The Bleeding Edge of Semiconductors: A Tale of Three Companies || Peter Zeihan

The semiconductor industry is one of ever-growing importance and its leaders—Intel and TSMC—are now fighting to be the first to bring the next generation of advanced chips to market.

Where to find more?

Where to find me on Social Media?

#semiconductor #intel #tsmc #chips

[masterchinese]

ASML was one of my clients years ago. I remember an engineer explaining it to me like this: "Imagine you use a laser pointer to target a coin. Now imagine that coin is on the Moon and you are hitting it from here on Earth." That's the kind of precision they work with.

[stephenandersen]

I’m part of the team building the factory in Ohio. It’s great to be a part of this

[DoubtfireClubWGPowers]

I am part of the team sitting here on the sidelines and watching all this happen by basically hearing about it on YouTube.... It's great to be a part of this

[chemist]

A couple things worth noting: it's standard practice in the industry to start developing future products 3-5 generations out. This is just how long it takes to develop a chip with tens of billions of transistors. Intel WAS resting on their laurels, but this is not evidence of that. Each generation had less improvement than was typical for the last decade or so because Intel cut R&D spending in order to look better for shareholders. They didn't start spending money on R&D again until after they were already behind and now they've been playing catch-up for the last 6 years.

Also, just for some context, high numerical aperture (pronounced like this for Peter's sake... /ăp′ər-chər/) extreme ultra violet lithography, or High NA EUV is super complicated from a technical standpoint, but for the laymen, all you really need to know is it let's more controlled light through.

If you want the slightly more technical explanation: what determines how tightly you can pack transistors on a silicon wafer (which is a pretty good indicator of performance class) is the resolution of your Laser. That laser shoots light through something called a photo mask, which blocks light in some places and not in others, and that creates the pattern on the silicon wafer that makes transistors and is how all computers are created nowadays. So every company building wafer fabrication machines is trying to get the highest resolution possible without effecting throughout of wafers per hour. In optics, resolution = lambda/2na where lambda is the wavelength of light and na is numerical aperture. So in order to make a machine with a higher resolution, you either need a higher wavelength of light or you need to let more light through your lens. building the first EUV machines took almost 30 years, the project was really started in the early 1990s and took till the late 2010s before it could be commercialized. So we're not at a point right where switching to a better laser technology is financially or even technically possible. Because of this, you can change the types of lenses we use, which is why the next generation of machines is called high NA EUV. The lenses and mirrors will have precision down to a Picometer. They will be massive. Zeiss, the optics company that will probably supply these lenses and mirrors, has had to spend billions of dollars in R&D and build special buildings and all new kinds of metrology and manufacturing equipment just in hopes of manufacturing these lenses and mirrors. In my opinion, EUV and high NA EUV are both moonshot projects greater than that of the space race. I hope this information sheds some light (pun intended) on the industry.

[FCS]

It’s so nice the way you get all of the relevant information out so quickly, with no unnecessary words. Thanks very much.

[arsenii_yavorskyi]

also worth noting that Intel also designs the chips that they make, whereas TSMC is just a manufacturer that makes chips designed by other companies.

[landotter]

Asianometry is an outstanding channel to learn about this stuff. 🤙

[rayvestrayvene]

When it's zero degrees no unnecessary words will be used. Love ur content keep it up. Thanks.

[Tore_Lund]

Funny how trekking in the mountains at winter, can make you contemplate on the past decades in the semiconductor industry?

[nathandanner]

One thing I find interesting about Semi-conductors is much of the 'new' tech has existed theoretically for decades but, hasn't been implemented. I recommend a great channel called 'Asianometry' if you want to learn way more then you thought you wanted to know about Semi-conductors.

[moremoola]

I'm Dutch. I've lived in the US for over 25 years now.
Not to toot my own proverbial horn, but engineering in the Netherlands is SERIOUS business.
Good to see Intel being in the vanguard this time.
The US needs to lead the semiconductor industry if we want to stay ahead of China.

[johnduguid]

Very insightful! Thank you Peter for helping me catching on SMH pace in a nutshell at lighting speed! As always, greatly appreciated!!

[dustydinoface]

I’m part of the team that’s not a part of this at all. It’s great to be a part of this

[daniellarson]

Kind of a nice peaceful pace where he is. As the world changes with it's ever evolving technology, it's good that there are places where nature still predominates as it has done for thousands of years. When I was a kid, they were still discussing improved vacuum tubes. About the time I finished school 741 Op Amps were the thing. I can't imagine where they will be in the next 50 or so years.

[carmenmccauley]

Hard to predict the future when it's changing every second. Thanks for this Mr. Zeihan. I never miss an episode

[casparcoaster]

Intel turns out to be GM in the chip business... sell crap cuz your name rec, not cuz its better...thats how Toyota conquered the world, and never lost its pole position. Something tells me, TSMC won't either

[mattfazio]

Love your content, thanks for all the information. But you can keep that snow up in Colorado, thanks.

[elijahmontgomery]

It would amazing if you do a Q&A for 10 to 30 minutes! I would pay for a question.

[danielblair]

The terms like 3nm and 2nm don't actually refer to particular lengths. They are processes and are essentially marketing terms.

[jamesowens]

I worked for ASML briefly in 2011 developing the EUV machines. Crazy-tight tolerances on everything to keep the beamwidth as small as possible. The reticle (which reflects the beam onto wafers) was moved around by a "stage". The stage had to be purged with Nitrogen to keep oxygen molecules from being ingested through the seals into the vacuum chamber. The reticle wasn't physically attached to the stage by mechanical means, because anything would result in warping it. Instead the stage was hyper-polished and the reticle stayed in place strictly by surface adhesion. Back then they were reliably cutting 13 nanometer lines. Obviously there's been a lot of improvements since then. Even after working there, it still boggles my mind!