Good morning, everybody. Great to see you all here today. Thank you for coming out to hear the Ceva story, and I know it's been a few years. We just found out it's been five years since our last Investor Day, so very pleased to come out and share the vision of Ceva . As I said, it's been a while. You might have noticed this morning and from the slide deck as well, that part of today we've done a brand refresh, changing the company, kind of bringing it to where we believe the company has evolved to. It's not really reflected in the brand and the image, and we'll talk a lot about that today and the different markets we're playing in.
But we announced this morning that, you know, the company going forward, it's really an IP powerhouse or an IP company that many different products, not just DSP. So we're changing out a few things, and we're gonna go through that. But before we start going through the slides and the agenda for today, we just have a quick video. It'll take about two minutes. We wanna show and give you an idea about what Ceva is doing these days and what the company is about.
All around us, everything in nature is interconnected. No matter the season or setting, there exists an invisible network of knowledge, understanding, and discernment flowing seamlessly together in rhythm and harmony. It's an inclusive force, driven by a need to connect, sense, and infer, coupled with a deep, underlying intelligence. Ceva is bringing this force to the Smart Edge. Ceva develops innovative silicon and software IP solutions that power the Smart Edge, so smartphones can seamlessly pair with earbuds, drones can sense their surroundings and sync with VR headsets, and autonomous cars can understand the world around them. That device you're using right now, chances are it's powered by Ceva technology. Our licensable technology includes Smart Edge comprehension of sound, vision, and motion, as well as wireless communication between devices and over cellular and satellite networks, ensuring you're aware and connected to everyone and everything around you.
Ceva technology innovation delivers efficient, ultra-low power processing platforms for high-growth end markets. Today, Ceva technology is at the core of more than 17 billion devices. Our proven IP solutions help shape a smarter, safer, and more interconnected world. It's why the Smart Edge runs on Ceva .
Okay, so for those of you who don't know me, my name is Richard Kingston. I run investor relations, market intelligence, public relations, a few different roles at the company. Been with the company for about 20 years or so, so been through a few different changes and it's always nice to see the next evolution and the next era of Ceva . Amir will talk about that during his presentation as well, about where the company has come from and where we're going. Before we start, just we have to quickly look at the legal notices here. Forward-looking statements are in the deck, and everyone can see it. Also, by the way, we filed this deck this morning with the SEC, so it's all available to the public online.
One other piece of important information, all the numbers and the information and, the data today we share about Ceva is with Ceva recognized as a standalone company after the divestment of Intrinsix, the design services business we recently sold and closed in Q4. All the numbers represented are Ceva standalone. So for many of the previous years when we had Intrinsix as part of the business, you may not have seen those numbers previously. They're included. In the back of the deck, we have an appendix with different P&Ls and so on, where Ceva is listed as a standalone and the Intrinsix numbers are not included here.
So like I said, as part of today, one of the items or the plans we've been working on behind the scenes is to really give Ceva a brand refresh. After many years of being known as a DSP powerhouse and a handset-focused company, we realized in the last few years the company's come a long way, and that's not really reflected in the brand that we've had since the early 2000s. And the decision was made to give it a facelift and bring it more up to date to where Ceva is today. Just to give you an idea, there is a lot of thought went into the logo and where the company is, and we kind of tried to explain the logo a little bit.
As you see it, it's different, obviously, to what we had previously from a couple of different elements. But when you see the C and the E together, they're connected, and that really reflects the strength of our wireless communications and connectivity business. That brings that to the forefront of the brand. The V, you know, it's shaped in a verified or trusted way. We wanna be a trusted partner to our customers, and we believe that we're making great headway in that respect. Many great customers who rely on Ceva over the years for IP, and in many cases, not just one or two IPs, but a portfolio of IP.
And the triangle that you will probably all remember for a long time at the top of the V is now being replaced with a circle or a rising sun in a burnt orange, giving the impression of, you know, a rising sun towards a new tomorrow. And the reality of it is, a lot of the technology that we're working on in R&D today will be in the market in four or five years' time, powering the next generation of devices.
A lot of the devices today, the IP was developed a few years back, and I think when you hear the story today, you'll realize that a lot of what we've been working on has changed the profile of the company, the customer base of the company, and the different markets we serve, and we're looking forward to sharing that with you. I should finally say, as part of this as well, the domain or the web URL for many years has been ceva-dsp.com. Today, it's changed for the first time to ceva-ip.com. So all our emails, everything now is Ceva IP, not DSP anymore. DSP continues to be a very nice, important technology within our portfolio, but it's not the core focus.
It's one piece of a big, broad portfolio, and I think Ceva IP is a much better reflection of where the company is today. And the smart edge, we'll – you'll learn a lot more about the smart edge today, what we believe or what defines a smart edge for Ceva, and the opportunities that we have in that space. I think it's gonna be pretty compelling. So just to look at the Ceva team. Today, we'll have three presenters. Amir Panush, our CEO, will present the first session, followed by Yaniv and Michael, who will follow on with their sessions. But we wanted to give you a sort of a look at the Ceva management team today as it stands.
We don't have all, obviously, all the R&D people and the different general managers, but from a corporate perspective, you know, there's been some nice refresh and good change in the company in the last few years. Gweltaz Toquet, for example, he's been with Ceva a long time. He's our Chief Commercial Officer now, after running sales in Europe and Asia for many years. He's now running sales worldwide and responsible for all the commercial activities. When you look to the left of the slide, I should also say today we have Iri Trashanski here, standing on the side. Iri joined us about four months ago, came in in the new role of Chief Strategy Officer. The role was created under Amir to help focus the company on a strategy going forward and look at opportunities and different areas like that.
Iri brings a lot of, you know, experience in other big corporations that can help to give Ceva some new ideas and new directions. And when you look at the left side of the slide today, between Iri, Amir, and Dana, who runs our HR, they blend very well with the members of the company, like Yaniv, Michael Boukaya, and myself, who've been around for a long time. And now we have this very nice mix of seasoned veterans from outside the company who bring experience and knowledge and know-how from companies like Qualcomm, InvenSense, Marvell, GlobalFoundries, and so on. And that's blended very well with Ceva's internal team that has a lot of in-house experience developing IP and running products.
Now we have a much more rounded management team that has great experience inside and outside of Ceva, and together, we can bring Ceva forward to the next level. So just a quick look at what the agenda looks like for today. So we're gonna start now with Amir. His presentation is about the Smart Edge era. We'll have a Q&A with Amir directly after that, as I know a lot of you are keen to hear from him, and we want to give him the spotlight for that. So we'll, you can answer questions or ask questions at least about his session, because there's no point in talking about the financials and stuff like that until after Yaniv has presented. But certainly for the CEO Q&A, let's try and focus on what he's gonna present and questions around that.
Then we'll have a break, just a quick coffee or whatever, and we'll come back with Michael, our Chief Operating Officer, will present a technology-focused presentation about what we're doing and innovating around the Smart Edge opportunity. And after that, Yaniv will finish up the day, CFO financial presentation, talking about the operating leverage and the margins that we can bring back to Ceva now that we're a standalone IP company again. And then, of course, at the end of the day, we'll have a wrap-up with a Q&A where all of our members and Iri will be available for questions. And, from there, we'll, you know, head out for lunch or whatever you want to do. So that's how we're gonna look today, but for now, I'll pass it over to Amir, and enjoy the day and enjoy the sessions. Thank you.
Richard, thanks a lot, and good morning, everyone. First, thanks a lot for coming, and we really appreciate the investments in our company and your interest in the company. Today I will talk about the Smart Edge era, but before I start with that, I want to share a little bit about my background. It's the first time for me to present to you, and this is started about 20 years ago when I was a software embedded developments engineer. Actually, that was my first interaction with Ceva, where I started using the TeakLite processor, and I found that our processors at Ceva, we have very low power and capabilities to develop the next generation of wireless technologies.
Then I moved to Texas Instruments and then to Atheros, driving lots of wireless LAN communication technologies, starting the new business with the team over there for the automotive market, spreading from outside the mobile into new frontiers. With that integration into Qualcomm over there, I built a team and enlarged it to develop and drive a new business unit related to connecting MCUs with the wireless LAN and the Wi-Fi technologies, and really bringing the new devices into the smart home applications. So from automotive to smart home. And in the last decades, I joined InvenSense, then later on, acquired by TDK.
Over there, I ran the InvenSense group, was able, fortunately, with a great team, to double the revenue in the last few years and diversify the company from an early success in the mobile area, with the motion sensor, to very diversified sensor portfolio of technologies, as well as very diversified sets of markets. And lots of data we'll talk today also, how we are doing things at Ceva. I'm very happy that to join the team and work with the team here and the rest of the group moving forward as we build the company. So before I go into the details of today's presentation, just a little bit about how it's been my first year with the company. The first 90 days was really to assess and understand the capabilities of the technologies, the team.
Our focus areas, going and traveling to meet with key customers and to assess the perceptions of Ceva in the marketplace. What I found is that we have a very, very motivated team with great talent. We have a very broad portfolio of technologies and that are very innovative. And also when I talk with customers, I was very happy and pleased to see that we are perceived as a very trusted partner, and for many years with lots of the different customers out there. But also, as I dig more with the team, I realized that there is needs to focus more the strategy to create clarity of how we invest our resources. And with that, I worked with the team on creating clarity on that, realigning some of the R&D resources.
We made a decision to divest the Intrinsix business and take it back to a pure IP play that I believe, and we believe as a team, have a very great future moving forward. As we look into the next area of our company, we see a very strong growth opportunities as part of the Smart Edge area, and I will talk about how with that, we will focus moving forward on a possible growth for the company in the coming few years. This is our missions. As we look at the Smart Edge in the near of the company, is to be the partner of choice for transformative IP solutions for the Smart Edge. The partner of choice.
It's very important as an IP company to work with our customers and build with them the long-term roadmap in order to be successful in the marketplace, to have a transformative IP. We'll talk today more about our AI solutions, how we're supporting early on and looking forward into Transformers capabilities, and all that in order to support the Smart Edge, and this is growing very, very rapidly in front of us. So let me take a little bit step back about the history of us as a company, how Ceva started, and how this is moving forward. We at Ceva started about in the 1990s in the PC area with our DSP technology, and then it grew very, very nicely in the smartphone area, where we developed our DSP to propagate it very nicely in that marketplace and grew the company.
