Million-Qubit Quantum Computer from Intel

Xpertise
99 followers
In this video I discuss Intel Quantum Computer based on Silicon Qubits. It is the first Silicon Qubits at Scale and based on this technology Intel is building a Million-Qubit Quantum Computer
#IntelQuantum #Intel #SiliconQubits
***
WATCH NEXT:
➞ New WoW Processor explained: https://youtu.be/-NeRIrRSFs4
➞ DOJO AI Accelerator: https://youtu.be/QurtwJdb5Ew
***
MY GEAR (affiliate links):
➞ Camera Sony Alpha 7 III: https://amzn.to/3dmv2O6
➞ Lens Sony 50mm F1.8: https://amzn.to/3weJoJo
➞ Mic Sennheiser: https://amzn.to/3IKW5Ax
***
➞ Support me on Patreon: https://www.patreon.com/AnastasiInTech
➞ Subscribe for new videos every week ! ❤ And tell me what you think in the comments below!
You probably heard of recent Google and then IBM quantum supremacy. All of these quantum computers
are based on the same superconducting qubit. However, today I would like to discuss an
entirely different approach to building quantum computers. It is so-called silicon qubits,
or they also called silicon quantum dots. It is actually a paradigm shift. Just imagine
similar to those in our phones or laptops to perform quantum computing. This technology
will enable a 1 million qubit quantum computer of a small size. The one which can actually
fit in our pocket. This is so exciting. And today I will be featuring a very special guest,
Dr. Stefano Pellirano, who is one of the tech leaders of Intel Labs working on silicon qubits.
You probably know the qubit is basically the quantum equivalent of the classical bit that
you have in classical computers. But the very key difference, very, very fascinating, is that
the classical bit can only be in one of two states, either zero or one. It's like a switch.
But the qubit can be in superpositional state, which means that it can be like zero and one at
the same time. And this is one of the aspects that makes quantum computing so powerful to
address some very specific problems. And there are multiple approaches to building a qubit.
The most popular as well as the most mature right now is superconducting qubit. This approach is
pursued by companies like Google and IBM and a bunch of other startups like, for instance,
Catqubit, Catqubit, Catqubit. And depending on how many qubits they are able to entangle,
applications becoming more and more relevant. With just 50 entangled qubits, you have already
more states than any classical supercomputer. It means you've reached quantum supremacy.
When you have 300 qubits, you get already more states than atoms in the whole universe. And
let's say with one million qubits, you can do already things like cryptography and even magic.
To me, the most important question is, are we alone? And I have a feeling that quantum
computers, as they mature, are going to help us answer that question.
Now, what is wrong with a classical superconducting qubit? This type of qubit is
relatively large and it also needs a simulation and a readout interface. So it's lots of wires.
Here is a setup by Google, which was used to control just 50 qubits. This is obviously very
challenging to scale. But the problem is that to be able to actually solve a useful problem,
we really need to scale the system to a very large number of qubits, like not the tens,
not the hundreds, but actually millions. So it's very important to target the qubit technology
that can actually scale to this number. And so when you look at the silicon-based qubits,
they are very similar to transistors. And then again, the key point is that this is a
qubit technology that can leverage all the lithographic process, the advancement in
scaling and all the experience that we've gained in the semiconductor industry to push it forward.
Silicon quantum dot technology allows the highest degree of integration. And as the name suggests,
it's based on silicon, so on our traditional CMOS technology. This means you can integrate
millions of qubits on a single chip. How does it sound?
Qubits are fragile. So due to some noise or measurement, it can collapse back to the
classical state. That's why we still need cooling here, because at 20 millikelvin,
at which quantum chip operates, the thermal noise is very, very low. Actually, losing this state is
not a problem, because you can read out this state pretty fast. But the error is the problem.
So there is a lot of work and research ongoing to develop the error correction algorithm for
quantum dots. They use millions of qubits, and some of them perform the same operation. And based
on that, they will be correcting the error. Now, how quantum dot looks like? When you look at a
spin qubit, it is basically like a transistor. Now, how do you implement a qubit? You know,
you can do it in many different ways. And basically, all you need is a system that has
basically two different energy levels, each one corresponding to, say, the zero and one.
But the system has to be such that it behaves in a quantum mechanical way. And so one example of
silicon qubit is the spin qubit. So this is basically is a single electron transistor.
So you encode the qubit state in the spin of one single electron, and this electron is effectively
trapped within a single electron transistor. And so, you know, now you can imagine that
how small a single spin qubit can be is basically as small as the smallest transistor that you can
make. So if you have a great process technology, and a very advanced lithography process,
you can make a lot of them in a very small volume or area.
Now, let's say we have quantum dots built out of transistors. But how do we control this
silicon qubits? And that's true. That's the part that I love, actually.
So let me try to explain this. So, for example, the state of a spin qubit can be manipulated
using radio frequency pulses. And this has to be sent to the specific resonant frequency
of the qubit. It's like you're trying to resonate some like a musical instrument.
And so in the end, when you look at the circuit that is needed to manipulate the qubit state,
it looks like a wireless transmitter. So what you have to do is you have to generate this pulse
at a specific frequency with an arbitrary envelope or shape. And these shapes, the form of the pulse
actually represent the instruction that you want to implement on a quantum computer. So different
shapes are different instructions for your computers. For this, Intel has built a Horseridge
SoC to control these qubits. It is fabricated using Intel's FinFET 22 nanometers technology.
