Quantum Computing and the Paradigm Shift

0:00 so good morning in the good evening uh juvenile tour welcome to this new session of

0:06 distinguished speaker uh of the j talk series my name is dominic a member of our technical

0:14 community and i am extremely pleased to present rob hayes who is the ceo of atom

0:22 compute just to introduce briefly who is wrong

0:27 is recognizing this industry leader and the founding of the

0:34 board member of internet alliance and is an executive sponsor or

0:39 for creation of a computer excellence lean consortium and last but not least is serving in a

0:45 steering committee of oecd ai compute task force in

0:52 in his past tenure rob has held important and senior position in both

1:00 lenovo and intel where he met our cto

1:07 raj and he'll probably chat a little bit how they come together he has some interesting tips on this one but

1:14 uh introducing the title of the presentation uh introduction to quantum computing uh it brings

1:21 some memories back to my youth when i was studying uh

1:27 computing and computer physics and effectively it was or it's still about

1:33 electron and holes but rob will will help us to this new

1:39 journey of quantum technology where we will discover uh a more uh

1:45 interesting and awesome word that will come from a new particle a new methodology

1:52 on how they effectively are transforming uh this quantum technology our world and specifically what is the huge impact

1:59 that this technology are airing on the computing which is exactly the quantum computer world

2:06 and so uh with the any further ado i would just uh uh

2:13 welcome rob on the stage and please uh to you

2:18 thank you so much domenico pleased to be here and thanks raj for the invitation to join you guys today

2:24 i actually i met raj it's probably been maybe 15 years or so when i was at intel and he was a fairly

2:30 senior executive as an intel fellow and i was a individual contributor product manager at the time in our ethernet

2:36 division and i was working on some remote management and security technologies that raj had a particular

2:42 interest in so we we became friends and he became somewhat of an informal mentor to me at the time and

2:47 i later i later moved on in my career to be a vice president intel and lead the xeon processor product line

2:53 um as well as other ethernet products and high performance compute fabrics as an executive there and then i went to

2:59 lenovo as the chief strategy officer and all through the way raj and i have kept in touch and found

3:05 ways for our companies to collaborate wherever he's been or i've been and so um pleased to get to speak to the

3:11 juniper team here and um i do have a good mix of background in both computing

3:16 technologies as server system vendor as well as cpu provider but also in networking

3:23 ethernet products ieee standards and other industry consortius so i've kind of been at the crossroads of compute and

3:30 networking through my career and and have enjoyed enjoyed all of it so right now i'm at adam computing i joined

3:37 adam computing which is a startup quantum computing hardware provider last year in july

3:43 as the ceo i actually got engaged with the company about a year and a half ago as a board member and an advisor to the founder and

3:50 then ceo and uh and really became enamored with

3:55 the both the technology and the market opportunity for quantum computing as well as the technology that atom computing has created so i'm gonna i'm

4:02 gonna talk to you today a little bit about um you know what is quantum computing what

4:07 are some of the different technologies that different companies and academic researchers and so forth are pursuing to

4:12 bring quantum computing to life i'll tell you a little bit more in depth about what we're doing at atom computing

4:18 and why why we think it's distinct and quite scalable and then we'll talk about what people are going to do with quantum

4:24 computers and how it might impact juniper and everyone on the call here and we'll open up for questions so

4:30 that's the agenda for today um and i'll get right into it so adam

4:35 computing is based in berkeley california uh it was founded in 2018 by ben bloom who is standing on my left to

4:42 the right in the photo i'm in the center and jonathan king who is

4:48 kind of over my right shoulder and over one and so they both have phds in physics

4:54 and chemistry respectively and are driving our quantum engineering teams as

4:59 well as our applications teams respectively uh we're kind of known for being the fastest to deliver 100 qubits the qubit

5:05 as we'll talk more about is the fundamental building block of a quantum computer um and the number of qubits

5:11 kind of dictates the size of the machine or the scale of the machine and when we announced our 100 qubit system which was completed last year it

5:19 was tied with google's for the world's largest gate-based quantum computer and we did it in only two years with 20

5:24 million dollars raised at the time so much faster and much cheaper than anyone else had done and that's kind of a

