The Next Wave with Young Sohn – Unleashing New Frontiers – From Chips to Industries
Hi Simon, good to see you again! Thanks for joining us for The Next Wave! You know, during this pandemic, it's really hard to do things face to face, but I'm glad at least we are able to do this virtually. So, let's start with Arm. You know, I think I am of course privileged and have been involved with Arm for many years as a customer, as a board member, as a partner, but not everybody knows Arm. So I think it'll be worthwhile for you to educate the audience about the role of Arm, what Arm does and (give) a little historical perspective as well. Well, first of all thanks for having me today, it's great to catch up! We've been friends for a long time and it's good to see you. So, Arm is a slightly unusual business. We often get described as a chip company but that isn't what we do. What we do in our core business is design microprocessors, but instead of putting them inside chips and selling chips, what we do is license the designs to sending it up to companies for them to use as a building block within much more complex devices that they're making. The company will be
30 years old in November, so we're about to celebrate a big milestone. And at the time the company got formed, semiconductor manufacturing technology was just getting to the point where it was feasible to start building much more complex chips that became systems in their own right, up to that point if you wanted to build a system you took a bunch of chips which had individual discrete functions and you built the system on a printed circuit board. With each generation of process technology, we've been able to cram more and more transistors into a chip and that gave rise to the ability to put those systems inside a single piece of silicon. So when the company was formed the idea was, well, we have this low-power energy-efficient microprocessor, I wonder if it's going to turn out to be useful to embed that inside a more complex chip. We didn't know what the killer application was going to be, but it turns out that in pretty much every application embedding a microprocessor adding intelligence turns out to be incredibly useful. Yeah, I mean I remember the discussions at the boardrooms, and we're looking at number of applications and it's just incredible how widely Arm's architecture has been adopted around the world. And obviously, for consumers that don't know much about Arm,
they probably have no idea that they actually are living with Arm every day. Absolutely yeah, the business model of licensing technology to other chip companies to enable them to put intelligence into a chip to build an ecosystem around it has proven to be incredibly successful. We have today over 500 different companies around the world building chips based on the Arm architecture there's a big ecosystem around that providing software and other design services to make that all happen, but the real impact though is in all the consumer devices, all the devices that go into enterprise networks and drive the internet. There's just so many of them that are now based on the Arm architecture. Over the history of the company there have been now over 180 billion chips built using the Arm architecture, which is quite an expanding number. It took something like 11 or 12 years to get to the first billion chips and now our licensees – our partnership – more than 22 billion a year. I mean it's quite incredible, there's something like over 700 per
second being manufactured. The scale is just astronomical. That is an incredible number and an incredible reach for a small company that came out of Cambridge and today it has a huge footprint that all of us in the world are living with. Anybody that has a mobile phone or anybody that has automotive applications or embedded applications like robotics has Arm in it. You may want to give some perspective around what did it take for you to build such a what I call the digital architecture of choice in today's world that came out of Cambridge, UK.
I mean I have to sort of pinch myself to think about just exactly what we've achieved. When I joined the company right at the beginning, I was lucky enough to be the 16th employee and we were in a converted turkey barn in the middle of the Cambridge countryside and the technology span out of a British computing company called Acorn. It was a joint venture with Apple, Apple wanted this processor to put inside the Apple Newton – if you remember that, the first digital personal assistant – and, you know, we just kind of looked for other applications where this technology might be useful. We went knocking on doors and trying to convince people that, if they wanted to put a processor inside their chip, it would be a good thing to help try and establish a global standard, a standard around which we could build this ecosystem that would drive down the cost of implementing a processor inside a chip. We knew from day one there was no way we could do everything ourselves and so we set out to build a network of partners who were going to be able to use the technology and to prosper from using their technology by building their businesses around this standard. So the goal really genuinely from day one was to create a global standard that would make it very efficient to put this intelligence inside chips, and that's turned out to be, you know, incredibly successful. That's the kind of philosophy that we've taken in to build a company based
around partnership and ecosystem. The Samsung team really appreciates the partnership that Arm has given and all of our mobile phones and TVs and others have an Arm architecture in it, like multiple of them in actually even one phone. So it just shows the power of the processors and the ecosystem that the Arm team has driven. You also took this strategy of partnership and building this ecosystem for others to join in, which made it even stronger. But also you had an angle of providing low power, which I think, if you think about traditional players – they focus on more on performance than power and are always driving a huge power footprint, and I think what you have done in some ways is idea of this innovation coming from below and just hitting it on the incumbent not being able to even deal with the kind of equation that you've been working on. This is an interesting lesson for how
this new architecture can come about. Of course, it took time, but it does show the new architecture can come at a right time when you are patient and if you build an ecosystem together with partners. It's an interesting discussion that I'd like to get your perspective on. At the time that the company was formed,
they said that essentially the process technology was really improving rapidly and it enabled computing to evolve in in two different directions at once. There was the 'how much performance can I squeeze out of these transistors that i put into a chip' and my only constraint is how much electricity can I get from a wall socket, and that was one way to go. And then there was this idea that, if you can eliminate as much power as possible, then you open the door to many many devices which could then run on batteries. And that was the angle that we took, so we always looked for how can we simplify the design, how can we really analyze the design to look at where the power is being consumed, and use smart circuit techniques, smart architectural techniques to make the processors as energy- efficient as we possibly can. So every time when we're looking at new design we're thinking 'what
is the trade-off?', 'how much power have I got?' and 'how much performance can I deliver with the technology within that power budget?'. And that's how we think about the design process. And then, you know, the way our business model evolved, it was about licensing to others. So we really had to think about our success coming when that licensee of ours actually got into production. So, as well as our own technology, we were thinking about who else do we need to work with, what about the companies that provide EDA software, the software that's used to actually design a chip. Let's go and work with them, let's go and think about how quickly a chip designer can get through the design process using our IP and their tools, let's think about the end manufacturing technology.
