On Thu, 11/20/2025 9:36 AM, bad sector wrote:
computing about to hit full-stop?
https://www.youtube.com/watch?v=IS5FovPfvf0
It appears to be AI slop.
Some transcription errors may exist, as I used DNN to convert
the speech to text.
***************************************************************
Almost no one knows this place exists, yet every major leap in technology starts here.
Today, for the first time, I'm taking you inside this secret lab. And you're about to see things
the public is never supposed to see. The technology that will power your smartphone, laptop,
and AI data centers, 10 to 20 years from now. But before we open that door, we have to face the
harsh truth. As a chip design engineer, I'm convinced that over the last 50 years, we lived
through a miracle. We've managed to double the number of transistors on a chip roughly every
two years. And that enabled MASSIVE performance gains. To understand how colossal this
progress was, let me show you something.
First, NVIDIA GPU. The GeForce 256 released in.
1999 was manufactured using the 220 nanometers technology. Fast forward to today and NVIDIA is
ramping up the production of its Rubin GPU in 3 nanometers. So in just the last 25 years we
managed to shrunk the transistor by more than 70 times and turned graphic cards into the engine
driving the entire AI revolution. That's the miracle we lived through. For decades, engineers
have been carving atomic scale machines out of sand. It should be impossible. And yet, year
after year, the chips kept getting smaller, faster, and more powerful. And for the next generation,
we are heading into the angstrom era. We're so tiny, they barely feel real anymore. But
here is a truth no one wanna admit. This era is actually ending.
And this chart proves it. Right now we are running into the hard limits of physics. And as you
will see just in a moment, transistors are now literally just a few atoms wide. They are
becoming mechanically unstable. Heat is exploding. The alternative materials we hope would save
us aren't scaling fast enough either. If we don't invent a new kind of device, not an upgrade,
but a true breakthrough, the computing revolution just stops. And with it, stops everything
that depends on the progress in computing. Technological advancements, AI, space tech, and even
your devices. All of it will hit a performance ceiling. And right at this moment of despair,
where the world's largest tech companies like Apple, Google, Nvidia, AMD, and even Even TSMC
had no answers left.
They turned to one place?a small lab in Belgium almost no one heard of. Yet, the future of
computing is being invented right here. What happens inside these walls doesn't just advance
tech, it decides whether innovation continues or stops. Well, if you trace back any advanced
chip, from AMD or Nvidia to Apple Silicon, it doesn't lead to Silicon Valley or Taiwan but to
a small quiet town in Belgium called Leuven, home to iMac. And this is not your typical chipmaker,
it's a research hub that invents technologies that chipmakers like TSMC, Nvidia, Samsung,
Intel gonna use in 10 to 20 years from now. Here, impossible physics problems turn into working
prototypes long before they appear appear in Apple.
Apple or NVIDIA Keynote No matter where you're watching this, on your smartphone, desktop, or
even TV, all modern devices run on thin-fet transistors. A transistor is a tiny electronic switch
that controls the flow of electricity inside the circuits. You can think of it as a microscopic
ridge of silicon with a thin fin sticking up. This thin-fet device was a hero. Because it
carried us. Through an entire decade of technological progress. From the first Apple silicon
to the GPUs that power the most advanced data centers on Earth. But its time was up. To keep
shrinking transistors, the key structure inside it, the fin, had to become thinner and thinner.
Imagine something just 6 nanometers wide and 60 nanometers tall.
You'd think that it's a pretty impressive size. But in the end, they literally started bending and
snapping during the manufacturing. In fact, when the challenge becomes too complex, too expensive,
and too risky to solve on their own, the world's biggest players Intel, Samsung, ASML, and TSMC
all joined forces with iMac. Because all of them want eventually the same thing?a place to test
impossible ideas before billions trying it on their own. By then, iMac had already explored every
path they could think of, including new materials, new architectures, and new shapes. The
problem was most of them went nowhere. And the stakes couldn't be higher. Because the wrong bet
wouldn't just slow us down. It would set the entire semiconductor roadmap back by a decade.