But also over time, those markets became more consolidated, while the data era really came to place, and we start propagating our different technologies across many different markets and much higher volume in the marketplace. And as we look forward towards 2030, this is where we see even more significant growth of the Smart Edge area. But even more importantly, what we see is more integration of technology and needs of these different devices. It was started with those devices become all connected. Now we see those devices all becoming more contextually aware, with more sensors and other capabilities around it, and also smarter in terms of localization, of inference and ability to make decisions, and also some personalization of those decisions. All that also, there is the tailwind of the overall semiconductor market.
Based on markets reports, McKinsey, and so on, the expectation in the semiconductor markets will grow to $1 trillion. So if we combine the overall semiconductor markets potential growth towards the end of the decade, as well as the overall number of units and capabilities that are needed, this provides for us significant growth opportunities as we move forward. But what really drives... What are the key mega trends? What drive the growth toward 30 billion units, and how this is related to us at Ceva? There are 5 key mega trends that we are, that we are tracking: the IoT, the 5G, AI, automotive, and cloud. For IoT, it's really the whole industrial 4.0 and the smart home, where more devices are becoming connected. Today, it's almost we can't find a device that is not connected.
5G, going to 5G Advanced and 6G, this is what drives the infrastructure to be much more capable with higher capacity, lower latency, enabling use cases, new use cases that are very important for us, because with our 5G, 5G Advanced and 6G in the future technology, that will propagate more use cases and markets and customer base. AI, which is, of course, a focus area for us and overall in the marketplace, this is really transforming the marketplace, where we see those devices needed, the AI localized with more capabilities integrated. Automotive, with the electrification, digitalization, automotive has been a strong market for us, and we will keep building in that market. Cloud, with the hybrid edge to cloud solution, we are at Ceva, focusing more on the IoT, 5G, AI, and automotive.
The cloud is less of a focus area for us, but with the hybrid network of between the cloud and the edge, this is where we play very strongly moving forward as well. But just to explain what we mean at Ceva at Smart Edge and how we look at that, there is the cloud infrastructure, with the edge platform and gateways communicating to the cloud. Those edge platform and gateways are part of the Smart Edge market that we are targeting, as well as all the end devices that communicate to the gateways and between them. The right side is what we call Smart Edge, and that's our target markets moving forward very clearly. Within that markets, every device is pretty much now connected and requires all the different wireless technologies, as well as, as I mentioned, they are becoming smart and more integrated.
Actually, quite interestingly, Morgan Stanley just had a survey with all the top MCU suppliers out there, and they asked, "What are the things that you see potentially are going to get integrated with the MCU?" They said, first, we will need audio capabilities because the interface to all those machines are basically become audio and voice in, triggered. But also what we will need is, of course, all the connectivity integrated with that, and that's a great opportunity for us with all the DSP and all MCU partners out there, to add a different connectivity, fully integrated. And what we see also, OEMs or end customers, they want to have more integration of those capabilities. Last but not least, those devices are moving from so-called being dumb to be smart. But this is where AI really becomes more localized in a different level of performance.
There are the edge devices that need very low power and still some basic AI capabilities for voice activation, for some video, basic video capabilities. And there are, of course, the more sophisticated AI solution out there that need more so-called local smartness in order to run big neural networks and to drive those capabilities. Just to quantify the Smart Edge and the overall markets and how this translates to us at Ceva, overall, the semi revenue CAGR between 2020 and 2023 is expected to be 8%, out of which, for the IP market, is expected to be above $10 billion IP TAM. Overall, as I mentioned, about 30 billion new devices annually, and the Smart Edge semi markets, which for us, basically excluding, again, data centers, memory, and those type of things that are less relevant, is about $600 billion.
So overall, both from SAM, for IP, as well as unit, this is definitely a very significant market for us to go and address. But this is not, this is not new for CEVA. We have been addressing these different type of market segments, as well as so-called the Smart Edge applications, for many, many years. We have shipped more than 17 billion devices to date, out of which about 10 billion devices are in mobile. This is so-called our earlier years of rampant success. But see how much we have diversified in terms of the different markets already deployed all around us. 6 billion devices in the consumer IoT, and that's where we see a lot of the different wireless communication, and then now smartness coming to play.
Automotive, I would say we are not necessarily perceived as the automotive type of a supplier, but we have already shipped more than 100 million devices, and we are looking to expand in that marketplace. Industrial, which is a great extensions of our play and success in the consumer IoT, where we see the same type of suppliers wants the technology to be able to be delivered at cost. They're using our IP, embedding that in their devices, and then go to the different marketplaces. PC, where with our sensor fusion and sensor IP capabilities, we'll talk more about later, very high volume already penetration. Infrastructure chips, the micro base stations, lower volume, but very high performance and capability that we drive moving forward. Overall, very broad base of customers.
We like to win with the winners and really build a partnership with the key players in the market. So what that all means in terms of how we look at our technology and what we want to enable, what are so-called the core common use cases in the smart edge? There is the connect, the sense, and the infer. Connect is where we see all devices using the different type of wireless connectivity technologies, and that's where we have been—we have done very, very successfully, and we'll continue driving that. The sense, the device needs to be contextually aware, where we see different type of sensors, from audio, like microphones and others, to vision, to touch. With our DSP technology, processing technologies, and the software that comes on, on top of that, we are addressing the different type of sensing, application and use cases.
And last but not least, which is a major extension for us moving forward, is to address the infer use case, the localization of inference in order to create that intelligence across the, across the smart edge marketplace. So what that all means to us in terms of some expansion? I talked about there are key mega trends that drive very high volume. There are the different applications out there, but how they translate to the SAM, to the addressable markets for us at CEVA. What we are projecting is that our SAM will go from $1.5 billion in 2022, to around $2.2 billion by 2027, or about 45% growth across the three use cases that we are now supporting moving forward: connect, sense, and infer.
This is already based on a very strong market leadership that we have in the marketplace. We are today the only complete wireless IP portfolio provider. We have everything from Bluetooth, to Wi-Fi, to Zigbee or 15.4, now to UWB, Narrowband IoT, and all the different cellular IoT technologies. We already have and establish a leadership as a Bluetooth connectivity supplier with about 30% market share. We are now at seven at the 10 top MCU providers there, using our IP already. As a trusted partner, we see some of them, and many of them are coming back to add more, more IP and, and technologies from, asking us to go and support them.
We have been an early innovator in vision and Edge AI, and Michael will talk more in his sessions about the new products that we are launching in the market and will come, and we'll start licensing next year. And we already have very good established base of customers, hundreds of customers out there that we partner with. So with that, to conclude, we are looking at about 45% of addressable markets for our growth based on our broad portfolio of technologies and capabilities. Speaking of which, when we look at our technologies, what we really so-called offer in the markets and will drive that same growth, we are focusing on three major pillars of technologies, and that's how we think about our R&D investments and how we allocate that moving forward.
First is we are already the number one wireless communication IP provider. We want to maintain this leadership, expand that leadership, and ensure that we are enlarging that business. In addition, as I mentioned, we are adding scalable edge AI, AI sensing solution, which is another important technology and pillar of growth for us, where we are already powering more than 500 million devices and looking to expand that. Last but not least, which is different than most of our competitors, as a silicon IP providers, we are also offering embedded application software. This is where we are helping our customers, and many times the OEMs themselves, to differentiate with our IP, very differently than our competitors are doing. That's why we have a more comprehensive portfolio of IP, so-called silicon IP, in addition, software IP on top of that.
All that to address the different market segments, which I'll go a little bit more deeply in the next few slides. The key points, as we look at profitable growth of how we invest at R&D, what I focus very much with the team every time that we look into new product category or technologies or something that we want to launch, is that going to be scalable across markets as well as across customer base? The more we are creating scalability of our IP, the more that we can drive the long-term profitable growth of the company. So just a little bit of each market. I will go through that quite quickly. Again, the six markets: consumer IoT, automotive, infrastructure, industrial, PC, and mobile. At the top, you can see our estimation of the SAM, TAM, and the CAGR for each.
At the bottom is the several TAM for each. I would highlight a few things. Consumer IoT, of course, this is a very big market for us in terms of SAM potential, of TAM of close to $2 billion. Automotive, it's a potential additional significant TAM for us moving forward. Infrastructure is smaller, but it's where we drive lots of innovation in order to leapfrog in our technology. Industrial, very good scalability from the consumer IoT and very high CAGR moving forward. PC and mobile, they have been traditionally very important market segments for us to address, and we'll continue to do that. But considering the consolidation in those markets as well as the low growth moving forward, those are the market that will be less of a focus for the company.
Ceva. At Ceva, moving forward, our focus is on those four key market segments, which again, for as an IP company, it's very critical that we are playing in markets that have good diversification of customer base and technologies. The more consolidation, the more challenging it is to really go and leverage our IP capabilities. So now I'll go a little bit more deeply about each of these markets to provide you a little bit of more insights, how we are doing in each market. I will highlight that also, so far, when we reported our numbers and talked about, we more talked about technologies. Here we are giving you and sharing with you how we see our distribution of capabilities across the different market segments that we are addressing.
The consumer IoT, definitely a very large market, and the most important things is really how big is the SAM and TAM potential for us. We also have done very well in that market for the last five years with $250 million, but that shows also the annual potential for us moving forward as we keep increasing our capabilities in that marketplace. Across the three common use cases, connect, sense, and infer, I will highlight several things, like with our partners that I mentioned here. We see our partners, the MCU partners, taking our different type of wireless connectivity technologies, plugging that into all their devices in the connected home, like light bulbs and switches and other devices. All the different robot cleaners that are moving around to be contextually aware and understands where they are heading and moving.
This is with our software solution, not just the silicon IP. We are providing additional value directly to those OEMs, and with the latest acquisition of the 3D spatial audio from VisiSonics, we are basically partnering, and we have the customers, THX, as the lead brand in the 3D audio space marketplace. Why we win in these markets? Why we are doing very well in these markets and will help us moving forward? First and foremost, we have a very broad portfolio of technologies to offer our customers, and we have lots of repeating customers that coming from one technology and they want the next technology. Second, it's already proven in very high volume, very important to our customers. But the other thing is that we are hardwired our technology to ensure that it's the lowest power offering out there in the marketplace.
So any customer that comes to license our IP in this marketplace, if they want this technology with proven and the lowest power, we are able to offer them from the previous standards all the way to the most advanced standard that will come in the future. And I will share later on a little bit analysis, very strong ROI, return on investment, to our customers, which again, propelling, propelling more and more their needs of our technologies. As we look at the automotive, this is a market that is important focus area for us. We are investing in order to have verified automotive qualified products in the marketplace. As I mentioned, we already shipped quite a bit of volume, 100 million units. We generate about twenty-five million dollars to date. The same expansion is meaningful.