So we're always thinking about energy efficiency. That's important in anything that you're carrying around in your pocket of course, but it's actually important everywhere. And then how can we optimize across an ecosystem to produce the best result. What makes it unique is the fact that, unlike other players, you're not just making products, but you are making and building an ecosystem, so that the customer, the end-customer can be successful by providing the better tools and better alignment to the manufacturing skill all those type of things it's incredible what you have done to accelerate the development of new electronics by not being able to design these very traditional custom compute engines and then you of course add other blocks like graphics and networking as well as interface logics and all those different blocks that people need to build the chips they made it easier for them to do it so i think i think this whole idea of creating digital architecture is really a modular architecture and i think i'm in a way pioneered it from traditional compute perspective and you made a migration to much a um a next-generation approach and the question i guess i have is now what's next coming given that some of the announcement that was made by nvidia that wants to have um so i'd like to get your perspective of what is your view of synergy why do you think nvidia will benefit uh by working with um to take care of and provide benefits to your customers the applications for computing have grown and grown and grown over the years the ability to put so much intelligence into a piece of silicon for such little cost um has just uh driven an explosion of the number of end devices and of course we live our lives today you know with mobile devices with just all this electronics that we take for granted but that the and whilst we've done a lot of efficiency and and advances in the compute engines that are in these uh these silicon chips um that the nature of that computing has been fairly constant over the years they're it's running conventional code that's been taught in universities for years and years there is a transition coming though over the last little while we've seen massive improvements in the field of artificial intelligence and we look at the next generation of computing being a combination of classical computing and ai driven computing where data is being used to make decisions and being used to make decisions in an autonomous way through algorithms that are effectively using machine learning and ai techniques to extract information and look for patterns within that data so i think the nature of computing is changing we've been uh studying that we've been uh expanding our own products to um provide machine learning accelerators to provide software frameworks to allow applications to be built and nvidia have been been driving that in you know to huge success of thinking about very high performance applications for ai now we both companies we see ai just going everywhere being as pervasive in the future as conventional computing is today and we believe that if we take the strengths of arm the ability to uh create pervasive technology and to be able to create ecosystems around what we do we take that and we add and we combine it with the strengths of nvidia and the all the work they've done on really high performance ai compute the work we've done on the the software stack actually enable applications to be built we put that together uh we think that creates a really compelling future in delivering how we deliver this next generation of computers so that's what's behind the acquisition given the importance of um and its architecture for really number of devices that are based on what we call digital architecture uh and i think you mentioned 190 billion soon by the time we are done with our interview maybe it'll be 191 billion it's accelerating but um you know the safety and security is something that people are really a a worried about as the architectural proliferates and reaching the mass number of devices so i want to just get your perspective around what are you what is i'm doing about the security yeah it's a huge subject for us and it and it's uh it's a sort of um horizontal um technology that cuts across absolutely everything that we're doing to your point you can you can build like incredibly secure systems and you find it's the smallest cheapest thing that created the entry point into the network so specifically around that we've been looking at how people build uh secure iot devices for uh some time now we we introduced uh what we call the platform security architecture psa which is a set of guidelines for how people build very small very low cost um iot devices but have a an architecture that then enables secure software to be built on top of them so go by defining the kind of building blocks that need to be within the chip um how to partition memory up how to keep encryption keys away uh from from main memory how to uh partition algorithms uh how to build a secure operating system that's really important because in a world of uh a trillion connected things a world of billions and billions of iot devices it only takes one for the bad actor to be able to get into the network and we want to do everything we can to try and prevent that we've also been thinking about how software on those devices get managed through its lifetime any computer that you might use you can have operating system updates pushed to it if there's a security issue we need that to be in place for absolutely everything that's connected to the network that's a that's a huge piece of work that's uh that is ahead of us for for the industry and then as you say uh security needs to go everywhere we're thinking about how big cpus are sitting cloud data centers how are the the applications within them separated from each other how does the hardware help support that um how do you contain um something if there is uh if some application does get penetrated how do you contain it so it can't get into the rest of the system it's driving changes in the