Essentially, the industry wasn't looking for a next upgrade but for a reinvention. And if any
place can pull off a miracle, iMac. After years of failed ideas and dead ends, the answer turned
out to be very simple. And it started with a rough catch on a whiteboard. And they asked a
deceptively simple question. iMac? iMac? No, it's not a simple question. What if iMac becomes
unstable? What if we just flipped it sideways? And just imagine. This simple idea became the
second biggest turning point in microchip history?the invention of gate-all-around-transistor.
Instead of one tall fin, engineers stacked thin sheets horizontally, supported from below, making
them far more stable to build.
And the first trick solved the mechanical problems that were killing FinFET. Right now, these
devices are making their way into your next GPUs and smartphones. This innovation definitely
bought us time. And it will carry us for another three chip generations. But it wasn't the end
game. Because right now, even with this gate-all-around innovation, we're still running out of
space on a chip. iMac had to go back to the drawing board again. When cities ran out of space,
they didn't stop building?they built skyscrapers. iMac asked a bold question?why not do the same
with transistors? Well, if we cannot shrink in 2D anymore, there's only one way. Up. Stack one
device on top of another vertically. And you immediately die. Dobble the.
Transistor density and now I will walk you through step by step how this technology is being
invented. Before anyone touches a wafer, they spend years simulating this device, tweaking
geometry, materials and then try every variation. Kill the weak ones until the numbers finally
say this might actually work. And that's because every idea that dies here will save you years
of effort and expenses in the fab. For CFAT, IMEC explored multiple ways to stack transistors.
Most of them failed, but one had the pulse and that became the path forward. And when it
finally worked, it changed everything. For the first time in history, we began building chips
in three dear and so.
You're gonna see it with your own.
I could not be more excited because right now I'm taking you inside the fat see something almost
no human eyes have ever seen one of the greatest inventions in human history first I suit up i
n a clean room bunny suit and of course mine is like two sizes too big so I'm actually looking
like a walking bedsheet with a badge they blast me with air just to make sure I don't contaminate
anything inside this place is insanely delicate clean rooms must be cleaner than clean inside
they are building technologies so tiny that one single hair, one speck of dust and a million
dollar experiment is gone Once we are in we are heading straight to one of the coolest machines
scanning electron microscope this is the only.
Actually see something that small because a modern transistor is about 100 times smaller than a
virus. Here is a catch. The moment you look at it, you cannot actually see a 2 nanometer
structure or even a 10 nanometer structure because our eyes simply aren't built for that. Our
eyes only work when light bounces off the objects, but to bounce from something that small,
light is simply too big. Its wavelength is hundreds of nanometers wide, like a giant wave
trying to hit something smaller than a virus. That's why to see transistors, light, and typical
microscopes are practically useless. We need something far more powerful than that. We need
electrons. That's why we're gonna use this scanning electron microscope to see the Unseen.
First, we place the wafer inside the chamber. Inside, a thin beam of electrons scans across the
surface of the chip, and this lets us zoom in hundreds of thousands of times. At first, what you
see looks like simple lines. But as we increase magnification 20 thousand times, 50 thousand
times, 150 thousand times, patterns begin to appear. What we see right now are metal contacts
that lead to the device. They are like tiny highways carrying electrical signals into the
transistor. And here is the brutal part. When the electron beam hits the surface of the chip, it
instantly heats up. Stay on one spot for just a half a second too long, and the beam will destroy
the sample. You get only a few chances before it's gone.
This time, we had to sacrifice this one just to see what's inside. Now, Felix fires up the ion
beam. And suddenly, a tiny opening appears. In just a split second, we cut out a slice thinner
than a human hair by a thousand times. And only after making this cut, we can see the most
important part of the transistor ? the nanosheet channel. The tiny layer where the electric current
actually flows. But even this super microscope has its limits. We can actually zoom in here down
to 10 nanometers. But we still cannot see the atoms. To go further, to see the most interesting
part, we actually have to leave the fan and do something way more extreme. I'll show you in a
moment.