5:30 testament to the technology that we chose additionally our qubits are very high quality

5:35 we've published technical papers that that demonstrate world record coherence times

5:40 which is we'll talk more about that but that's how long a qubit holds the quantum state and so that's very important for the performance of the

5:46 computer um

5:52 you know quantum computing is one of these kind of once in a 50-year paradigm shift in computing performance uh it is

5:58 going to touch many applications and many industry verticals we'll talk more about how it's different

6:03 than classical computing but at the end of the day it won't replace classical computing it'll supplement it in the

6:09 same way like gpus and fpgas and accelerators accelerate workloads off of

6:14 standard cpus quantum computers will offload certain portions of workloads that are calculations that are that are

6:22 amenable to quantum subsystems and and really accelerate the

6:27 performance of many applications we'll talk more about that later

6:34 first of all what is quantum computing right so classical computing as everyone is aware

6:39 is uh we have bits and bits can hold uh one state at any time and it's either on or off zero or one or one or zero right

6:46 so um each bit holds a value of zero one in a quantum system we have these things

6:52 called qubits which is just short for quantum bits and they actually don't hold a zero or a one they hold a very cons they have a

6:58 complex coefficient value it's represented as a vector in three-dimensional space uh

7:05 often depicted as a sphere is like what you see in the in the diagram here but each

7:10 qubit holds uh with that one complex number which has a lot more information than just a one or a zero on or off

7:18 and when you string multiple bits or multiple qubits together this is where the power of quantum computing really

7:24 shines so in a classical system again each bit holds a zero one and you can string you know as many of them together

7:30 as you want in your system uh and you can hold one value and that value can represent a a number from zero

7:38 to two to the n minus one right so you put these bits together you get one one larger number as you add

7:43 more bits to it in a quantum system remember we already have complex numbers held in each

7:49 quantum bit and we have a property called entanglement and when you get multiple qubits together

7:55 they're able to what's called entangle and that system can now hold two to the n values at the same time so instead of

8:02 one value from zero to two to the n minus one you can hold two to the n values which means that these cube the

8:08 performance and the power of these qubits grows exponentially with the system which is where quantum computing gets its power from

8:16 to give you one example this is a google example a fairly famous paper they put out about a year

8:21 or so ago where they claimed quantum advantage on a given workload it was a

8:26 workload that is one that's well known for in quantum computing to be something that quantum computers can do

8:32 much faster than classical computers could do and they demonstrated on their

8:37 system which is a 100 qubit system that they can do the calculation in four minutes uh and that calculation would be

8:44 estimated to take about ten thousand years to perform on a classical computer so that's an estimate i don't know if it's

8:50 exactly right or not but you can see that we're talking orders and orders of magnitude different

8:55 performance and so i think that's really kind of a good example of why people are so interested in quantum

9:01 computers is because there's certain calculations that just aren't practical or cost effective to do on a classical

9:08 system so people just won't try them but they can be done on quantum systems um and this is like modeling natural

9:15 systems like fluid dynamics or chemistry molecules

9:21 energy and things like that are are things that people are very interested in modeling on quantum systems

9:29 another kind of comparison this is the summit supercomputer at oak ridge national lab

9:34 this is right now it's the second most powerful quantum computer in the world uh it it as you guys probably all know

9:41 uh super computers are very large this one is the size of two tennis courts they're often the size of a you know a full warehouse

9:47 um this one consumes 10 000 power of 10 000 homes 185 miles of cables 200 million costs uh for for an oak ridge to

9:55 buy this from ibm uh and it the performance you're getting out of this is the equivalent

10:01 performance of what you can get out of 50 to 60 logical qubits a logical qubit is an error-corrected

10:07 qubit um we'll talk a little bit more about that as well but um you can see that you know massive

10:13 performance out of relatively small systems is the promise of quantum computers but here's the problem

10:19 in order to get that power we need to build large-scale quantum computers meaning more and more qubits and we need to do that with error correction

10:25 and error correction requires more physical qubits than what's called a logical qubit because

10:31 you need to sort of over subscribe the system because these qubits are prone to noise and you have a certain error rate

10:38 and in order to kind of mathematically cancel out that error rate uh you need to kind of over subscribe the number of

