SemiWiki.com - Podcast EP311: An Overview of how Keysom Optimizes Embedded Applications with Dr. Luca TESTA

Episode Date: October 15, 2025

Daniel is joined by Luca TESTA, the COO and co-founder of Keysom. After studying microelectronics in Italy, Luca obtained his PhD in France while working with STMicroelectronics on analog/RF circuit design. Dan explores the charter and focus of Keysom with Luca. Luca describes how Keysom is providing an automated and reliable… Read More

Hello, my name is Daniel Nenny, founder of Semiwiki, the Open Forum for Semiconductor professionals. Welcome to the Semiconductor Insiders podcast series. My guest today is Luca Testa, the CEO-O and co-founder of Kisom. After studying microelectronics in Italy, Luca obtained his PhD in France while working with ST Microelectronics on analog RF circuit design. After several years of experience as an analog RF engineer, Luca began working with amplitude lasers, where he became head of the electronics department until 2018. When he co-founded Concepto, an engineering firm specializing in embedded hardware and software

solutions, he then co-founded Kisom in 2022. Welcome to the podcast, Luca. Thank you, Daniel. It's great to be here. So what brought you to Kisem? Well, Daniel, this story behind KISOM is quite simple. As you mentioned in 2018, I co-founded together with Cyril, who is the president of KISOM, an engineering cabinet in France, which was specialized in embedded systems.

And back then, we had the chance to work on many different projects, spacing from IoT for agriculture to laser and defense. And each time the processor, chosen for the application, was over killed, causing a waste in silicon area and power consumption. And this was due to a simple fact. SOC designers could only choose among a limited list of off-the-shelf general purpose course. So, well, we decided to change that. We wanted the semiconductor industry in it for a core

to be able to choose the content of it, quickly with no deep expertise in core architecture, and providing a tool, Core Explorer, that delivers all the software ecosystem associated with the hardware. So that is how KISOM was born. We are today a semiconductor IP company specialized in customizable processors based on the Risk 5 architecture. Oh, interesting. So the value proposition is smaller, faster, and more efficient processors tailored for the job at hand. Well, exactly. We observe that on average, 40% of the instructions of a general purpose processor are not used in the lifetime of a specific application.

And it is to me a nonsense that different products, such as motor controllers, DCDC converters and intelligent sensors, embed the same core while they need different complete features. So our solution allows to integrate only those instructions and building blocks that are needed. I mean, if you don't need interrupts, get rid of the interrupt controller in a click. So we generate tailored Risk 5 cores focused on what is essential, avoiding unnecessary complexity. Right. That's what reduced instructions said is all about. So Kisim is doing something quite innovative with 32-bit processors, reducing, unnecessary instructions. Can you explain how that works? Absolutely. Well, traditional cores tend to be general purpose, meaning they come with a broad

set of instructions grouped into extension sets. So you have the base integer i, for example, plus many extensions such as M for multiplication and division, or A for atomic instructions, etc. many instructions of each extension are never used in a specific application. This creates overhead in area, power consumption, and increased design and verification complexity. So at Keystone We designed Core Explorer, which is a design space exploration tool to analyze the customer's workload and then streamline the tailored processor, basically removing instructions that are not needed. That can be done either by the customer through our simple graphical user interface or automatically having the customer

application as input to Core Explorer. Basically, well, think of it like tuning and engine. You get rid of what slows you down while keeping everything you need for performance and compatibility. Okay, but isn't there a risk of breaking compatibility? well well daniel that's that's a main challenge indeed customer they want optimization but they also rely on the risk five ecosystem the compilers the debuggers the tool chain so our job here is to ensure that even with a customized iSA they don't lose access to that ecosystem so we achieve this by staying within the risk five framework and providing tool chain support for every optimized core we deliver. Okay, just to be clear, you're pushing innovation, but without isolating the customer from the standard. Yeah, exactly, Daniel.