I wouldn't say this is as big as consumer IoT, but you can see also that the TAM is very large. So the more technologies that we add, like the AI and other devices that we can deliver to that market, that will propel more and more potential growth for us. You can see several of the partners there. With some of those partners, they are right now going very soon to production with our AI vision DSP for their ADAS systems. We have MCU partners that are using their MCU in the automotive space with our different wireless connectivities. And for vehicle-to-vehicle communication, using our core DSP again as a core wireless capable devices to enable all the different so-called 5G IoT use cases, and we will discuss a little bit more about that moving forward. Why they pick us?
First, we truly have long-term commitment to those partners. We started very early, almost a decade ago, in that space with our technologies, and we keep investing and supporting our customers, as well as qualifying the technologies. The other things, we are working with them jointly, especially the big ones, the roadmap moving forward, which is very important to those customers, and on the highest performance of what they need in their solution. And lastly, making sure that those solutions are scalable. What we see in the automotive, even more than other markets, those customers need the scalability up and down of the technologies because they take one platform, and they take it to the highest tier, to the mid-tier, and the low tier, and they want a good return on that investment.
They cannot afford to change every core every time that they need to go to a different level of tier of their technology. If we look at the infrastructure, so-called the micro base stations, and now coming also the Satcom, the satellite communication, this is a market I would say it's, it's smaller than the others, but this is where we've established very strong leadership. And even more importantly, this is the markets where we drive the highest performance, and we leapfrog in technology. We partners with the leaders in that marketplace. We develop the most advanced in performance, lowest latency, and lowest power that is needed for these type of devices. And then we leverage that technology across the other market segments and customer base.
We have two of the top micro base stations customers as the lead partners for us, and we see more and more expansion and potential in satellite communication as a new so-called infrastructure or gateway opportunities for us, partnering with all these type of customers where the wireless communication is not core to their business necessarily, and that's where we come and enable them with our IP. Last but not least, industrial. This is a market where we see very good leverage for the consumer IoT, and we have done very well. We've already high volume shipments, now about $25 million. We have lots of partners that actually will go and scale in the coming few years very, very strongly. You can see some of the name here.
What we offer in the industrial space that is somewhat different, this is where they need so-called software-defined radio. What they need is the scalability to use similar core and support it across different type of wireless communication technologies, because they want, again, to deploy the same core in different sub-markets of the industrial space. So sometimes they will need a, let's say, Wi-Fi communication, sometimes they will need narrowband IoT communication, sometimes it's the 5G IoT next generation, like RedCap, reduced capabilities. With one core, so-called software-defined radio, we are enabling them to propagate through all their so-called MCU or CPU capabilities into all the different sub-markets. We are also working very closely with... for imaging type of vision AI, in this marketplace, and you can see some of the names here, as well as smart meters and smart trackers.
If I go back before I move to the next slide, in every of those markets, what we see customers are starting with one technology with us, let's say even smart speakers. Just with a simple core DSP at the beginning, then they want to add the Bluetooth, then they come again and they ask for the Wi-Fi. All those things are connected and need contextual awareness. And now what we see, we have lots of engagement of how to enable so-called AI or the smartness, all the way from, again, high performance to lower performance of those devices. So now let's switch, as I mentioned, from the markets to our customer base. We at Ceva, addressing the three major type of customers segments: the small fabless, the large fabless, and the OEM, with the OEM becoming stronger and stronger focus for us also moving forward.
The small fabless are the ones that are dynamic, moving fast, and enabling us to propagate the markets with our IP. The more we propagate our IP, we can basically drive the innovation in the markets and enlarge the accessibility of our technologies. But we have a very, very strong focus with the large fabless. Those are the ones that we are jointly working on the roadmap. They share their needs, we share our roadmaps. That's where we innovate in terms of the new technology capabilities that they need, and they drive scale and volume for us as we penetrate the markets, like the different MCU, like the different vision AI solutions out there.
The OEMs, this is where we see more and more vertical integration by the OEMs, either by software and hardware combination that they want to drive in order to improve the performance and create special use cases and capabilities for their end customers or consumer, as well as much more semiconductor integration to their solution. We're already engaging with, with OEMs, where they have their own internal SoCs, and they want to integrate more of other capabilities into that, and they want to do it internally, utilizing Ceva IP technologies in order to drive that to market. Also, that's where we're addressing directly lots of our software IP capabilities, like 3D spatial audio, sensor fusion, other audio capabilities to go and address that markets. We are very fortunate to have a very large base of customers and across the different so-called customer base, and markets.
Now, let me give you an example why those customers come to us and come again to work and utilize our IP. A little bit insights about the return on investment to our customers. I give you an example of Wi-Fi 7, a technology that we are developing currently. On the left side is what it estimated to be internal developments, not of the full solution, but development of that IP, so-called internally, development and then maintenance, versus what does it mean to buy from us the technology, assuming multi-license deal over five years, assuming roughly 100 million units, so-called volume, just to give some perspective how this model works. We see about 3x-4x cost saving to our customers. With our IP technologies, the more that we do it with the different customers, the more we can scale it and provide the value across.
Similarly to so-called the fab and the fabless domain, in order to enable our customers to utilize that. The other things that we see more and more as we address it with customers, those IPs are complex, but sometimes are quite standard IP, not necessarily the core of what our customers wants to, to develop. But the other things, they need time to market, and also sometimes the market is moving so quickly. So, so-called building that IP every time, every two years, it's very, very non-cost effective or very high cost. And that's where we come in, help them basically to differentiate with what they, what they want to differentiate, and the rest we can offer them with a very, very good ROI on their, on their internal investment.
So here's much more statistics about our customer base that we want to share with you guys today. Over the last five years, we've licensed more than 275 deals, out of which to more than 160 unique customers. More interestingly, and even more valuable, is that we have about more than 100 new first customers in the last few years. But these are the two things that I think talks highly and the most about our IP and the value that we bring to the marketplace, and how we are perceived as a trusted, verified supplier in the marketplace. We have more than 60 customers that are repeating customers in the last five years. They licensed one thing, and then they came again to license either extension of the technology or something else.
And then with that, more than 60 licensed more than one technology from us. And this is very important moving forward, and I, I want to emphasize it even further. We are building a very diversified portfolio of technologies. We are moving from what originally was more a core technology in one core market, to a very diversified technologies and diversified markets and use cases that we can address. This, what will help us to propagate more and accelerate those customers coming back to us. Going back to the mission statement, this is why also to be a very strong partner that is trusted and, co-develop, so-called, with our customers, co-developing with our customers, super critical for our growth and success. And it, and it's already validated in very, in the last five years, and we expect it to expand more in the future.
Here are some testimonies from very, very important customer that we have, we appreciate very much the relationship and the trust that they put in us. Renesas, very, very important partners. They actually have licensed all our wireless technologies, from Bluetooth to Wi-Fi to UWB even, which is a new technology, in the marketplace, to cellular IoT. A broad, broad, broad base of partnership between the two companies, and they are really great partners for us, and we appreciate so much their business. As well as, of course, as I mentioned, as we go from connectivity use case to the infer use case, we are now addressing more with them, the infer in the automotive space... the vision AI type of DSP solution, and so on. The next, moving from connectivity to sensor IP solutions, LG. LG is an OEM. We work with them directly.
They actually have, I believe, really strongly innovated and differentiated the way that you can interact with the remote controller and with the TV. They have so-called, what is called smart remote controller, where you can point, you can move things, all by motion detection. And all the motion detection, all that smartness, basically how to get very accurate positioning constantly without any drift, without any issues, and very reliably, is based on our technologies. It's what so-called, called the webOS technology or their own or their own operating system. They are propagating across all their so-called consumer devices, and that has been very, very great, strong, long-term relationship that we see that moving forward for the coming few years and keep propagating.
The last, which is related to our AI and vision DSP and those things that are moving forward, Novatek, which is one of the top semiconductors or suppliers in the marketplace. We have partnered with them on multi-generation of SoCs, where they're using our DSP for vision AI, and we are looking forward to keep expanding that relationship with them for all the different devices that they're deploying in the markets. Again, very nice testimonial from the different partners that we have, and we appreciate their business across connect connectivity, sensing capabilities, and software IP, and the infer or the AI capabilities that we have developed and we are moving forward, developing even further.
So what that means in terms of some expansions for us towards 2027, and why we believe that we will see a significant growth in SAM, and from there, our potential to drive more revenue. There are three major components that, again, when we look at what we invest in, we always analyze whether that really will drive for us, those three core important ingredients to get to larger SAM or larger potential markets for us. First, are we in markets that the number of devices is going to grow? With the focus on the four key market segments of the Smart Edge and how we are addressing the Smart Edge, we are expecting a very significant growth of number of units, and with that, our ability to charge royalty and to drive our business.
The other thing is really the number of IPs that are needed with each of these devices. With our very diversified set of technologies, we are going to enable more of the multiple technologies within one single device. So I hear lots of your questions sometimes about this technology and the other technology, what will be the royalty for each? In addition to our royalty for each is really the combination of those things, and we'll talk about that. We are basically building more combined solutions to the marketplace. Last but not least is how we drive the average ASP that we can get per unit. This is by the increased complexity of technology that we are offering, as well as, and this is where Michael will cover a little bit more deeply, the platform and the solution that we offer.
At the end, the more complete offering, we have more capabilities driving the average ASP up. All that together is expected to drive the same growth by 45%, as I mentioned previously, towards 2027. I want to switch a little bit here into M&A and how this is related to strategy, a little bit from so-called the organic to M&A and the history, and my, my direction with the team, how we look at that, analyze that, and moving forward. I'll start with RivieraWaves acquisition. That was in 2014. That was originally for Bluetooth technology. We expanded to Wi-Fi technology. That has been extremely successful acquisition, with 20x more in revenue growth comparing to when we acquired the company.
But the key here, beyond the success, is that that was when we focused on wireless technologies, that back then we knew very well, and targeting the same type of market that we knew very well, the mobile and consumer. The second was in 2019, Hillcrest Labs. That was for software IP, sensor fusion, and this type of technologies. Again, successful acquisition of 2x more in revenue and providing significant royalty contribution to the company. That's where we expanded more towards software IP, started working more directly with OEMs like LG that you saw, and enabling more value in the, in the value chain in the market.