10:43 physical cubits to logical qubits and it turns out this isn't like a a a ten to five ratio or ten to four ratio

10:50 or something like that it's like hundreds to one a kind of ratio to to really be able to cancel out all the air

10:55 corrections for very large circuits for all the errors and so there's multiple factors that go

11:02 into uh the performance of a quantum computer it's not just the size of the qubit or sorry the number of qubits in

11:08 the system but it's the error rate like i just described so people talk about fidelities this is a very common thing

11:14 for people to report as they say have 99 fidelity or 99.5 percent fidelity that's simply one over error rate

11:21 um so if you have one error out of every 100 you know uh calculations performed on a qubit

11:27 that's a 99 fidelity um and if you think about like a circuit depth i'm going to run a series of

11:33 calculations in order if you you know ran 10 calculations in order and you've got one error at every 100 you would be

11:40 99.9 to the 10 or 99.99 to the 10 and that would actually be a pretty high

11:46 probability of an error so that's why we want to have error correction in there um and a low error rate to begin with

11:52 also coherence so uh these quantum systems uh recart depending on what um

11:58 kind of a technology you're using are often kind of prone to falling out of what's called coherence which means they

12:04 lose their quantum state fairly quickly um superconductors which are the initial kind of quantum computer are

12:11 have a very short coherence time that's one of the downsides of them they only can hold the information for microseconds and then it's lost and so

12:18 if you lose the information while performing a calculation then you have to start over and it turns out with error correction

12:25 there's a feedback loop going on in the system where you're trying to detect and correct errors and you need to hold that information for a long time so short

12:31 coherence means difficult error correction long coherence means lots of time to form deep circuits and lots of

12:37 time to perform error correction so um so that's a very important metric

12:43 there's something people talk about is connectivity connectivity is um qubits get entangled i'll tell you i'll have a

12:48 little demo of that in a few minutes here but basically you can have two qubits and they can interact with each other and

12:55 share information during the calculation and um and there's a there's only

13:00 certain qubits in a system that can connect with each other depending on the topology of the system so in our system for example

13:07 only nearest neighbor qubits in in space the ones that are literally like next door to the one that you're using can

13:13 connect to each other other systems you can connect to qubits within a chip but not across multiple chips and things

13:19 like that so all of that has some impact on performance uh error correction you know do you have

13:25 the ability to detect and correct errors can you do mid-circuit measurement where you can actually detect and correct as you're executing through a circuit or

13:31 you can only do it at the end that will affect performance and then finally gate speed or the clock cycle

13:36 and some um some gates or some cubits take longer to perform calculations than others and so that'll all that adds up

13:42 to a multi-dimensional space kind of problem and trade-off space for for performance in the system

13:50 i'm not going to talk about benchmarks today but there's things called quantum volume and other benchmarks that try to

13:55 take all of these things into account and give some kind of numerical representation of the overall performance of a system

14:02 um but right now what i was going to talk a little bit about some of the different technologies or what we call modalities in the industry for um for

14:08 quantum computers the first two uh quantum computing type systems uh i should say i'm only talking

14:15 about gate based systems here uh there's a type of system called an annealer or a simulator um d-wave is the most popular

14:22 or you know kind of oldest company that's been building these this is kind of a different form of quantum computing

14:29 um it's it's useful but it's only useful for a subset of applications and it's not really the type of systems that most

14:35 people are are looking for for the broad set of applications so within the gate-based quantum computers which is the the kind of more

14:42 broadly applicable type system superconductors were the first technology to build um to build qubits

14:48 for these these quantum computers google ibm and righty are the three leaders in this space basically what

14:54 they do is they build chips using semiconductor manufacturing technology these chips hold qubits in there they

15:00 can basically get the electrons into various states on these chips that represent

15:06 the quantum states within a qubit and they have to wire each chip up together and so that they're all

15:12 connected so they can they can uh they can all operate as one system and so you get these these beautiful photos like

15:17 you see on the left here of what ibm calls the golden chandelier um and that's just simply because you've

15:23 got all these chips and these chips have these gold wires connecting them so that they can they can interact with each other

15:28 um the the they've been doing this for about a decade or more um the qubits are relatively high quality they've got the

