 Hi, I'm Thomas Aguay. I'm managing all the activities related to IoT embedded form. And so we met, we did a video I think two and a half years ago maybe? Yeah. And that was the launch of the RMV8M, right? Correct, yeah, sure. So, but now we're seeing, so I just did some video. I saw some M33s, the shipping. Yeah. Are the shipping 23, 33? Yeah, correct. So in fact both processors are now available into a real chip. It's very nice to see all the investments were made into security, defining the RMV8M architecture, making a treasured available into the small MCUs. And it's very nice to see how it'd be a fridge now and how the security is becoming more and more important for IoT and embedded. It's not only as a CPU itself, it also we are bringing more ways to accelerate the proliferation of this technology. For example, we have just announced last year what we call the secure foundations. The secure foundation IPs are a set of tailored IPs covering the CPU, but also subsystem, secure IP, OSCs, tool, trusted firmware, which all together forms a kind of cornerstone to quickly develop a secure system. So you have to provide the full package IP. But that's what ARM has been doing all the time, but now it's a different package or? Well, basically now, I think now what we are taking more and more into our way of thinking and way of working is to better align all the availability of the IPs whether it's a CPU or a subsystem and the sort of which go with it so that we get all the elements of the solution at the same time. For example, also we have introduced earlier this week we have announced the PSA certification program. So PSA is our platform security architecture. It's our framework to have a coherent and consistent way of addressing security for the industry. And now we have just released our certification scheme with different layers and levels. It's included also the early adopter of the technology and also very important we have in the board security people like security labs which will help us to satisfy the solutions. So I did a video last year. I think Cypress was talking about some PSA stuff. Correct. And so you have good partnership with them, I guess. They are very good at taking your security stuff and the next generation security is even more better than the previous, right? Yeah, we're in Cypress because we're working with many partners. I mean, this announcement I've done earlier this week we have also NXP, we have ST, we have Silicon Labs, we have Microchip and I'm sure I may have forgotten some of them. We also have an OS provider called Zaya, who is also a PSA compliant from the start which is very interesting. So it's not working with one partner but working with the key innovators in the field. And how do you manage that? Because there's so many companies, everybody wants to work with ARM. It's a lot of management. Well, you know, we've been working with all these companies for many years now. Even people who are coming from the New Vuitton. I forgot New Vuitton for example, which is more Taiwanese suppliers who has been very early into the program. I thought that we stay fair. I mean, in the end we are here to serve our partners and customers so taking the inputs and seeing how we can help them to build a more compelling solution. Because two and a half years we needed the video, it was just about after the vote of Brexit, right? And also the acquisition of SoftBank was everybody wants the independence of ARM to continue in a smooth way. And how has it been, the feeling in the last two and a half years? Has it been some acceleration and innovation in the world? What have you done? I mean, first of all the great thing with the acquisition of SoftBank is now we can really leverage all the connections of SoftBank. Also, we have better control on how we can work on the medium and long-term. So it has been really a very great addition for us. I forgot to ask a question. Because there's a lot of investment to do in the embedded. Correct. And where do you put that investment? Where do you prioritize? What do you do? Okay, so there are different axes. First of all, you may have seen SoftBank on the show. Well, it's not only anymore only about MCUs, but more and more also about MPUs. So you will see the balance between MC and MPU now changing a bit. You see a lot of rich embedded systems running rich OSCs like Linux. And we see a lot of accelerations also like NN, so neural networks or machine learning on these platforms. You know, we have released a year ago the ARM NN framework. We have even just released it and donated it to the open source. It's now managed by Valinaro. We are bringing this kind of technology, machine learning, down to the smaller devices. And we have just announced earlier this week, just two weeks ago, I'm sorry, the ARM Helium technology. So this new technology is building on the ARM VHM architecture. And it's bringing the so-called vector extension to the M profile, which would deliver a massive compute uplift both for signal processing but also for machine learning. We're talking about up to 5x up to 15 times the performance compared to joint solutions. So is this a little bit like the theme of this show? Is AI an embedded and neural networks an embedded? Or is it like a new thing? Let's say we see more and more compute at the edge. Whether the compute is AI, ML, or more signal processing or more heavy lifting scalar compute, we see more and more interest to have the compute happening from where the data is collected. We don't need every time to send the data up into the cloud because of latency reasons, because of privacy, because of security, but also it costs energy. You know, when you are sending data from a node up to the gateway, sometimes you may have a better usage of this energy to process it locally in the node and assess if this data is valid or of interest before sending it up to the gateway or to the cloud. So embedded well is a big world, right? Does everybody have to think about AI in the embedded? Is this something that affects everything? Or is it just a niche or a small part of the market? I think AI, it may become a kind of the paradigm shift. Yes, so I think the way people have been developing into embedded