In this case, again, it was about technology that we know how to integrate quite well in the same markets, providing new value to our customers, both of them with a pure IP business model that we know how to scale very well. Then in 2021, we acquired Intrinsix. That was for doing design services in the U.S. for the aerospace and defense industry, and that's where the fit wasn't to the same level as the others and what we want to move forward, which is focusing on the pure IP business model that we know to do very well, and a business model that we want to scale across markets and across large customer base and globally. We are very global companies. All our customers are basically global. With that, we made a decision to divest the business and get back to pure IP model.
Recently, we acquired spatial audio software, VisiSonics. This is to augment the Hillcrest solution from sensor fusion to complete 3D spatial audio, working directly with multiple OEMs. I mentioned THX as a brand and a key customer, but that will propagate to many more customers as well. Moving forward, as we look into IP, my goal with the team is to drive a significant inorganic growth as well through M&As. But this is really to, again, going back to the core strategy, which is we deliver IP. We want to deliver a large portfolio of technologies that can scale across similar market segments and customer base. The more technologies that are synergetic with what we have, the more we can offer in the market and create more value to our partners and customers.
Also, I will finish here that our team is already structured in terms of capabilities, in terms of reach, and knowing how to integrate very well different type of IP businesses related to the technologies and markets that we are already addressing. Let me now summarize all that, what that means for us financially moving forward. Yaniv will provide much more details about that later on. If we talk about the megatrend that will drive the volume, our focus on four key market segments moving forward. Our three pillars of technology: Wireless Communication, Sensing AI, and Edge AI within the Smart Edge. We are expecting to grow organically about 1.5x between 2023 and 2027.
I mentioned that we look and we have Iri and a new office working on strategy and M&A to look into other opportunities out there. Overall, I would say that the IP market, within that market, you can see a lot of different assets that they have hard time to scale. Scaling in IP is not necessarily that easy, and there are more opportunity for consolidation in that market, which I think will help the IP providers and honestly, as well as to our partners and customers, because they can get more from a one single place that is trusted and verified as a good partner. But in addition to the revenue growth, very important for me and the team is to be very strongly focused on how to drive operating margin growth. So-called making sure that we're investing the R&D where we see the scale.
Scaling volume, scale across technologies, and scale across customer base and markets. With that, we are targeting to reach about 10x of operating margins from where we are today by 2027. Yaniv will share more about so-called the model itself, and we can discuss it later. Before I finish, I want to summarize several things about really where we were and where we are heading as a company. Historically, and how we enabled the growth in the early days in the marketplace was about the mobile handsets markets. Today, we are really a different company. We are addressing the Smart Edge, very broad wide of market segments and use cases. From a use cases, we have been more focusing in the past on wireless cellular technologies.
Today, we have the technologies in connect, sense, and infer, and that's what we will want to look and augment more technologies as we develop organically or inorganically. We have been selling, and that is still core and strong asset that we have as a DSP technology, but we have expanded way beyond DSP. To some degree, what I found out when I go to customers, that we are still perceived more as a DSP company than the broad wide of products that we have. Clearly, we have very good assets in wireless, sensing, and Edge AI. Edge AI is a big frontier for us moving forward to drive additional growth. With that, we changed the URL from Ceva DSP to Ceva IP.
My vision with the team is to build a strong IP company that is diversified across the key markets that we target, across the use cases that we are already addressing, and making sure we're creating a profitable growth as we're moving forward. So just to summarize that, we have the market tailwinds in terms of more connected devices and data-driven era. We are already addressing, and the growth smart edge market. Those four markets are the ones that are really will grow and help us with the tailwind. Diversified, as I mentioned, growing customer base in our journey, and that will continue, as well as innovative technologies that are very well diversified.
The more we diversify and have those capabilities, but using the same core engine, and that's what Michael will talk a lot about, how we use the same core IP across, we will drive more and more the profitable growth engines that we are looking forward. With that, thank you very much, and, I'm open for Q&A. Yeah, please. Yeah, it's over there. I think they are coming. Yeah.
Thanks, Amir. Thanks for the presentation. Suji Desilva , ROTH MKM. You talked about three vectors of royalty growth, units, content per device, and ASP. Of those three, which do you think is the most potent in the next year or two, three? Just to kind of get a framework on your vision here.
I would say the first two are the largest driver first, but it's really the mix of three. But the average ASP, keep in mind, when we talk about the average ASP, it's not necessarily that a technology itself only going in ASP, which will help, which will happen as well. It's our ability to go to a more sophisticated type of a solution, even from going from a Bluetooth to Wi-Fi to 5G, as well as build, creating a platform.
Should I hijack my question?
So, yeah. So out of the three, the two are the more dominated, but overall, it's really a mix of the all three. So we are going to see growth in number of units, growth market, as well as how much we are delivering in terms of volume. We are going to see also more of our portfolio within one single device of our customers, so-called delivering much of, more technologies. And this is something we see again and again and more and more. Customers that use one of our technology coming to add more technologies that are fully integrated into one single SoC, and then of all the complexity of our technology that will drive up the average ASP.
Then one other question on the inorganic element. Of all the technologies you have in place, and it's quite a few, are there any you perceive to be holes in the marketplace versus your strategy overall? Or do you feel like you have the pieces you need, by and large? You wanna scale more things through, so I'm wondering if there's any kind of obvious, kinda gaping holes there.
So the inorganic is really split between two. There are so-called, I would call it the tuck-ins, where we see that we have a gap in our technology portfolio, and then we add it. For example, what we did with 3D spatial audio, we had sensor fusion. We want to combine this one solution. Honestly, we already partner and deliver that to some of our customers. We want to plug that in on the whole solution and provide it as a value-add technology to our customers. There will probably be in the future also some kind of so-called tuck-ins to make our technology portfolio more complete or more comprehensive. But on top of that, we are looking for more significant acquisitions to add additional technologies within our portfolio to scale up. As I mentioned, I think the IP marketplace, there is a need for more consolidation.
We are well positioned for that from many, many different aspects, and we really know how to take multiple technologies to the same market segments and the same customer base and deliver that value. If you look at even that marketplace, the strong IP suppliers are the ones that are adding more of those capabilities. When they go to the big guys, to the big customers out there, they don't come with one technology, they come with a portfolio. Okay?
Hi, Jeff Bernstein, Silverberg Bernstein Capital. Just on the satellite opportunity you talked about, I think you referred to it as ground station. But, is there an endpoint, a part of that, and what's the ASP like? Is that closer to a base station kind of opportunity?
So first, the communication is between satellites and the endpoint, and the idea is basically to provide the technology in both ends. With that on the ground, that will help to propagate much larger volume. So the idea is to provide ubiquitous, so-called WAN connectivity through the satellites. We see some of those satellites players moving from low bandwidth, just a so-called SOS type of communication, to provide you a really good broadband communication no matter where you are. We see those customers investing in that. With that, they need the next generation, so-called 5G Advanced wireless communication between their edge devices to their satellites. This is a great opportunity for us.
In terms of the ASP, I'm not going to go specifically into ASP, but it's definitely more than the lower end of the scale that we are selling. It's not necessarily the on the level to micro base stations, but close to that. Okay?
Thank you.
Hi. Is, is there a relationship with on the satellite side between having your DSP or your IP up in the satellite as well as on the, on the terrestrial device, whether it's edge or mobile?
So I'm not sure I got your question clearly, but the technology, the DSP wireless communication is a platform. Actually, in this case, we are providing more as a platform. Michael will talk about it. It's not just the core DSP, it's the full encoder, decoder, and the software stack that come with that. Because again, the difference with those players, those are not the smartphone mobile guys that have done it for 20 years, and they are so-called internally consolidated, and they know how to develop this technology. Those are new players that are not experts in 5G and 5G Advanced. We provide a more solution. That solution definitely go to the satellite infrastructure side that they need, and potentially the devices that they want to deploy basically to so-called end users and so on, right? Through a partners.
Right. So, your opportunity is more up in the satellite-
Uh-
initially, and then later at the ground station, or?
It's really both, but if you want to put timeline a little bit, yeah, at the end of the day, they need to make sure it's going to work on the satellites. But definitely it's both, 'cause they are coming with those end devices. They don't want to rely on external partners to deploy that in high, high volume. Not high volume as mobile, but still high volume. Yeah.
Hi.
Yeah.
Hi, Kevin Cassidy-
Yeah.
- from Rosenblatt Securities. Just, is there any change to your strategy for the split of revenue between licensing and royalties? And then along with that, are there... As you go to more complex, does the royalty rate go up? Is there any upside to-
... Yeah, so definitely our expectation is that royalty will go up. As I mentioned, both by the product mix that we are offering, as you mentioned, the complexity of the solution. Good example is just what I was asked. Instead of developing and providing the DSP only, we are providing a more, we call it PentaG, platform of DSP, encoder, decoder, the whole stack, and the software stack. That will elevate the ASP versus so-called just the core DSP. What was the other question that you had on that?
The split between, licenses.
Ah, yeah. I wouldn't say that our business model is changing such that there will be a dramatically different split between the two. Both, we expect them to grow. I would say with our software asset that we have, and we'll talk about that later, the potential of our Wi-Fi growth in terms of royalty, because we have lots of technology that haven't matured yet in terms of volume penetration, like Wi-Fi, like the IoT, so-called industrial IoT that we have. Those are the things that will drive more the royalty growth. But generally speaking, we're expecting both to grow as we move forward.
Hi, this is Martin Yang from Oppenheimer. Can you elaborate on your comment on OEM becoming a stronger focus? You know, do you need a different set of go-to-market to expand on your OEM opportunities?
It's not different sets of go-to-market. Definitely it's a different set of customer base. The focus is already started a few years ago, and I'm basically focusing the team and putting more efforts into that. It started with the software application that we are offering to our OEMs. But really what in addition drives it and where we see lots of potential and value, is those OEMs are more becoming vertically integrated with their semiconductors needs and want to integrate it. So we have multiple partners that are coming to us and say, "Okay, this is what we want to build. We've used to supply, get it supplied from many different type of players. We want to own it.
Help us with your IP to go and build it as a complete solution." Those are the OEMs, the big brands out there in the consumer space and other places. We know how to reach them. They are within the same ecosystem. In many cases, even when we sell silicon IP to the semiconductor companies, we are still working so-called with the OEMs, to talk about roadmap and other things. Like, for example, automotive, we talk with the car makers and so on, so-called more, even if we don't sell to them. So I wouldn't say it's a new go-to markets, but it's definitely more market opportunities and stronger focus for us moving forward. Yeah.
Great. Taking a question, Gus Richard.
I don't think they hear you. Yeah.
Now, is it working?
Yeah.