15:35 noise out they've got the manufacturing kind of dialed in the problem is scaling with this um

15:41 every time you want to add a qubit you have to add another wire so this is a picture of of a system with a few cube few dozen

15:48 qubits on the left you can imagine if you got like 10 000 or a million qubits you'd have a million wires literally

15:54 each of them sending an rf tone down to the qubit um and i don't know how you would you would wire up a million wires you know that

16:01 supercomputer we saw on the other page was probably uh i don't even know 20 or 30 000 nodes probably so that's you know

16:08 185 miles of cable was probably only going to 20 or 30 000 things not a million things so

16:14 so you can imagine the complexity um more recently companies like ion q which went public last year in continuum

16:20 which was a recent merger between honeywell quantum systems and um and cambridge quantum computing out of

16:26 the uk they've been manufacturing these things called trapped ions it's a different type of qubit technology they also they

16:33 make chips but these chips have little um like channels in them uh or voids and

16:39 then they create magnetic fields in the voids and then they load ions which are just charged particles into those voids

16:45 and they get kind of suspended in air these systems also have to be wired up um and connected together they both of

16:52 these types of systems have to be cryogenically refrigerated and dilution refrigerators and as they get bigger

16:58 more wires larger refrigerators bigger and bigger engineering problems so so while both of these modalities

17:05 have very high quality qubits and they've been around for a while they're both very challenged to scale

17:11 and that kind of leads to what we're doing and what many others in the industry are doing so

17:17 there's a number of companies and academic you know institutions looking for a more scalable

17:22 approach so that we can create many more qubits because we ultimately need millions of qubits right in order to to have a few

17:30 thousand logical qubits and and perform very large calculations and so these are kind of the these aren't all

17:37 of them there's a couple others that are kind of more in the research phase but these are the the seven kind of

17:43 um you know probably most prominent alternatives to superconductors and trapped ions

17:49 uh we're pursuing neutral atoms sometimes they're called cold atoms i'll tell you how it works but you know the 30 second version is we basically

17:56 trap atoms in a vacuum chamber with lasers and they get stuck in free space to that laser so they're not moving at

18:03 all and and if they're not moving they're cold by definition so that's that's why they're also called cold

18:08 atoms and we can manipulate the quantum states inside the atoms nucleus using other laser tones

18:15 um photonics is probably the second most promising of these scalable ones or

18:21 furthest along there's a very quiet startup called sciquantum that is pursuing this and

18:28 xanadu out of i think they're in canada is a little bit more vocal but these are two companies and they're basically

18:34 uh using silicon photonics to get photons to act as qubits

18:39 there's a number of others here you can still it's quantum dots that intel's pursuing silicon spin um and others but

18:45 these are all you know they're all different approaches all different technologies different ways of manufacturing systems but all going

18:50 after more scale because they because people generally see that superconductors and trapped ions will

18:56 will hit some kind of a scaling wall um you know in the not too distant future

19:04 um so to dive in a little bit more into the neutral atoms and how that works and what we're doing at atom computing um so

19:11 neutral atoms are simply um atoms in the second column of the periodic table they're

19:16 alkaline earth metals they have what's called a closed shell they're also called neutral because it

19:21 basically means they have the same amount of protons and electrons which basically means that they don't interact

19:27 with anything in the environment they're just very stable um and that stability is what gives them

19:32 the long coherence times we're able to use the nucleus of the of that atom to create spin uh remember

19:39 if i showed you the represent the uh the coefficient inside a qubit with a sphere well we can actually literally rotate

19:45 the the um the nucleus of the atom to to be in wherever we want in space um i'll show you a demo how we trap them

19:52 with optical tweezers but what's really uh beneficial about this is that these these atoms can be very

19:57 close together we we in our system keep them about four microns apart um and uh we we trap them in a grid and

20:05 you can actually see looks like my little gif animation's not working on this one but usually that little

20:12 diagram in the bottom right is kind of moving and you can see the atoms turning on and off as we're running the system but

20:18 basically we trap these atoms in a ray they're four microns apart our current system is 100 cubits so it's 10 by 10.