worlds, moving from assembler to C 25 years ago, now we're using more and more of the Chef software was very good. Now we're looking at ways how we can solve differently the problems. And I think this approach is complementary. We may not need to use AI in every use case. I think today AI and machine learning is very successful in vision, for example. We're looking now at other kind of problems to be solved like what you call a voice control. Machine learning and neural network can also work very well in this area. Or what you call, for example, the third V. So one V was vision, one was voice. And the third V for us is vibration. So how we can better analyze some signals coming from sensors. It could be, for example, vibration on the motor. We could do some predictive maintenance, for example. And here machine learning is indeed a very interesting path to address this challenge. And it's very important to do it at the edge to save bandwidth, because there's not enough bandwidth for like a trillion devices, right? Yeah, bandwidth, but also, you know, it's somewhat of a latency. But privacy, I mean, do you have to share this data up to the gateway into the cloud? Maybe not. So some countries are more reluctant in terms of privacy of data. And if you can really use the data locally and make the best use out of it and not to overload the infrastructure and the cloud, then maybe it's a better use of also energy in this case. So Masayoshi-san, two years ago and a half, he said he wants a trillion devices. Are you on track? So is it growing constantly in the Cortex-M section of the market? Is it growing a lot? Yeah, correct. So I think to date we have our partners have shifted more than 130 billion devices and it's always growing and growing every quarter. So yeah, I think we are trying to fulfill the vision of Masayoshi-san. My thinking, but I don't know if that's the right way to think, but you know, some people are talking about risk five, but the arm is like you have the IP that keeps evolving all the time. If somebody is doing something that's free-license, it is not going to be able to keep up, right? Well, in fact, in the end, we are here to provide technologies. We are here to provide the best technologies, the best enablements, and also we're looking at more than just a CPU architecture. Here what we want to offer as a value is how we can speed up our partners with solutions, with systems, with software, with tools, which was all together. So I think we are more working at this kind of a different level, a different scale, more at solution level. It's kind of long-term support, right? Also, I think, we have some great initiatives. Recently, we have taken our greatest and latest Cortex-A core. We have been adding some safety support to it. We have created the Cortex-A 76AE, the Cortex-A 65AE. It stands for Automotive Enhanced, and now they are being used in Automotive and bringing a split lock, or drag-call lock step, into Automotive, which is a great innovation. So I think here we are really looking forward how we can innovate and bring the next big thing to our partners. There's a special milestone right here you're talking about? Oh yeah, definitely. So we are very glad to announce with Raspberry Pi Foundation that we have achieved together 25 million units of Raspberry Pi being the ship and distributed. I mean, this is a great project. It started to be more focused on education and reaching out to students to help them to play with accessible computers, accessible both in terms of price or accessible in terms of ease of use. And now we are very proud with the foundation to have achieved this very nice milestone, so 25 millions. So you're going to give out 25 million more now? Oh yeah, very great. So tomorrow on the last day of Unmitted World, we have the so-called student day. And during the student day, we will give away 25 units of the Raspberry Pi to the first students coming to the armed booth, sponsored by the other foundation. So thank you very much to the Raspberry Pi Foundation for this sponsoring. It's a big success story, right? It's one of the most popular development board systems out there. Yeah, definitely. We are seeing more and more users of the Cortex-A to embed it. And this platform is great. Currently the Raspberry Pi 3B Plus, I think, is using a quad-core Cortex-A53. And it's very affordable. There are a lot of interfaces around it and so many use cases. At the beginning, it was mostly used by students at the beginning and for educational purposes. But now we see them being used in industrial applications. So it's a very big success. Recently launched Embed Linux, which is related to the Cortex-A, right? And how big is that going to be for the embedded? Well, Embed Linux is also a very important milestone. It's another way to connect to the cloud. Embed has started predominantly on the M-class in the past years. And now we are extending the outreach of Embed OS to Cortex-A platform, so the rich platform. It will ease the combination of multiple applications coming maybe from a different horizon and to be able to connect to the cloud in a secure manner using Embed. How has Embed been successful so far? Is it a big success? Is it big use? So yeah, we see a lot of pickup of Embed and Embed OS. I think what's very interesting for the people is not only an OS but it releases an entry door to the connection to the cloud. And here you will see through the acquisition we have made, like stream and treasury data, it's a bit bigger story than just OS. And everything is open source? So other distributions are using the innovation you're doing, right? And it goes to other things too. I think it's very good. I mean, in the end, we are here to share innovation. And we are a strong supporter of open source. For example, we have created the so-called Linaro Foundation, which is a cross-company effort to maintain Linux on ARM. And we have all our partners who are contributing here. And Embedded is playing a bigger role in the Linaro. Correct. For example, there is this one working group called Linaro Lite, which has been called to life two or three years ago, which is focusing here also on IoT use cases, whether it's M-class base or A-class base.