Cool. Gus Richard from Northland Securities. An adjacent IP to your wireless basebands are the transceiver. A lot of companies have trouble with that. Is that something that you guys are working on, or is that something you might acquire, or do you have partners for that piece of-
Yeah, transceiver, you mean the RF IP? The RF—you mean the RF part, the transceiver? Okay. Yeah, very good questions. Let's start with where we are today. Today, we are not supplying, and we don't have internal RF IP or RF, so-called technology that we offer to our customers. Today, we partners with different partners in the marketplace, and that's definitely a potential expansion for us. I was asked about, you know, how—what are the gaps in our technologies, and do I see more additional capabilities there to go and augment that? That's one option. We are looking into all those things, but that's definitely something that we would consider moving forward.
What, what's the bigger opportunity for the company? Is it, moving customers from a make to buy philosophy, or is it competitive displacement?
Very good question. It's honestly, it really depends on not necessarily the type of customers, but where they are and what they are looking for. Okay? So when I talk, for example, on those OEMs, it's, it's less about with competition, it's more about how they internalize that, that's technologies. When we look at the semiconductor companies, and I would say first, in some cases, we deal with the make versus buy, and that goes back to the slide that I showed, the ROI, number one, and then how core and how much they want, so-called, to uniquely own this technology internally. And honestly, sometimes a little bit their own internal NIH and how they allocate resource and all that. So that's definitely there. In technologies that we dominate in terms of capabilities, it's mostly make versus buy.
In technology, there is more competition, in addition, it's competition landscape and decision, which comes with it. Yeah.
Okay, so if there's no more questions for now, we'll take a quick break. You can grab a coffee, whatever you want, and we'll come back in about 15 minutes or so, and Michael Boukaya will be up next with the innovation section, talking about R&D. Thank you.
Hello, one, two, three. Do you hear me? No?
We're here.
No, stop. Now? Perfect. Richard, let me know when I can, I can start.
Go ahead.
Yeah?
Or you can.
All right. Right. So good morning, guys. I'm very happy to be with you today. My name is Michael Boukaya, and I serve as the COO for Ceva. With Ceva from the very beginning, for many years, one of the founders of the company, and heading technology and engineering in Ceva. So following Amir, I would like to focus today how our technology portfolio is unlocking the next generation of smart edge devices. And the way we structure that into three main pillars, one is connect. Connect is the ability to give to these edge devices to be connected everywhere with seamless wireless user experience, with very high data rate, super low latency, meaning you don't want to wait for a prompt on ChatGPT.
You want things happens now without latency, with very, very long battery life, meaning very strict budget for power consumption. So in that front, we managed with the team to be and become number one in the wireless IP. And we have today a comprehensive portfolio for all wireless technology, starting from connectivity technology like Wi-Fi, Bluetooth, and we recently added UWB. And beyond that, all the cellular expertise that we had for years, and we add more and more capabilities and new innovation, looking for 5G, 5G Advanced, and beyond to horizon of 6G. So all this portfolio and the way we come, very holistic, positions us as really number one in the market today to lead and enable Smart Edge devices to be connected in such a way, from very tiny connections, from few megabit per second, up to multi-Gbps .
I will share to you more detail on that front. The second one is sense. Smart Edge devices need to offer much more, the next level of immersive user experience, meaning the way to have a more natural experience surrounding you, more contextual awareness. The way we go to the market here is via embedded software technology, where we made strategic acquisition a few years ago with Hillcrest Labs, recently with VisiSonics, to combine things together and offer a unique portfolio of software packages to enable really new level of immersive realism for Smart Edge devices. Last but not least is infer. And here I'm talking about AI technology.
Here we are facing a new era of technology coming from classical AI use cases like classification, segmentation, making the edge more intelligent to take decision and predict something, to the era of generative AI, giving the ability to create content. This will reshape our life, and we, we see that with the mega boom with OpenAI, and reshape completely the technology and the way architectures are done for smart edge devices. That's the reason we innovate new NPU technology, stands for neural processor unit, and I will share with you today how we are looking to this new era of AI capabilities and offering scalability, which is extremely, extremely important for us.
As Amir said before, how we leveraged one single IP development and technology over multiple use cases, and give very high scalability from very low-end use cases with few TOPS, up to one- more than 1,000 TOPS for very high-end use cases in, for example, for ADAS system. But to be really, a leader in IP, there is a DNA. The DNA is to really predict and anticipate the future needs of our customers. Meaning that the way the time zero of this, of our customer design, start the design, we need two years, even three years before, to be in the position to predict what will be the key pains, the key issues that we want to solve. And that's the reason we put in place in our three segments, in wireless, in sensing and AI, research labs.
With very talented guys, but what I call the forward-thinking about identifying the key, the key trends and pains of our customers. That's number one. Number two, you don't want to make a bet, but you want to validate your trajectory. What we created with large customer base, that we have a very deep relationship with the top guys in the industry, the big, the big guys that drive the market, that shape the market, and we have this relationship with our customers to collaborate together and make sure we are on the right, the right direction. In this way, the combination of our ability to predict and to anticipate the future needs way far away from the design start, enable us to come and to do the right things.
Number three in the DNA is, of course, to be able to execute and deliver with a high quality, on time, and that's enable us to continue a very long relationship with our customer, to come and make the integration, to escort them during the shipment. Our goals by the end of the day is to go to shipment, to create volume, so we support them, and in that way, we can collect more royalty for the future. Taking that into consideration, taking the three pillars in our technologies, the wireless, the sensing, and the inference on AI, we are leveraging and combining a unique portfolio to leverage our investment to come to many markets, to many markets with many customers, and have the scalability to go up and down in term of efficiency or high-end, or high-end high-end use cases.
This enables us to really leverage on our investment and to come to many—the key markets that Amir just related before. Consumer IoT needs connectivity, needs sensing, so we have the two, combine them together and create, and create value. This one does not work?... The projector is not using the email, so no. Not a question. Here we go. Great. Thank you, Richard. All right. So, I would like to start with our first pillar on connectivity technology. When we look to how the future for Smart Edge in wireless opportunities and, and, and edge, we see that the next generation of connectivity standards, for example, moving from Wi-Fi 6 to Wi-Fi 7, will come with a very, very big step in term of how the bandwidth we go up with 4.8x data throughput.
This means enable you to go and transmit 46 Gbps in Wi-Fi 7. This is enabled by the way the technology is giving multi-link operation, meaning that today on your AP, or access point, you can select on 2.4 GHz or 5 GHz. Wi-Fi 7 is offering you the ability to transmit over multiple bands at the same time, at 2.4, 5 GHz, and 6 GHz, giving much more capacity. The modulation of 4K QAM giving much more bits in the transmission. Overall, offering to the users 46 Gbps. So moving from the two standards, a huge step of 4.8x in more throughput. And nowadays, it is really urgent because we all need more data. Home office is a fact today with all what we experienced in recent years.
Users needs much more, you want to internet in a fluent way, virtual collaboration between people, large data files without any latencies. Wi-Fi 7 is a must for the next generation. But for us, it's not only about technology, it's also about new opportunities, creating more, a new wave of devices. It's not only for smartphone and PC, but everything connected at home, like gaming, audio system, TVs, et cetera. So Wi-Fi 7 is a key core technology update for offering new opportunities for us. Same come from Bluetooth technology, moving from Bluetooth 5 to Bluetooth 6.0. Bluetooth connected in what consumer IT connects everything. Now, giving the ability for Bluetooth standard to support 8 Mbps, with much higher modulation of 16 QAM, open new edge and new wave of use cases.
Imagine your earbuds now are be able to, to come and transmit much more data with lossless audio, high definition audio on your earbuds. This create a new, new wave of new earbuds, headphone, gaming, gaming headset, et cetera. So overall, this big step of these two guys, which are the two main cores in the connectivity world, are opening us many new opportunities, and we believe that we are really, really well positioned today. I think with the perseverance over the years, acquisition we made, the team that we have, we have managed to have, to manage to have a very good partnership with more than 100 licensees. And the market validate us with very nice volume of more than five billion devices CEVA inside. And these things are possible when you invest and you persevere in investment.
You know, one of the thing and one question I, I heard today, how we come versus the competition? To provide a Wi-Fi 7 solution in the market, you, you need Wi-Fi 4, Wi-Fi 5, Wi-Fi 6, Wi-Fi 7. You need all the backward compatibility. For each one of them, you need to design, optimize, certified, deliver, go to production, and make sure you are carrying on IP. It's a... We put here a huge investment to get all these backward compatibility in-house, and this created very, very high entry barrier for potential competition. And the value that we have today with this investment is the holistic portfolio that we have on the all the technologies, on Wi-Fi, Bluetooth, UWB, and Zigbee, and the ability to combine things together. When you come to an consumer IoT-...
Our customers need Wi-Fi solution plus Bluetooth, they come to CEVA. Or you can imagine a combination of UWB and BLE, they come to CEVA. No, no reason to go someone else. So the fact that we have all the combination, all the portfolio in-house, create a real power for us to deliver. Beyond connectivity, let's have a look how CEVA plays an important role in cellular technology. We are in this market for years, and let's take a look to the evolution of cellular technologies, starting from 4G with LTE, that puts strong foundation for with OFDM technology inside, and the transformative leaps of 5G and 5G Advanced, that offers much more bit rates with 20 Gbps and reduce the overall latency of one millisecond latency.
Looking to the horizon of 6G, it's an amazing capability and capacities with 1 terabit per second, and take a look to the latency, under 1 ms latency. So what does it mean for us? In the first beginning, 4G really targeted mobile phone, smartphone. 5G and 5G Advanced, and looking to the horizon of 6G, are operating much beyond smartphones, much more use cases. And Amir related to some of them today, 5G Advanced, for example, the ability to create 5G technology on satellite, and that's applicable with, with your question for the base station and the consumer and the handset itself. So it's, it's a big opportunity on the two sides, and we are targeting the two sides, point-to-point, as we made for base station and mobile, same strategy, and I will explain later how.
But other use cases, like vehicle to vehicle, vehicle to network, vehicle to pedestrians, that needs super low latency. You cannot, you cannot have low, bigger latency when going from such, such systems. And to the future for 6G, the ultra-low latency, the so-called eURLLC, Extended Ultra-Low Latency Communication, will really unlock the metaverse vision. That's today is starting, but needs really strong cellular backbone to enable that. So to enable these new opportunities, new use cases, we have two pillars in our portfolio. Number one is our DSP. We delivered Ceva-XC, which is a famous product line for baseband processor. We announced this year the sixth generation of our XC technology. Why it is so important?