Now, this is where things start to get really wild because we left the fab and leveled up to
something far more powerful. Transmission electron microscope. This one can achieve magnifications
of up to 50 million times and we are about to look at the invention that took over a decade to
build and cost the industry hundreds of millions to make real. Unlike the first microscope that
shows you only the surface, this one let us see through the material almost like x-ray at the
atomic scale. This one can zoom in far enough to see actual atoms and nothing else on Earth can
do that. Getting to the atoms is a whole science on its own.
Do you remember we just cut out a tiny slice of the chip 2 nanometers thick, called lamella?
That's a slice thinner than a strand of DNA, about 13 nanometers thick. Any thicker and electrons
can't pass through it. Now you see Felix carefully placing this tiny slice on a copper grid and
load it into the microscope.
I cut it short, but this preparation took us at least half an hour. And now, finally, we are
zooming into the seafoam transistor. Now on the screen, you can see the bright shape. These are
the metals, the contacts, and the gate. The gray regions are silicon and silicon germanium.
And running through the image, almost like a thin thread, is the nano-sheet channel. Any path
where electric current flows, can be carried out by the cell like a tube. And what does this
mean? It means that the energy is passing through the cell. Right. The current flows when the
transistor switches on. So this is the DPP type channel at the top, and here at the bottom, you
have the N-type channel. And it's a rocket science to manufacture this channel so tiny.
So those are atoms, you can see them here.
What you see right now is a glimpse into one of the humanity's most advanced, tiniest creations ?
the seafoam. For the first time in history, two transistors stand on top of each other, taking
more power into less space and opening the door to a new era of computing.
And now, if we zoom in...
You can see the single atoms of the channel. And this view is as close to the heart of the chip
as you can get. And this channel is about 30 atoms thick. When I saw this, I just froze. Because
this was the physical limit of reality. Right in front of me. There's a catch, though. Because
to look at it, we literally shoot electrons at the device, and every second we rotate the person.
And every second we risk to destroy it. Fun, right? But also terrifying. Every image feels like
a one-time chance. Like a shooting star. You're admiring it, knowing that it's gonna disappear.
Why does this process ? we went just through, SAM, NTM, which is so expensive and requires a lot
of training to do ? matter so much? In fact, this isn't just for cool visuals.
These are two critical steps in inventing every new chip technology.
Because modern chipmaking involves building these tiny structures layer by layer, which involves
thousands of steps. And at each stage, engineers need to check if the structures are still perfect
or if something went wrong. And if something went wrong, they need to figure out why and fix the
process. What you are seeing right now actually decides everything. It decides the fate of this
innovation, if it moves forward or if it ends right here. Most people think new chip technology
just appears every few years, as if someone wakes up and says, this is the next transistor. But
the reality is very different. In fact, from the moment a new transistor is imagined till the
moment it will end up in your phone, you are looking at 18 to 20 years.
And so far, we have explored just a tiny part of this innovation process. But I'm sure you already
can see why this is one of the hardest engineering challenges on the planet. And I was lucky to
talk to Serge Bissamans, who led the invention of CIFAD technology at IMEX. He is basically
leading the innovation that NVIDIA, TSMC, Intel and Apple will depend on in a decade from now.
And even when you come up with a perfect design, if the tools physically can't build it, the
idea dies. Because in this industry, a new device isn't just a design challenge. It's an
equipment challenge. In some sense, some people say Moore's law isn't capability law.
If the tool does not exist to print smaller features, if the tool does not exist to remove certain
materials selective to others, you would not be able to physically build the complicated structure
on your safer. And CIFAD invention pushed tools harder than anything before it. In fact, to stack
two transistors on top of each other, they needed an entire new generation of tools. You have to
be able to precisely etch and build tall and narrow structures without letting them collapse.