base architecture it's driving changes in the way that software is written across all of these devices and again we're taking a very partnership based approach uh to try and help address these problems working with people um thinking about what solution is going to be needed not just tomorrow but five ten years from now and thinking through the changes in the architecture that are going to be required to deliver that so it's a huge piece of work but one that's crucially important for us as you know when i was in the boat mobile phone was a major driver and it was in two ways the volume was going up and you know it was hitting one and a half billion units per per per year and then they were shipping multiple arm processor per phones so this multiplier effect was really great for arms growth in during late you know earl late i guess the 2000s and early 2010 era but now with the mobile phone a bit saturating because everybody has a phone uh what do you see the next big growth form going forward yeah i mean the growth of mobile has been phenomenal for for the company and we've had two ways of it the initial wave of uh digital mobile phones and and then smartphones um and and over the years uh you know we've seen arm technology being used in many other applications but i think i think we're at a moment now where um the the growth of semiconductors looks really really positive for the for the years ahead and it's a number of markets that are driving that growth um we're seeing that the deployment of 5g 5g is way more complex than 4g so it's driving the adoption of more um processing within all the chips that build up the 5g network and then on top of that you you have the growth of iot that's billions of devices they're really small but they're enormously high volume devices that's a driver we have autonomous vehicles highly complex you know data centers on wheels that need to be driven pardon the pun in a way that makes them really energy efficient but enables a massive of compute power to do everything an autonomous vehicle requires so there are all of these things going on at once and if this combination of um ai as a the kind of core element of processing 5g networks iot this is all happening together uh and i think is going to drive the growth of the sonar capture industry for years to come and then the other place uh the dimension is is the data center and you know the cloud is is growing like crazy um and that's another area where through our focus around energy efficiency we're able to provide some real benefits there as the cloud expands um with the performance that we're now delivering uh we're starting to see more deployments of armed processes in the cloud as well and that's an area people you know once upon a time thought we'd never be able to address but uh but today we are so a lot of these applications are very exciting and actually these are all the footprint of what's driving the demand for the uh devices as well as the applications that are making our lives better right uh technology and compute power can help improve um our day-to-day living but there are so many other areas where this deployment of computing the deployment of iot devices the use of ai can be used to address some of the hardest problems that the world's going to face over the coming years and things like climate change things like growing populations things like pollution you know everything that's uh encapsulated by the u.n global goals these are areas where technology can make a big difference and it's it's
something i know it's an area of focus for you personally young and there is an area of focus for samsung and it's something that we're big believers in that the technology can help play a role i actually do appreciate um also being supportive of the initiative that i also started called extremetech challenge startup companies are using technology and innovation entrepreneurship to solve the sustainability issues and i'm hoping that all of us in corporations they can be able to work with startups to accelerate these innovations so i'm going to continue to support and push extremetech challenge and i also know that you have a 2030 goals that you've been pushing as a part of the same agenda in a different scale in the time frame yeah i i think that the key thing about about what we're doing in various different ways is um that this evolution of compute power is it's been incredible it's it's giving us these you know mobile phones that we can play games on and do cool stuff but the the the potential impact goes way beyond that and what i've seen through arms history is um if you continue to drive down the cost if you drive up the performance if you make it easy for people to get access to these technologies then what you do is is kind of spark their own imagination you you enable people to come up with the ideas that you can have possibly have thought of no matter how many smart people you've got within your company because you've got these kind of business uh objectives that you have to deliver on but if you can put all those tools in the hands of other people then they come up and they create great ideas and through string tech challenging through what we're doing in 2030 vision um it helps uh highlight that and and with these these technologies that we put in people's hands i'm really optimistic that we're going to see great solutions coming around and that do solve some of the world's biggest problems yeah we need to partner power up and solve some legal problems and make some impacts right i'm so happy to hear that i'm so glad also we can be a partner in multiple levels including this particular agenda which i believe is really an important one well simon i really appreciate your time today we had a very you know variety of discussions but we have a long history and i believe we can really work together and i think fundamentally um really bring the partnership into the ecosystem and that's what made it possible in my view well thanks for sharing i really appreciate that well thanks a lot you know a lot of questions
2020-12-22 14:33