Because our customer need programmability for future-proof, programmability to differentiate, and also scalability to come with different use cases, from low-end use cases, like massive IoT, up to mobile broadband with very super high-end data rates. So for that reason, we crafted a new, innovate a new solution on XC20, which is designed really for advanced node, achieving 2.5 GHz, which is super high frequency in this node. And we improve the performance 3x compared to previous generation, giving the ability to support new standards, new use cases with special instruction, special mechanism. And we are proud that-- to say that we are the first in the market that introduce dynamic multithreading in baseband processor. This feature is critical to achieve super high efficiency. Our customer cares about power, our customer cares about cost and die size.
Having multithreading mechanism inside create and enable us to achieve very high efficiency in these two fronts. Last is the scalability, and here again is the way we invest and leverage our investment in R&D. We are not create one product for one use case, but we are creating a family of DSP. XC20 is a family, architectural family, that can go down for very low-end use cases and give an optimal efficiency and up for very high, very high performance. The second pillar for that is the ability to help our customer to go to the market. These new use cases like V2X, like RedCap, reduced capacity modem, like satellite, satellite modems. There are new, many new players and newcomers in the market. The fact that they lack wireless expertise, they are not the big guys like Qualcomm, like MediaTek.
They want to integrate IP to save cost. They are not going to Qualcomm because it's very expensive. They will not be ready to help them to customize their need. The natural way is to go to IP, and here it's not enough to deliver one single component, like a DSP, but we need something much more comprehensive. That's the reason we introduced the second generation of our mobile platform, PentaG2, which is a platform composed of our backbone technology of DSP, the XC20. But on top of that, we deliver a comprehensive platform, including 5G Advanced hardware accelerators, AI, NPU, which is completely brand-new technology for wireless, that we introduce in our mobile platform. We are also providing on top of that software, which is qualified, and ability to make simulation in FPGA.
All that together is a combination of very strong portfolio and offering that can enable our customer to reduce, reduce risk and time to market, since we integrate everything, we qualified everything.... It's a modular architecture, so again, in term of investment for us, we invest in one platform and address multiple use cases, from RedCap up to satellite or O-RAN, O-RAN use cases with this platform. So overall, the performance is also scalable for massive IoT, with few megabit per second, up to more than 10 Gbps . So if I summarize the wireless pillar, why we win, there are four main reasons for that. First, our perseverance.
We've invested in this domain for years, the first beginning, and as I said, the compatibility, the way we go one standard after another, one risk after the other, and make another layer, make us really in a very strong position in terms of entry barrier for other competitors. Second point is time to market. It's extremely important for us to come ahead of time and ready for our customers to be there for their design. And timing is super, super important. I want to say something here. Even equipment, but test equipment for wireless standards are coming to us, with us, and working with us to test their equipment because they know that we are mature and we are ahead of time. Number three is performance, best-in-class. You cannot mislead here.
We need to be careful about ultra-low power and highest performance. These are two KPIs when I run the team. It's in front of us. Make sure you are not above what we need to do. Ultra-low power and highest performance is super important, and completeness, as I explained before, is the completeness of the portfolio. It's not only Wi-Fi, it's not only Bluetooth, it's the ability to have all the technology in-house and offering something comprehensive that can serve multiple technologies. I'm moving to the next pillar on sense. New Smart Edge devices need to serve today and deliver much superior, much superior sensing and immersive realism user experience. This needs much more sophisticated technologies to enable that. We are categorizing that into four main cores.
One is spatial audio technology, giving you the ability to have 3D sensing and sound experience around you, which is important for earbuds, gaming, AR/VR devices. It's a key parameter, and our customers and customers are willing to pay to have these features, a must-have. Number two is motion tracking, the ability to track movements. It isn't just on a cursor, it is your head, like head tracking. And think about earbuds, to wireless earbuds. You need these two technology together in order to track the head movement in six dimensions in the space and giving this spatial audio. The combination of the two is very, very valuable. That's the reason we made these strategic acquisitions of Hillcrest Labs, that gives us and contribute the motion technology, and VisiSonics recently, that gives us the ability to add spatial audio.
The combination of the two give us the opportunity to give and address multiple use cases again. Last but not least, is AI voice assistant and voice clarity. Voice technology is critical. Voice assistant today at home, super important, and it's deeply, deeply based on AI because it started from very simple command like, "Hey, Siri," "Hey, Alexa." That's all, okay? But with much more sophisticated detection and recognition of much complex sentence and commands. So you can imagine why AI is so important here. And last is the voice clarity, is the ability for, for example, smart speakers to recognize the user in a very noisy environment and to beam form on someone, and to really understand what you say, and to cancel the noise, to have very clarity of your voice and your command.
So these, for us, are the four core, technology that enable our, our partners and OEMs to differentiate and giving more value, more immersive realism. The way to go here, the go-to-market strategy for us here, is to go with software. Software, embedded software versus, standard IP silicon development, and we go directly to the OEM, enable the OEM to differentiate on this platform and, and deliver. So if you look to the portfolio that we have, we have all the pieces. The RealSpace is a product that gives immersive 3D sound, with head tracking together with the motion tracking from Hillcrest Labs. Today, we have the motion engine ability to track everything. We have voice technology with voice clarity of ClearVox and WhisPro, that gives all the voice, assistance and clearance, as I explained before.
If we look why we win, there are a couple of key reasons. Well, number one is a flagship spatial audio technology. It's really, really, really demanding. With our OEM insist on that and with VisiSonics acquisition, we are able to show and demonstrate very high quality. But it's not enough. The combination of technologies together, as I explained before, is a very valuable to our customers... OEMs want to have one-stop shop for everything. For example, take an earbuds, a TWS players like boAt, for example, they come to us for one simple reason: we offer all the portfolio and all the technology in one delivery. The spatial audio with motion tracking, the head tracking, plus the voice. The voice technology gives a full solution for a headset device. Last is the quality.
This must be qualified to go to production very quickly. Remember, the life cycle of an OEM is much quicker than regular silicon development. We deliver our software technology, and they go to shipment very, very fast. So it must be qualified, and today we have a track record of more than 300 million devices shipped with our software. The third pillar, and that's a very, very important growth engine for us, is Edge AI. When we look to the trend in the future, I would like to spend some time with you to talk about how we see the future with generative AI. Generative AI is the next frontier and the catalyst for the evolution of Edge AI. The first beginning, generative AI starts with the cloud, with, and that's the reason NVIDIA was so successful and is still very successful.
But when you look to the future, the workload on Generative AI will be hybrid and will come to the edge, and there are very good reasons for that. Why coming to the edge? First, economic reasons. An inference on, with generative AIs, and for example, you ask the Google, is 10 times more expensive than a simple, simple ask and inference. 10 times more expensive. So now, if you want to scale up with million of users and million of accesses to the cloud, it's become more problematic. So one reason is a cost reason, having a low-cost reason to decongest the volume of workload on the cloud and moving on the edge. Second point, which is super important, is privacy. You don't want your data, you don't want your prompt to be available for everyone. You want to keep your privacy.
You want to generate things, which is, you know, generative AIs give you the ability to create new contents, create text, create image, create sound. The sky is the limit, but you don't want that to be available, and you want to share that everywhere. Second, and the second point is lower latency. You know, you make a prompt, you don't want to wait for one minute to get an answer to get something. You want not to go to the cloud; you want to have your answer very, very quickly. But if we look at that, these are the drivers, the opportunity is here. 2.7 billion gen edge devices are expected to be in the market in 2027, and 30% of the inferences will happen on the edge. So that's a great opportunity for us.
But here are the challenges. On the other hand, an inference with generative AIs is 30x more consuming in term of battery and power consumption. On the compute, it's 10x more complex than a classical inference, like a vision, vision inference or audio, NLP inference on the edge, 10x more complex. And last, but not least, is the cost. When you have such network with generative AI, classical AI networks needs 100 of parameters, let's say 5K parameters. Now, generative AI on the edge today consumes 1 billion parameters. When you look to the future, we talk about 10 billion of parameter on the edge. So there is a cost issue, and we need to treat that. So these, we have, on one hand, a big opportunity, but on the other hand, also big challenges.
So how, how we treat the problem? And here I want to say something super important. Classical AI architecture like GPU on the edge or CPU plus something on the edge, or even NPU, neural network processor, the classical architecture will not fly anymore to treat such big problems. We need to rethink, we need to recreate the architecture. We need to renovate it, and that's... And when I said before, we need to anticipate problems to be on time, our research lab, our AI research lab, already understood this paradigm more than one years ago to treat this problem. And today, we are proud to say that we announced in August this year, a new NPU version of our NeuPro-M to treat and to be the best Gen AI NPU for the edge.
For me, the best things is need to be measured with some KPIs. Number one is performance. Performance mean how many token? Token is a word. If you put on ChatGPT some prompt, is how many words you can compute at the same time, okay? And this impose a latency. You don't want to wait for two minutes to get the answer on your prompt. You want to get it one second, no more, right? So for that, you need to proceed a lot of token per second. Today, our NeuPro-M is able to calculate 3.4 million token per second. KPI number two is energy, power consumption. You remember the problem. We invested a lot of intelligence and innovation to create super low power design and the way to reduce dramatically the compute and the power consumption...
Today, we achieve 1.6 million tokens per second per watt, which is a metric of efficiency in energy, which is today with our, we scan the competition, the best-in-class efficiency in power consumption, looking to the landscape. Last but not least, is the cost saving. You remember 10 billion parameters? We need, and we will show you some of the techniques in my next slide, to reduce the cost dramatically, and we put in place techniques, very advanced techniques, to reduce the footprint reduction with 4x reduction. Last, I want to talk about scalability. Again and again, when we develop our IPs, we think about scalability, how we can leverage over multiple use cases, from simple use cases to very high-end use cases.
That's the reason we provide in our new NeuPro-M NPU scalability to go from 10 up to more than 1,000 TOPS. The second part in the problem to resolve it is not only a hardware issue, but mostly, if you ask AI users, is software issue. And if you don't have a full stack, comprehensive stack, AI available on your NPU, it's useless. So that's the reason we deliver IP with a full stack AI, with ability to make system optimization, model efficiency, and retraining the network to go and to be much, much smaller. The ability to make a very good graph compilation of the networks, and last, to run and infer on the hardware. Here are some details how we come and treat the problem, our approach.
Again and again, it's years of R&D and research to be really, really fit the problem and understand what are the real issues when we talk about gen AI networks. First is the ability to have unstructured sparsity, means to take a full network, understand where we have a lot of zeros in a network, in weights and data, and just remove them in unstructured way, meaning that there is no really good rule where the zero are located. Randomly located, our hardware is able to identify where the zero are, just remove them, make the network much more compact, reduce the compute, reduce the power.