20:25 um and so that whole um all hundred qubits fit in 36 microns on

20:30 the side right and when we get to a million um qubits we'll be in a 3d array

20:35 100 by 100 by 100 that's a million four microns apart that's still less than half a millimeter on a side to get all

20:41 million qubits so very very small very different than that golden chandelier that you saw from from ibm and google

20:48 um and all of this is controlled wirelessly because we're doing it with lasers and so instead of having to have a wire that

20:54 goes to each atom we just have a point of light that goes to each atom and that's how we control them

20:59 so i'm going to show you kind of like a cartoon like demo to try to make it uh maybe a little bit more clear what we're doing and then we'll talk a little bit

21:05 more about like what what are you going to do with these quantum computers so

21:11 we basically have a system where this isn't working very well

21:18 but my animation might not be working on uh okay well i'll just talk through it and the animation may not work so we have

21:25 atoms the atoms are basically floating around in a vacuum chamber there's nothing in there except for what we put into it which is strontium

21:32 in a gaseous form and what we can shine a specific color of light

21:37 on on the atom and it and it gets stuck in the focal point of that light

21:43 and it just freezes in space it freezes to like four micro kelvin it's just so still it doesn't move

21:48 um shoot the animation's just not working so let me

21:54 um so we shine that beam of light we capture the atoms

22:00 i'm just going to try to get through this um we but we don't just capture one we capture many so we capture it in a grid

22:06 of light so we just have multiple points of light in a grid the atoms are floating around we capture them

22:12 as they cross the beams of light if if we miss one um we can rearrange the atoms we can just steer the beams of

22:18 light over if we want so what we actually have is a system that's a larger array than 10 by 10

22:24 and we capture some you know something less than 100 or 100

22:29 of the atoms in there and then we just rearrange it into a tight fully populated grid uh to get entanglement

22:35 um and then within uh it within the atom we use the nucleus itself to create the

22:41 qubit and the spin apologize the animation's not working um we can create entanglement with

22:48 what's called rydberg states on these atoms so normally the atoms my fingers are kind of representing the electron

22:53 cloud and the orbit around the nucleus we can actually with different colors of light excite the atom so that the

22:59 electrons a cloud actually goes to a different orbital radius

23:06 and that can be calculated and measured so that's uh it's something that we can control and when they get into what's

23:11 called the rydberg state the the electron cloud is very large and so these atoms that normally are quite far apart from each other four microns which

23:18 is very far compared to the size of the atoms themselves um when they get into these rigberg states

23:23 they're actually bigger than that four micron so the energy between the electron clouds can interact with each

23:28 other and that's how we get entanglement and then we can create what's called

23:35 there's a full gate set that's defined it's i think 22 gates that are out there and we can run

23:40 single cubic gates two cubic gates and we can do all that in parallel so this is just representing like the full array

23:47 running different um gates which make up a circuit and then at the end we shine a flash of

23:52 light on it and we can read out the state of each each qubit um and then we can shine another flash of light on it

23:59 and and all of those qubits get reset to their ground state and we can start over and run another circuit so

24:05 again apologize the animation didn't work over zoom for some reason but that's kind of how it works

24:11 if you uh if you want to kind of see it from a system diagram perspective this represents this blue box in the

24:16 middle represents the vacuum chamber it's got little windows in it this is where the qubits are

24:21 we shine different colors of light into it via different different wavelengths of lasers we

24:28 control all that with a rf control system um interestingly uh it looks a lot like a cellular base

24:34 station um it's a micro tca chassis it's got rf you know line cards in it we're

24:39 running software defined radio network over that so we control everything with software and we're just sending different tones of light or tones down

24:45 the down the cables the rf cables to uh to the lasers to to change the light pulses

24:51 and we're just controlling amplitude phase and and uh frequency on that light in order to go do that and where it is in space

24:58 now we read out with a camera it's all this is it's it's focused through a microscope objective so it's a

25:05 kind of a regular high-speed high-resolution camera that's just focused through that microscope so that's how we read out and all of this

25:10 goes um back uh into a rack of standard servers that sits off to the side of the

25:16 system that runs our software stack which would be our schedule our operating system our

25:21 apis our storage and all that kind of stuff so that's kind of like logically what the system looks like uh physically if