Grow and remove materials layer by layer without damaging anything underneath it. And you have to
be able to deliver power from the backside of the wafer, something that has never been done before.
That required new etching and deposition systems, new epitaxy techniques, new backside power process.
And on top of that, new tools capable of.
Measuring details that once were literally invisible. That's why iMac works closely with tool makers
like ASML, the key edge deposition and EPI suppliers, and key material suppliers to co-develop
equipment and materials required to turn these ideas into reality. And these tools aren't cheap.
The EUV scanner alone, the queen of all chip making equipment, now costs over 250 million dollars
per unit. Most of the others range from 10 to 50 million each. And to run even an experimental Fab
line, you need about 200 of them. And once the early version of these tools exist, they don't go
directly to JSMC or Intel, they first land here, at iMac. That's where researchers test them
directly on silicone.
Long before any chip maker is even allowed to touch them. Right now, here, at the iMac Fab, they're
assembling the new high NA UV lithography machine from ASML. The most advanced lithography machine
ever made. They didn't let us film that part, of course, but trust me, this single machine will
fill an entire call. If the results look promising. Only then I'm sure. iMac moves the work to its
pilot line. Its test kitchen where manufacturing flow is perfected. Because to make C-FAT real,
iMac isn't just developing the new device, they're reinventing the entire recipe. A recipe where
new tools, new materials, and new techniques all work together in a perfect sync. And while the
journey to make C-FAT manufacturable is still underway.
It's already working on what is next after CFAT. Because in this industry, if you wait until the
current technology runs out of steam, you are already too late. Now here is the part no one likes
to admit. The future is still unknown. CFAT buys us a decade, but after that, for the first time
in the chip history, there is no clear roadmap after CFAT. No one even iMac doesn't know what is
coming next. The only thing is certain. CFAT cannot lower power or heat enough for the future AI
demands. Here we would need something more to break the loss that limits the silicon itself. Now
what comes after CFAT, the only thing that I can think of is that we start stacking even more
layers. The reason that we stack is because we can no longer. because we can no longer scale
X and Y.
I don't think that there is any idea right now out there that can easily scale X, Y. So that means
we have to make use of the vertical dimension. And CFET is just the first generation of learning
how to do so. If we master that by the mid-30s, next decade, I believe that opens up room for not
only hybrid channels, but also hybrid technology. Well, that's interesting. And we have to talk
about this. Because if you look at the history of computing, first, the progress came from one just
one thing. Shrinking transistors. But it's not the case if you look at what's happening right
now. Around the 2020, the physics of making transistors smaller and smaller started to flatten.
Costs exploded. Heat exploded. And physics stops cooperating. This is the wall we spoke about.
CFET will extend this line a little, little bit further.
It buys us time. It might carry us towards 2030, but it doesn't restore the exponential growth.
This era is gone. Right now, the industry is forced to change the very definition of Moore's law.
Because if you cannot make transistors significantly better, you have to make the whole system
significantly better. In fact, in the modern chips, performance doesn't come just from smaller
transistors. It comes from how fast many chiplets talk to each other. Now, instead of building
one giant chip, companies combine multiple smaller chips side by side or even vertically to
unlock far more power in a single package. And that approach changed everything. Right now,
chiplets on the chip sit centimeters apart. And at this distance, communication.
Starts from the tip of their tongue. This is like shouting across a football stadium. It's wastes
energy, and it slows everything down. A promising solution is to let these chiplets talk using
light instead of electricity. This is silicon photonics. Today, this technology is used to
connect GPU racks across data centers. But soon, it will move inside the chip package itself.
Letting chiplets to communicate with light at incredible speeds. Right now, silicon photonics
still consumes way too much power to be used everywhere, especially inside laptops and
smartphones. The lasers run hot, waste energy, and the whole system is extremely sensitive to
the temperature. You often need heaters just to keep the light stable, and this sort of defeats
the purpose of saving energy. This is where iMac steps in. They are not trying to prove that
silicon photonic works.
