needs, from dot club to dot space, to dot family or pizza! Take that first step in creating an identity
online and visit domain dot com. For more about the problems with ever shrinking
computers, watch this video on the end of Moore’s law and the whole new chip design they’re proposing. Fun Fact: The first point-contact transistor
was invented in 1947, and it was made from strips of gold foil, a plastic triangle, then
pushed down on a germanium crystal. Boy have we come a long way. Thanks for watching


  • TreeMobile.

    I wouldn't mind if optical transistors are a bit bigger if they would perform much better than a traditional silicon chip.

  • 優さん

    We need another world war, then we will have fusion energy, inter galactic space travel, cure for aids and revival from cryogenic state and all sort of other things in matter of years!

  • CSLFiero

    High quality graphene is very expensive and can not be produced in mass, only super computers will use nanotubes. We could move to germanium or indium gallide, but that it rare mineral and could be a geopolitical problem. Chip makers can retool for alternating layers of nanotube to reduce need of high quality graphene, that will buy us one extra Moore law "tock", but still leave mainstream behind. There is cpu on memory(memristor) which won't advance the shrink, but give us novel ways to add more cpus. Finally, the best route, indium gallide "alloy" to most effectively keep cost rise in check and retool to layer that in 3d for maximum length of forge service. Such a gaafet structure should bring us into the 3nm range with acceptable yields at the highest possible density. From there, use that same concept to do cpu on memory at highest possible density. That should buy us 7-10 years of time. By then global computing should enable us to leverage all the world's cpus collectively, including the several ultra powerful graphene super computers which should be operational by that time. Then cpu power will become linear in advancement and we might as well consider either biocomputer human augmentation or nuclear war. Either way, expect either humanity to evolve into superhumanity by 2035 or completely stall technologically for the next 100-5000 years.

  • johneygd

    Silicon can’t be smaller then 5nm ,but carbon can be just as small as 1nm, then we will have finally reached the end and then chips will get bigger in favor to squeeze more transistors on a chip, just image getting quatillion transistors on a microprocessor.

  • William Zhang

    People made the same prediction for the 10nm node less than a decade ago, saying that would be the end to silicons, now we are 7nm. I read that Samsung is planning 3nm Gate-All-Around FETs in 2021, so maybe 5nm isn't the end.

  • John Spotts

    Im still the guy saying technology is awesome, but it seems technology is being used against us by our own goverments and corporate masters everytime. Im starting to think the Amish might be onto something.

  • amar ek

    The real problem with shrinking is that we are already at a good convenience level and people will prefer to upgrade the hardware than directly jump into the new tech which will be very expensive. So normal people will benefit very less from this and the price will only becomes lesser slowly.

  • takeadayofff

    Next big wealth development for investors:
    It would be great to have a file system that you just cannot kill, running distributed on Blockchain via direct device multi-linking, on a fast SSD SOC.

  • Nagy Andras

    by design todays cpus are RISC cpus behind a decode unit. there is allways the path to use a shorter pipeline and process an instruction in less cycles.

  • Skip Sanzeri

    Why not mention quantum computing? Quantum computers are already working, but not in full production. IBM, Rigetti, D-Wave, Google and Microsoft are all working on quantum computing.

  • John Valley

    I don't understand the part at 1:26 when he says "if the gates are too thin they won't be able to stop electrons", could someone please explain ?

  • mangoleaf

    Quantum computing it is then we could strap it to our brains and watch cat videos while still avoiding eye contact on our way to work and everybody will just assume we're on crack

  • Some Name

    Quick FYI, the manufacturing process is 14nm (and shrinking down to 7 soon) and 12 is available. The transistors are actually much larger but we can laser etch the fins down to a width of 7ish nm or 14nm for the chips referenced in this video (First gen AMD 500 GPU's, first gen AMD Ryzen chips, etc.)

  • Milorad Menjic

    So if they make any more silicon chips you are wrong ? You are so wrong that you even can not imagine how much. If you invent 100 000 better and faster technology right now, with proof and test examples and even build factory what is your prediction for how long would rest of the world continue to use and produce and even improve THE SILICON CHIP until we come to THE END OF IT ??
    Unliked, unsubscribed, thank you.

  • Dave Maverick

    as he stated in the end that one of the best uses of technology, his example for amusement, is watching cat videos, while thanking nerds, you cant be more basic than that, so detailed info is secondary here, its mostly advertisement

  • BackFischKuchen

    A closed gate does not not represent a "zero". For a "one" a gate connected to the high voltage is open and for a "zero" the gate to the ground/low voltage is open.

  • René Pirovano

    The ultimate stuff is diamond, it has a energy gap of 5 volts, Si has a energy gap of 0,7v, and the heat conductivity is 10 times better. There is one problem, it's hard to make but not impossible.