That's what we try to do. That's what makes our approach, in my mind, both safe and effective. And importantly, we deliver everything in one single run through a drag-and-drop design tool. It's built so that even non-experts, hardware development can generate a tailored processor without deep RTL expertise. And that's really key from a business perspective.

By lowering the barrier to entry, we allow companies to adopt custom processors without heavy upfront investments or specialized teams. That means faster adoption, shorter design cycles, quicker time to market, which directly translates into competitive advantages for our customers and moreover core explorer is very easy to integrate into existing design flows whether customers use legacy tool chains or modern EDA environment and even for the deployment we offer flexibility our solutions run on hybrid architectures or fully on premised setups depending on customers requirements

Okay, just one more question about that. What kind of improvements have you seen in practice? Well, one great example is in IoT devices, where battery lifetime and costs are critical. By tailoring the instruction set, we are able, on average, to cut silicon area by 30% and reduce power consumption by 25%. and that translates directly into longer battery lifetime and lower manufacturing costs. And this is not just statistics and simulations. We have proven results in real tape outs. For example, in 2024, we delivered the tape out of the core, which was specifically tailored for a crypt application that was making use of the AES-256 algorithm.

And with Core Explorer, we reduced power consumption by 40% and silicon area by 70% compared to a standard implementation. And that is an amazing factor, and it clearly demonstrates the impact of our approach in practice. Right. So what does the landscape look like? How do you position yourself against the competition in the processor IP space? Well, we start now to see some consolidation in the RISC 5 IP provider space, but you are right that there are still big competitors out there. And, well, I have two answers for you, and the first one is strategy. At Kisan, we've chosen to focus very clearly from the beginning on small, efficient, 32-bit course for embedded applications, such as IOT and EJ. and instead of trying to cover every segment of the processor market we prefer to do one thing but do it extremely well focusing is about say no used to save steep jobs and I believe that applies to key some as our course we say no to being generic so that we can say yes to being the best at lean optimized 30 to bits course where power consumption and silicon area are critical and my

The second answer is ecosystem. Customers come to us with a vision, a market target, and the risk five core is only part of the solution. KISOM is not only an IP provider. We deliver a complete solution, hardware, software, framework to accelerate time to market. So let me illustrate this with an example, edge AI. We not only provide a core for the customer

seeking to integrate AI in a digital, component. We provide the complete workflow that allows the customer to deploy with no effort, the small AI model into our core. So basically, we deliver a complete solution. So Daniel, yeah, I can say strategy and ecosystem are the values that make KSOM stand out in this competitive market. Great answer. So KSOM is a European company, but now you're making a move into the U.S. What's driving that decision and how are things going? Well, the U.S. is obviously a strategic market for semiconductors. Many of the companies building next generation systems are based here.

And to better serve them, we have decided to hire a VP of sales in the U.S. giving us a local presence to engage directly with customers and partners. Yeah, that's a good idea. Risk 5 is really picking up, ramping up in the U.S., so I think you'll do quite well. So just one final thought, Luca, if you have to leave our listeners with one key message about KSum, what would that be? Well, I think my message would be core IPs don't need to be one size fits all. With the flexibility of Risk 5 and our customization, customers can get an architecture, that is truly adapted to their needs.

And looking at the business side, a tailored approach translates directly into competitive advantages, better PPA, lower costs, and faster time to market. Yeah, I think you made that message, Claire. So final question, how do customers normally engage with Kiesm? Well, we have different channels to reach out customers. Our sales team is growing, both in Europe and in the US.

And we also carefully selected a few partners, sales representatives that cover Europe and the US for KISOM. And of course, we are also involved in technical community exchanges, and we participate to main events such as Embedded World and the Risk Five Symposiums. And of course, you can have a look at our website. Great. And we'll see you at the Risk Five Symposium. Is that sure, right? That's for sure. Thank you for your time, Luca. It's a pleasure to speak with you, and I hope to see you at the conferences. Well, thank you, Daniel. See you soon. That concludes our podcast. Thank you all for listening, and have a great day.