We have put in place very advanced system architecture to enable parallel processing, to let all the units in the NPU to work in parallel, not to waste time, not in serial way, to get to the performance I discussed before, to treat huge number of tokens in parallel. Last is to understand what is the topology of the network, what are the basic components in the network when you talk about gen AI. Transformers, the transformers are the new type of elements which need an operation that you need to treat. Comparing to classical AI networks and operation, which is more convolution. Convolutions, yes? This new gen AI is using transformers. So we designed the hardware to be super efficient for this new type.
The last but not least in the feature set is the comprehensive optimized mixed precision ability to take a network in floating point and to quantize it to have integer with 16-bit, eight-bit, four-bit, or to stay in floating point or be floating point broken floating point with different types. The ability to have this mixed precision gives us the possibility to reduce the network, to reduce the complexity, reduce the power and the compute. And that's the bottom line. We achieve very, very network utilization. That's the key benchmark. 95 utilization on network, and we compare ourselves to serious competition, and you can see here how efficient we are in power efficiency with RoBERTa, which is a large language model.
That's the term very well used in gen AI networks. RoBERTa is very well known, 25 more efficient in TOPS per second per watt. But also in classical AI use cases like ResNet-15, 35 more efficient than the competition. So overall, we have very high confidence that with this new generation of NPU, we will be able to conquer and to capture more and more value on edge device. So just to summarize where we are, as you see, that we have a very nice portfolio in our three segment. In infer, with a combination of NPUs, with DSP technology, in sensing, with software packages and connectivity, with and with connect, connect cellular and wireless, and wireless technology.
During 2023, we upgraded nicely our portfolio with giving a very good focus on how we can leverage these new developments for multiple use cases for all the different segments. And we are really committed to continue to innovate looking forward. So stay tuned. We are going to announce in CS important announcement for different portfolio, new members in our portfolio and Mobile World Congress for cellular technology, since we are recommitted to Smart Edge innovation. So all together, we start from innovation to full solution to create more value to our customers, starting from the ability to innovate, to anticipate.... with the right innovation to bet on the right things, with deep relationship with our customer, giving solution to the increased complexity of the system that we need to solve today.
If it is in AI, if it is in sensing or wireless technology, but create more and more complexity, and going to the market with a comprehensive portfolio, not one single component, but really something holistic to go to the market and address, and address each segment with not one component, but a solution level, allowing us to give a much more nice offering, including combos, technologies, software, and baseband platform. This means higher ASP. This means the ability to capture much more value in our, in our business. Thank you for your time, and, we pass that to Yaniv, and we'll be, really happy to answer your question later on.
Okay. Thank you, Michael. Okay, I'll start. So, good morning, everyone. Great to see you and see the faces and get back to New York and to this type of event. Obviously, as Richard said, time flies, and five years, less a month, has gone by from the recent or the last first Analyst Day presentation that we have done here, same room. And it's. We wanted to do this much earlier. We wanted to do it sometime early in 2022, but then with management changes, it didn't really make sense to do it with Gideon talking about the future, and then we needed to give Amir some time, and you heard him earlier today, to learn, understand where we are and really make the decision about the next steps and the focus.
So we're back here today. With that said, I do want to recap 2019, what we had in mind there and what we achieved and didn't achieve by the end of 2022. We wanna talk about some interesting, different perspective and different way that we are showing today the revenues, and we have done some work around that. Talk about and explain what Michael said, all these R&D efforts. How do they work out on a, on an IP business model? Talk about the product life cycles translated into the financial models and obviously, capital allocation in ESG are topics that you're all interested in. And then I'll summarize and then talk about future models as well. So in 2019, January, five years ago, we talked about three main revenue targets.
We talked about licensing revenue growing 10%-20% of a $40-ish million level from 2018. We talked about royalties doubling from a $40-$45 million level, and we talked about unit shipments of three billion devices by 2022. Obviously, with more royalties and more units, the gross margin plan was to increase significantly, to double around 30% non-GAAP, and the bottom line to triple as well from the 2018-ish non-GAAP levels. That was the plan back in 2019. Let's share with you what worked and what did not work as well. So on the licensing front, we did manage to increase the number of deals, got over 60-ish. Again, CEVA standalone without the Intrinsix business, so that 50-60 new deals a year was achieved.
The Bluetooth and Wi-Fi business plan exceeded our expectation. We had a lot of very successful deals that we shared with some of you, we shared with you earlier, today. We managed to grow licensing much more than the 10%-20% that we had in our initial models. We doubled, doubled our royalty growth from the base station IoT. So on that part of the targets, that was met, and we did increase the volume quite significantly. Back in 2018, we ended up with shy of one billion devices. 70% growth there got us to 1.7 billion devices by the end of 2022.
Last but not least, and we talked about it today, very nice, and a successful acquisition of Hillcrest happened in 2019 after the Analyst Day, doubled the revenue so far. So this has worked out, extremely nice for us, in the last four years. What didn't work out? Obviously, handsets. The handset market, undertook significant consolidation. There was a lot of changes in the industry. Obviously, Apple had its issues with its, suppliers, Qualcomm, Intel, and that changed the plans for us and, and for others in this industry. Some of the markets we talked about today and back in 2019 took longer to, grow.
We're talking about cellular IoT, 5G, base station, automotive, industrial, the whole market, some here and there customers, but the overall market was slower. With any Intrinsix acquisition that Amir talked about earlier, had its toll, both from a profitability and a focus point of view. With not having enough royalties to back up the story, obviously, the operating margins and EPS were missed, and they were not in line with the targets we've put at the beginning of 2019. I want to show you now the focus that we have talked about earlier today of a pure IP play. One of the easier examples to give is Arm. Obviously, Arm is public again after quite a few years, seven years of being private.
What we said a decade ago and what we could today discuss and show investors is that the CEVA business model, obviously much, much smaller, is identical to Arm's. It's a licensing and royalty business model. I'll talk about and give you again the high side of how this works for us and how this works in the industry. On top of that, this is something that we've added that Arm does not have, is software. It has a little bit of different characteristics than the traditional core technology licensing business. Today, what we call hardware, these are the cores, these are the different technologies that Amir and Michael talked about earlier. We are managing to run four to five different R&D programs simultaneously a year. It is a long design cycle.
It still takes a year to two years to design something from scratch. It's shorter if it's using the same technology, like Michael explained, and enhancing it. Then we have the next phase of taking our IP, offering it to different licensing companies, semiconductor companies, sometimes OEMs, that integrate that IP into a chip. They design a chip. It takes some time. Today, we're talking about 50-60 new designs every year. And the next stage of that is going into production and enjoying a royalty rate. If you look at the concentration of royalty payers today in 2023 versus 2019, at the time, we had a concentration of the top five royalty payers contribute north of 80%. Today, we're about 50%.
So it's a very nice expansion and diversification in markets, in technologies, in royalty payers, we're about 80 versus 60, four or five years ago. And we have north of 100 customers today in different segments with different technologies that are working on their development or ICs and chips that potentially could go into production over the next couple of years. So that's the traditional IP business model. This is something that we are very focused in, and after clearing up the service business, this is what we are doing. On top of that, there's an interesting element of software, and here the timeframe for achieving the revenue and the royalty growth is much, much shorter. First, when we start the design, it's less than a year in most cases, of a design of a software piece or package or product.
We could run two to three different projects simultaneously every year. The design cycle with the OEM or with the customer is much shorter, within a few months, up to six months maximum. We talked, Michael mentioned that earlier. The OEMs have a much shorter lead time. They need the product. They want the product to work, they need the software implemented and out to the market. It's a matter of months before we could see the royalties kick in. And royalties, higher ASPs, because we deal with OEMs, and it's a locking mechanism. Those OEMs want those products to have the right set of features. They want to enhance those features every year that the product gets, or every few years, that the product gets enhanced and improved.
And we are there for a multiyear relationship with most of our customer, with most of our OEM customers. So here we have a nice add-on that have started with the Hillcrest acquisition, moved into sensor and the spatial audio with the VisiSonics recently, which adds on top of the traditional IP business model. The next few slides is looking a little bit different... Sorry, one more, one more slide before revenues, and summarizing where the typical IP play is today. So if we're talking about an IP company, about 50%-60% of their R&D is invested or revenues invested in R&D. Ceva is probably on the higher end, 60-ish% in the last two years. The market is probably closer to 50%.
This is something that we will talk about in a minute, that we want to focus and change. We're talking about off-the-shelf reusable technologies. This is one of the advantages of the IP business model, and you saw some of the products that we come up and generate. And when you have a pure IP business model, we are talking about back to 90% gross margin business. This is what we like in the business, this is what investors like, and this is with having no manufacturing needs, but it's pure IP and R&D efforts. This is what generates the revenues and the high profitability. With the royalty streams coming in and control over the overall expenses, the operating margins and the bottom line should improve.
This is something that we are putting an emphasis on, and this is something that we are changing today and have talked about it throughout 2023, is really diversifying our R&D expenditure and markets, but being much more focused and higher ROI in those R&D dollars, and to try to focus those to generate more revenues, reusable revenues, and so on. With having more control on our overall expenses and expense growth being lower than the overall top-line growth, this should cause and generate a overall margin expansion, and this is the plan for us going forward, when I'll share with you in a few minutes, the slides and the forecast for the next four years.
This is one of the more interesting slides, I think, in my presentation deck and today, to understand where Ceva is today, because this is something that on the left-hand side, you know, and have worked with us and followed the company, and this is the total revenue, licensing, and royalties of the company for the last five years. Ceva overall, 11% CAGR from 2019 through 2022. What you probably don't know is that inside that, if we look at Smart Edge, which is the theme of today and going forward, this is the blue. Versus the, our traditional modem business, when we look at the overall business, we have managed in the last five years to more than double it, from $50 million to $103 million of revenues, overall revenues coming from that market.
That segment of the market for us grew 27% from 2019 to 2022. What we were suffering is the royalties and the deals around the handsets, which I talked about in the previous slide. This is what has changed in the overall handset market, the mobile space, and it took the numbers down from where we were in 2019. The Smart Edge enables us, going forward, to participate in a much, much larger market, many more markets than we were focused on in a single solution and a single market back a few years ago. The growth for us is coming from that $100 million, full year 2022 and onwards, with a much more diversified market, customers, and applications.