So Horseridge has been designed to operate at 4 Kelvin in the fridge. And why this is important?
Because now all the very high precision analog signals that are needed to, as we saw before,
to control and read the qubit now can be generated directly inside the dilution refrigerator,
you know, just like, you know, this is the actual setup of our switch behind me near the qubit chip.
And so this greatly simplify all the signals that have to go basically in and out of the fridge.
While our switch sits at 4 Kelvin in the fridge today, the qubit chip for now is actually at
colder temperature in the milli Kelvin range. So they are not currently co-packaged. And so when
you look at where we are today, still the interface between the cryogenic plate at 4 Kelvin
where our switch sits and the qubit chip that sits at milli Kelvin, it's still very challenging
because this is still based on cables. And so it's very challenging to scale. So we really need to
push integration further. And so as you mentioned, similarly to other universities and industries,
we're also been looking at this hot qubit. And so this promise to operate at higher temperature,
say in the order of one Kelvin. And so then once you achieve this, now you can think of
putting the controller chip and the qubit chip maybe on the same package. And so now your
interface between the control and the qubit is not based on cables, but is based on packaging
identity interconnect. And so this is key to scale to very large number of qubits. And it's also
very interesting for us because as you know, Intel is a very advanced packaging technologies
that can be actually leveraged for this application. Now, if you want to push things
further, so we like to dream, and this is the holy grail of the integration,
would be to integrate the qubits and electronics to control the qubits on the same die.
And this would create what you've called the quantum SoC.
And also I asked Stefano a couple more philosophical questions about the future of
quantum computing. I hope you will enjoy it. Do you think at some point of time we will come to
quantum computer with reasonable cooling, so no liquid nitrogen, or even at, let's say,
room temperature or something like 300 Kelvin? Yes, yes. Of course, you know, when people see
the way this experimental setup are, you know, it's fascinating, but of course, you know,
it's not something we can keep in your pocket, right? So it's very difficult to answer these
questions, of course. But what I like to do is I like to look at history of what the transistor
process technology and integration can actually do. So, you know, you look at back, like in the
mid-70s, and we had the Cray-1, right, the first commercial supercomputer. And that was a beast,
you know, it weighted more than five tons, and they had pre-owned refrigeration. If you
compare to what would cost today, it would be about like $30 million. So big, big
impressive engineering achievement. But then after 50 years, you look at today, we have phones in
the pockets, in our pockets, that are like already 10 times more powerful at a very small fraction
of the cost and the volume. So in the end, you know, no, I will not say that it's not possible.
And, you know, I'm looking forward to be to be part of this engineering challenge.
Silicon quantum dots seems to be one of the most prominent approaches to quantum computing.
But not only Intel is taking this approach. There is a couple of exciting startups out there
working in the same direction. For instance, there is a quantum motion UK-based startup.
These guys are also working on building a quantum computer based on electron spins. They have a
similar vision as Intel. So they want to leverage a standard seamless technology to build a scalable
quantum computer. Then there is an equal one startup. I don't know if you've heard of it yet.
It's a cool one. They're also working on CMOS based quantum computer, but this one is specifically
tailored for AI. The thing is quantum based system is inherently by its nature, looking at multiple
complementary states simultaneously. And that's the reason why it is considered to be much more
natural hardware to be used for AI in contrast to traditional digital chips for AI. Equal One
already fabricated a prototype chip where they are integrating a quantum neural network
with control electronic on a single chip. Now they target to scale it up to more qubits to run even
larger neural networks on it. What is your opinion is the bottleneck of the modern quantum
computing? Is it software? Is it hardware? Is it algos? What do you think? So it's a complicated
problem, of course. And as you probably know at Intel, we're looking at the full stack. So we're
looking starting from the bottom where the quantum device sits all the way up to the computer
architecture, the software and the algorithms. And I think this is very important to have a quantum
program that covers the full stack. However, I think that today the main bottleneck probably
comes from the hardware. So because, you know, even if we are able to produce millions of qubits of
good qubits, right, you still have the wiring and you still have to interconnect all of them. And
you still have to bring all these signals for control and readout. And this is a really big
scaling channel. And in the end, it's not even all about the number of qubits, right? I like to say
that, you know, having good qubits without actually solving the bottleneck of interconnect
would be like, you know, driving a Formula One car while stuck in traffic. So in the end, what's
our strategy? Well, first of all, as I said, we believe that by leveraging the silicon-based qubits,
we can actually benefit from our 50 years experience in the semiconductor industry, right?
And then the cryogenic electronics, with that we can address the interconnect scaling bottleneck.
And so by combining these two, we will be able to deliver quantum computers that
have millions of qubits. And so we can finally use this to solve the usable problems that
eventually will really change our lives. I think quantum computing will make for us
many surprises this century. And it will be definitely fun to watch, but we for sure have
to keep an eye on silicon quantum dots. This approach has enormous potential. In case you
want to support me creating these videos, the link to the Patreon is in the description below.
Now you may like to check out another video on my channel about radiation-hardened chips.
So basically chips which go to the space, chips beyond the earth. So basically this is one of my
favorite videos on my channel. So I will link it here. Make sure to check it out later. And yes,
if you want me to cover a particular topic on this channel, leave me a comment below.
I definitely need more ideas for future videos. Anyway, I hope you enjoyed my train of thoughts
and special thank you to Dr. Stefano Pellirano for his interesting and exciting insights on
silicon-based quantum computers. And I will see you in the next video. Ciao!