  • dragoola69x

    Truthfully I'm surprised we haven't used the engineering and technology that goes into to transistors on Silicon Wafers to make small and or Nano batteries that can be layers upon layers to increase energy density tenfold because what is a transistor BUT a p-type anode and n-type cathode battery essentially a bry voltaic pile on a microscopic even Nano scale but what do I know I'm just a dishwasher.

  • TaidaDave

    graphene? cpu in sli? quantum? performance servers with new streaming era upcoming.. one technology ends and the gate of new possibilities is open, with qualcomm entering pc industry intel amd nvidia (in that order) will begin panicking AND this is a good thing because it results in something new

  • Nocturnal RS

    If you're the driver avoiding eye contact in your commute to work is probably a good thing, otherwise that reference seems random lol.

  • Mark Arnott

    ??Gets a story That's trending Hmm ? gathers pics off some site OH ill make a youtube Video that 20 other have done b4 me haa ????

  • Khondker Rifat Hossain

    Why do we avoid eye contact with each other?? Even a smile is a charity….we should respect and appreciate our fellow Human beings. Just think how would be if there were no one around you

  • Na Qiong Trading LLP Na Qiong Trading LLP

    Hei, U Still never Explain 96 layer Samsung 970 Evo Plus stack up like Building 3D Vnano chip…!! Yet..How it Works…??

  • Lost Space

    Well wasn't the issue the heat density got too high in Pentium 3 hence why we are all dual quad core trend? Smaller = more heat in small space = hotter than nuclear reactor.

  • Bibek Bhattarai

    How come we keep making chips smaller? Why dont we make the chips bigger? if the computer chips are bigger can you put more transitors or how the heck ya spell them

  • Rehoboth Farm

    Optical gates would be super but can we at least get optical interconnects? This would save so much power and reduce so much heat. It isn't like this stuff already exists or anything. Come on Intel get off it already. GAs substrate is super dirty. I'm not sure that is really even an option for mass production of processors.

  • NY000ooo

    Next technology: Memory Gel (this smiliar working like your brain, but it's simpler, and can mass production and this very cheap) But this only on future now 🙂

  • Abhi B

    Your added advertisement is like vacuum tube.. it occupied considerable amount of time in your whole video

  • Final Apostate

    Graphene based Computers, Used by Google.
    Not accessible by the common Peasant, And all Triple A Videogames will be though Stadia ( and other Videogame streaming Companies )
    Because of the high performance of the internet and these servers… Theres no need for " Gamers " to have Consoles or Computers.
    I didn't think i would be around for this level of Control

  • Threelly AI

    Jungle conclusions….

  • Threelly AI

    2019 anyone?

  • Wu Li

    This guy knows nothing about the subject. Spintronics are likely next, and have been used in a limited capacity since the hard drive was invented. Optical circuitry has made faster progress, but not for home use, and spintronics are not far behind and made for commercialization. They use practically no energy and work about a million times faster than our current processors.

  • Rick L.

    Proof of concepts are almost always tremendously expensive. Then some jerk comes in and says "wow your dumb. Here let me do this in half the time and cost."

  • Kittywampus Drums

    Quantum tunneling happens by negative entropy doesn't it? So just add more energy(mass) to the gates and the electrons won't tunnel passed them.

  • E-Nonymouse

    Until we acheive or exceed 1-2NM in size there is no point in trying to go beyond silicon based semi-conductors. We've been readily exceeding Moore's law since it's inception.

  • Peter Twiss

    Why didn’t you discuss other types of semiconductors like InGAs? Need infrastructure to get materials with higher electron mobility to be much cheaper and this starts from technologies to make harvesting raw materials more efficiently (cheaper chips)

  • John Savard

    Transistors have three terminals: an emitter, a base, and a collector. Now, there's this fancy special kind of transistor, a field-effect transistor, that has a source, a gate, and a drain, and it happens that all our microprocessors these days are built from those fancy special transistors, instead of ordinary (bipolar) transistors – but you can't just say that "a transistor" is a field-effect transistor! By default, they're bipolar.

  • JoeAceJR

    1:15 I think you meant 12 to 14nm. Ivy Bridge, Haswell, and Devils canyon were on 22nm and the ladder of the 3 was released in 2014. Broadwell was released in 2016 with the 14nm process, then came Kaby lake, then Coffee lake, then Coffee lake refresh which is all on 14nm. On the AMD side, Ryzen 1st gen is on 14nm, Ryzen 2nd gen is on 12nm, and the 3rd gen is on 7nm.

  • Vhsss Tv

    Samsung already announced their roadmap for 3nm for 2021… they'll be able to shrink it to 1nm before quantum tunneling is inevitable /too costly.

Leave a Reply

Your email address will not be published. Required fields are marked *