25:27 you walked into our lab in berkeley this is what it this is this is it um

25:32 so this this system here is our first it's a prototype it's not a not a product um

25:39 a hundred cubits it's 5 feet wide 12 feet long most of it is an optical table that you

25:45 see in the middle off to the right is the vacuum chamber with the glass cell and the microscope objectives that

25:52 where the qubits reside all that stuff in the middle that looks kind of complex is um

25:58 hand placed optical acoustical devices mirrors prisms things like that little micro motors

26:04 um and all of that is kind of hand placed for maximum flexibility today for the for the prototype so that we could

26:11 experiment and make changes at will moving forward all of that stuff has been cad designed

26:17 and manufactured so it's all kind of fixed to what we want it to be so that takes the

26:23 flexibility out but it makes it much smaller easier to reproduce and higher quality less error prone

26:29 so the ironic thing as the system grows in size of qubits from 100 to 1 000 to 10 000 and beyond this the physical size

26:37 of the system will shrink substantially because all of that hand-placed stuff goes goes away into little manufactured

26:42 um modules basically um

26:48 and this is i guess our version of the golden chandelier where the qubits sit they're in a little glass cell which is

26:53 where the high vacuum is it's two plates of glass with the frame around it

26:58 and it's high vacuum the atoms are in there and the and the lasers and the microscopes are sort of like fixed up

27:05 against that um it's about the size of a silver dollar to give you some perspective and it

27:11 won't change it'll be that size all the way up through millions of qubits so again the size of the system will will

27:17 the quantum part of the system right here will stay the same the the infrastructure will all get smaller

27:23 um and the performance to go up so kind of a little counter-intuitive but pretty cool and that's why this technology is so interesting for scaling is is that it

27:30 you know you don't have to worry about all these wires this operates at room temperature etc

27:35 um and then here's just like a little kind of photo of like how do we how do we see

27:40 the system right we don't actually have people looking at it computers are looking at it but um but the the system

27:45 is just read out and when we read it out the atom is either glowing or not glowing to say what the answer was 0 1

27:51 for each individual qubit and all the complex computation happens during the calculation

27:57 which is pretty interesting let's see

28:03 so what are we gonna do with these computers um there's a lot of workloads that uh

28:10 companies are starting to invest in um commercial workloads uh i talked to

28:15 you know like fortune 500 companies pretty regularly financial services transportation um energy pharma biotech

28:24 and a lot of these companies already have quantum development teams they've hired quantum physicists and

28:30 computer science and data science professionals who have gotten some training in how do you program quantum computers or been trained on the job

28:36 and they're starting to invest in um in thinking about their workloads and how do they kind of evolve from

28:43 maybe their current hpc workloads or their current um you know like monte carlo simulations

28:49 and things like that and take into account the capabilities of quantum computing and so

28:55 they're already starting to buy time on some of the systems that are out there from ibm or ion q or d-wave and

29:02 running experiments learning how to program these systems and getting some feedback

29:07 while their application teams are are starting to write you know algorithms that can eventually run on quantum computers when they reach sufficient

29:14 scale um you know frankly there's not a lot you can do with 100 qubits today from a commercial perspective it's really

29:20 around experimentation and people are looking for you know thousands and more qubits to be very very useful for them

29:26 once you take into account error correction but these are some of the the early use cases so battery development clean

29:31 fertilizer traffic optimization drug development artificial intelligence is a really

29:37 interesting one i feel like one of my slides got missing here but there's actually four classes of workloads um

29:43 that people are looking at uh optimization problems you're trying to optimize a portfolio

29:49 optimize uh multi-path routing like think packages coming from like a fedex packages you've

29:55 got millions of packages coming from different addresses every day going to different addresses and they get picked up by trucks doing loops in

30:00 neighborhoods and put onto bigger trucks and trains and planes and then fan back out to trucks doing loops and

30:06 neighborhoods that's a multi-path routing problem and quantum computing is really good at that

30:11 or it will be really good at that when we can actually put the maps of sufficient scale into the system

30:16 um and so you can imagine that you could do like a daily map optimization rather than having a

30:22 fixed function sort of loop of these trucks and planes and stuff that could save time money uh