This is a done part. But they are trying to make it efficient enough so it can be used everywhere.
Right now, they experiment with advanced optical materials you can't use in the normal fab, like
barium titanide. For your audience, which I am one of them, by the way, this particular one is
too detailed. But it fits the material story. So, photonics exist. The good thing is it can be
fabricated in any production line. Photonics is still a little bit power hungry. But there are
materials that are optically very, very interesting but that are not fab compatible. We can at
least deposit those materials on 300 millimeter. We can pattern them. We can do the optical
characterization in a realistic... So, those are the type of research. So, new materials to
scale the power budget of the photonic circuit. That's kind of, I would say, the tagline of what
we are trying to do here. If IMEC is able to pull this off.
It's the whole new leap in computing. Just imagine, this kind of technology that today only
NVIDIA can afford to use in their giant GPU clusters could one day run inside your laptop.
Or your smartphone. iMac Vision goes far beyond that. Right now, they're working on stacking
entire wafers on top of each other. Imagine taking two switches. And perfectly stacking one
on top of the other, with every street aligned. And if this works, we could eliminate most
of the communication wiring between chips. And this means lower power consumption, higher
speed, and most importantly, the ability to stack different types of chips and different types
of materials on top of each other, not just silicon. And this leads us to a future where chips
won't be flat anymore. anymore. They will look like it.
It's like 3D computing. Cube. There is still one more curve ahead. One that might completely
change what a processor even means. To eventually use light instead of electrons for computation.
Here is the interesting part. iMac isn't betting just on one future. Right now, they have
multiple teams racing in parallel, each exploring a completely different path. To what might
come after silicon. And no one knows which one, if any, will work. Right now, they are testing
ternary logic, reversible logic, spintronics, cryogenic CMOS. And you are lucky, because
all of these technologies are already covered on this channel. But eventually, we might need
an entirely new device. One that follows different laws of physics.
Siliconabase. I'm kinda proud to say that I probably touched upon all materials in the Meninov
table, except the ones that radiate at night, the, um, the, the, the radiative materials. But
other than that, we've probably touched upon most of them, if they are available in a
300-millimeter clean room environment. Right now, they are experimenting with germanium,
graphene, 2D crystals like transition metal dichalcogenides. They're used to create another
thin layer of passes. Even carbon nanotubes, materials just a few atoms thick. And they all
look amazing under the microscope. But the challenge is turning them from tiny lab samples into
the real technology, ready for mass production. The future is not decided. IMEC, in
collaboration with its partners, saved Moore's law not once, not even twice. And right now,
they are racing against limitation. against limits of.
Time, energy, and physics itself. The future of computing might be already here, in one of
the experimental wafers in Leuven. Even IOMAC doesn't know if it will work. What makes this
unique IOMAC model to work isn't just engineers and tools and partners, but neutrality.
IOMAC is non-profit. It does not compete with anyone. And their secret edge is that they have
an entire semiconductor factory on site, built just for the research. And now they're
expansioning it to more and more buildings to run more experiments. Most universities dream
about one clean room. IOMAC has an entire production line. The mix of brains, machines,
and neutrality makes IOMAC something rare. Not just a lab, but the world's.
World-famous product. They want to create the next generation of chips. So now you can
literally forget Belgian chocolate or waffles, because IOMAC wafers is what the entire
tech industry is really addicted to. Now, if you enjoyed this episode, share it with your
friends and colleagues, and watch this video where I explain what it takes to build a
semiconductor factory and to manufacture these devices. You will love it. Thank you,
and I will see you there.
***************************************************************
I'm sure when I post this, it'll set off the sporge detector on ES.
It's a good thing I got dsnote working, saved me watching the vid.
Paul
--- PyGate Linux v1.5.1
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)