30:27 fuel and all that kind of stuff so that's one example financial services like

30:33 risk modeling and portfolio optimization is another one simulation's a big one so simulating

30:39 natural processes like molecules attaching to diseases for drug

30:45 uh discovery and drug design is a good one

30:51 and things like that also encryption and security is also a big one that gets a lot of probably more um

30:57 probably more news than it should but it is a big deal and um and then machine learning

31:02 optimizations is sort of another another big category um so that's kind of my spiel on like

31:09 what you know what we're doing and what others are doing and how we're different um and some taste of what some of the

31:15 early applications are you know if i uh domenico asked me to kind of think about like well what does

31:20 this mean for you at juniper and and two things came to mind um and i invite domenico to comment on this as i uh

31:28 after i sort of describe my thoughts here and then we'll open it up for questions so um the first one that i that is kind of

31:34 front of mind for me is hybrid compute architecture so um i think for about the last decade or

31:41 so hybrid compute architectures have become the norm or the standard both in the

31:46 cloud service providers as well as in like high performance computing uh in the enterprise

31:53 uh and what i mean by hybrid computing is that you just have different kinds of nodes right you have cpu only nodes that

31:58 are maybe running the base application user interfaces and things like that you've got gpu nodes that are that are running um

32:05 you know vector processing workloads you've got maybe some network optimized nodes that are running

32:11 uh you know security software or communicating with the outside world

32:16 and all of these things have sort of been put into a cluster and there's some kind of a cluster controller and

32:21 scheduler that steers the right workloads or portion of the workloads to the right or right nodes within these clusters right

32:27 and i think and this is just a kind of a photo i grabbed from amazon that just kind of describes the current

32:32 architecture and i just tacked a quantum computer in there because honestly i think that's how it's going to be

32:38 we will have quantum computers that get integrated in with these cloud services and or high performance compute clusters

32:46 and certain portions of workloads will be offloaded to that quantum system and the results will be returned and

32:52 integrated back in with the rest of the workflow and you know from a juniper perspective uh

32:58 you guys are are uh both at the heart of how some of these things

33:03 get connected together within a data center as well as how data centers get connected to the internet and how data gets transferred you know across the

33:09 internet and um you know every time a new architecture gets introduced into one of

33:15 these hybrid compute nodes there's opportunity for network optimization for bandwidth

33:20 latency you know quality service control security and all the rest of that so i

33:26 think it would be great if juniper could think about how to

33:31 uh integrate all of these different subsystems with quantum subsystems in the future to make sure that

33:38 everything works well together um and is secure and then the other thought sort of on

33:45 the right is quantum computing is great and that's what my company is focused on and where my personal interest is but there's

33:52 other quantum physics applications and sensing networking and

33:57 communications and encryption are the three probably biggest ones and arguably will be deployed before

34:03 quantum computing quantum sensing's already deployed to a large extent in the military for

34:10 navigation for aircraft and submarines and and

34:16 and ships and uh and that's probably the earliest market for you know for global

34:22 positioning systems and all that for quantum technologies uh quantum networking is another one and this is

34:27 how do you uh transfer information or states uh you know

34:33 quote-unquote instantaneously between two different places in the planet or in the solar system

34:38 um and also how do you connect quantum systems together so maybe you don't have to convert from quantum to classical back

34:44 to quantum right we can just stay in a quantum domain and what performance or capabilities might that bring to the

34:49 table and then the third one there is encryption and encryption i think applies to everything you know all the

34:55 above and the space including the classical computing but um you have to think about it both from

35:01 an offensive and a defensive perspective um you know most encryption protocols as

35:06 you guys are probably all aware are based on you know prime factor or you know prime number

35:11 um factoring and things like that and um and this is one thing that quantum computers are known to going to be

35:16 really good at is is how do you find all the factors how do you find all the prime factors of any given number no matter how large the

35:22 number is right so um so we nist is already working on new standards that they think are going to

35:28 take a few years to roll out um that will be more um

35:33 uh defend against you know the the threat that quantum creatures uh um

35:39 you know uh pose to current you know like shah standards and things like that but um also how can we use 

35:46 quantum computers to provide better encryption on the sort of offensive side