Another way to look at the markets and the revenue split for us is to consolidate the six different pillars that Michael and Amir talked about into three major ones. One is the consumer IoT, the other is the mobile, and then the third is the industrial IoT. And here, we also, just for simplicity, we added the automotive infrastructure and the industrial. What you could see here, that back in 2019, almost half or 45% of our revenues came from mobile. Four years later, the first nine months of 2023, only 15% came from mobile. And we were able to offset that and to grow the consumer IoT, essentially to double, from 30%-60% of our overall revenues, royalties, and licensing together.
There is a time difference between them, but this is for the whole for the whole use case. On top of that, the industrial IoT is still an interesting market or markets for growth. We kept that at about 25%-ish, 23-24% of the business. Different use cases back then, but this is still an opportunity of growth for us going forward, and we are still investing our R&D dollars into that market. So this is a little bit of a different understanding. What did work out very nice, and this is what we have talked about today earlier and what could grow and drive growth going forward versus the handset market, which is how Ceva started: single market, single type of technology, and much more focused years ago.
In order to get to those revenues, there's a lot of R&D efforts, and Michael discussed about the different R&D projects that we are working on these days. If you want to take a snapshot and look at the last five years of R&D spend, we're at $274 million non-GAAP. Those R&D investments managed to generate or more than cover itself with $285 million of new licensing deals. But on top of that, also the $210 million of royalties that came in in those same five years. Obviously, there is a time lag between the two. This doesn't...
The investments that you do in the same five years usually represent royalties in the years to come, and the licensing also has a little bit of a one, two-year, or three-year type of delta as well. But if we just take a snapshot of five years, this means that the focused R&D efforts that we are putting today do represent and cover themselves both in licensing and with more opportunities into royalties. Three slides in market segments now on different product lines, and how they contributed revenues, and how they sucked in those R&D investments over the last couple of years. I'll start with Bluetooth. Bluetooth is one of the more successful product lines, very mature, that we started back in 2014.
If we look at the first four years of investment years, this is where the R&D dollars were put in place, this is where Bluetooth 4, and then later on 5, was actually developed, some of it even before. We started gaining a good market share, and we started gaining customer base, and from 12 deals, we went to 16 and 18 deals in the first couple of years, and from almost no volume of 40 million devices, we reached 200 million devices a year, single digit market share. And that was the—this was the initial ramp-up of the product line.
When we had a critical mass of customers, when we had a critical mass of use cases, and when we had new technology, next generation technology, like Bluetooth 5, and Michael talked about some of the differences and the different standards that evolved in the different technologies, the numbers started to look completely different. From single digit, we reached 30% market share, and from 200 million, we reached one billion devices, and we have over 100 customers. This overall product line, in the last five years alone, generated $107 million, Bluetooth product line, royalties and licensing. Obviously, it's quite combined. There are in those $107 million, this is just the last five years. There's much more royalties because we're in high volume, and it also represents new deals of licensing.
So this is one of the first successful, diversified products that we added on back in 2014. If we look at Wi-Fi, that started much, much later. The market was not mature yet. The use cases were not there. We started the investment around 2018. First four years, again, of investment, R&D dollars are working on the first Wi-Fi 4, Wi-Fi 5. We moved to Wi-Fi 6 standard, 4 deals, going up to 12-15 deals, single digit, five million units of Wi-Fi going up to 170-ish, and still a single digit market share. This is where we are today.
But if you look at the Bluetooth and extrapolate that, that Wi-Fi is about to start, the maturity with 50+ customers, the market demand, that Wi-Fi is all over the place, the jump, a significant jump in performance and capabilities of Wi-Fi 6 and 7 versus the older technologies and the older standard, that gives us the ability to forecast much, much growth, much faster over the next couple of years, reaching 25% market share, maybe somewhere in the neighborhood of 1 billion devices with much higher ASPs, as these are much more expensive, sophisticated chips. So if we look at the, where we are today, and this is just the initial part of the Wi-Fi opportunity for us, we're at $75 million of revenues, licensing and royalties.
But because we're still in the initial stage, most of it is still coming from licensing activities and not yet the royalties. So from a much, much younger product line versus the Bluetooth, we're $75 million versus $107 million, but still most of it in licensing, there are still a lot of interesting opportunities for growth in revenues from the Wi-Fi product line. Another third example of a separate product line that we talked about earlier, the family of the XC. XC is a very old type of product. We started out a decade ago, but we have continued to evolve. It's the same platform with different features, with different add-ons, with different adjustments in order to win different market segments.
With that same platform, we started with the basic 4G modems, went into 5G, base station, cellular IoT, satellites, drones, vehicle-to-vehicle communication. One of the advantages that those same R&D dollars do not need to jump from one market to the other and from one technology to the next technology, but it's in the same platform, the same base, just using and extrapolating it into different markets and different customer base. $65 million of combined revenues coming from that in the last five years. Richard, by the way, it doesn't work the timing, so just give me the time. Thanks. Moving on to capital allocation. So as you heard today, the most important focus today in the company as a pure play IP is M&A. And M&A in the right place, M&A in the IP core, segment.
We talked about earlier, four years ago, the Hillcrest Labs. Now, we're, we talked about VisiSonics today, smaller acquisitions, add-ons to software pieces or other technologies that we need to fill the puzzle and to build a broader, product portfolio. We have done that. We have divested Intrinsix, and this is the key focus going forward. Also, Amir talked about this at length of our capital allocation. On top of that, we continue to be active in buyback. We have done $24 million of buyback over the last five years. We have just authorized the plan of 700,000 shares, representing anywhere between $14 million-$20 million. We're active, in the market these days with, with that, and this is something that we believe, it will continue as we go along.
Next topic of the top of the capital allocation is ESG. It was mentioned earlier, a topic that maybe around COVID, got, a little bit out under the radar. Wanted just to share with you that, an IP company with no manufacturing facility, with pure R&D, play, which is computers and office lights, we are very, very clean and friendly company and business model to the environment and to the planet. The fact that we have mentioned here over and over again about cost savings and power efficiency, and if you take that power efficiency and extrapolate that to connecting the home, the house, the cities, the cellular networks, the IoT networks, that power saving or battery could save a lot, a lot of trees, probably much, more like forests.
We probably need to improve our reporting capabilities because we are behind that. But if we do that, that's a little bit on top of our picks, so we'll have to work along it. But we are very, very friendly to the environment, save a lot of money. Our customers are able to do that, and of course, corporate responsibility, code of business ethics and all that, is something that we believe in. That's in the DNA of a company and has been with us. So we'll try to improve also the grades and the around ESG in the years to come. Summarizing all the different opportunities in front of CEVA in the that you have heard today, and I'll try to present it in four different pillars. One is the ubiquitous connectivity.
This is the strong stand of Ceva in the markets and the technologies and the design wins that we have. We are gonna target one billion devices, 25% of the worldwide market, in Wi-Fi. These are higher ASP chips, higher royalty contribution, and another opportunity in gaining market share in the new version, like Wi-Fi 7. We're gonna continue the licensing expansion in 5G Advanced, that means RedCap, that means satellite. There is a lot of demand outside the handset business. And higher ASP. That will enable us to have a higher ASP product mix than we had with one billion devices of Bluetooth exiting 2022. On the industrial IoT automotive, we saw some of the examples today. AI ramp up with two leading OEMs, that generates royalties and still more design activities around that.
A robust pipeline for sensing AI connectivity. This is where most of R&D efforts are put today, the AI segment and the connectivity Wi-Fi segment. We are anticipating in both of these industrial IoT spaces to ramp up in the next couple of quarters and years with volume and revenue. Edge AI. Edge AI market tailwinds. It's one of the hottest markets that are out there today. Michael talked about how difficult it is for newcomers to get into that, and how advanced our solution is to win market share into this space. We're talking about new, innovative product offerings. We're talking about a good pipeline in those in this market and cross-sale of those different technologies. And the software piece, we talked about that. It's shorter time to market. We have invested two acquisitions around that.
That's a nice add-on to our business. Faster time to market from royalties. The 3D spatial audio has a lot of interesting design wins ahead of us, to enhance the growth and comprehensive software offering and sensing experience to all the different use cases are in the world. All that gets us to the model. If we look at the 2023, start with the base of where this year is looking like. I'm using the first three quarters of actual non-GAAP results and the guidance that we gave in the last earnings call. We'll talk about shy of ninety-eight million dollars of revenues. We're back to the 89% gross margins, operating expenses anywhere between $84 million-$85 million-ish.
Still, with more investments and less royalties and licensing this year, only 2%-3% operating margins this year and $0.13-$0.15. This is a snapshot of 2023. For next year, you'll see a few TBDs. We'll leave it for next on our next earnings call to talk about what we believe the licensing royalties revenue could look like for next year. And obviously, that also will talk about the operating margins and the EPS expansion in our next earnings call. What we did say, and what we do still repeat and believe, is that we wanna keep the gross margins back to the 90%. That's the target... for the next couple of years, not just 2024.
We, from 2024, we want to keep the same expense levels, non-GAAP, overall, OpEx and cost of goods like we had last year. So with some revenue contribution, there should be an expansion in operating margins, but we'll give more insight around that when we have our next earnings call. When we look at the longer-term models of the next four years, 2025-2027, here we're coming up with a combined CAGR of 8%-12% of our overall revenues. This is a combination of both licensing and royalties, contributing from all these different markets, the four pillars that I just talked about.
We are going to keep as an IP pure play, the 90% gross margin in this business, and we want to increase the expenses of the company at a lower pace than the top line. Meaning, if we are at 4%-8% CAGR for the next four years, this can and should contribute to about 20% operating margins. Numbers that we have seen in the past, we know how to get there, but this is what we're sharing here is how we want to get there, which is 10x, more or less, where we are in 2023. And obviously, if all this also falls to the bottom line, we're looking at north of $1-ish type of EPS, non-GAAP, in 2027, significantly better than where we are today.
And all of this is based on the combination of continued growth in all these markets that we've mentioned, a bit more disciplined investments, continue to invest in R&D, but at a slower pace and much more focused on return ROIs. And there, the model works. If that's, if those are the parameters, then, the rest falls to the bottom line, to you guys, and to create shareholder value. So this is the model, and, Amir, let's summarize the day and the formal presentation.
Thanks a lot, Yaniv. I just want to summarize before we go to Q&A. I want to thank you. Can you hear me? Hello? Okay. Yeah, thanks. I just want to summarize the whole day of what we have shown today. I want to thanks my team and to you, all of you, also joining us today. The summary of where we are heading moving forward: First, we are back to pure IP business model. We believe we have all the capability to go and scale that business. We are addressing high-growth Smart Edge opportunities as we move towards 2027 and the rest of the decade, so we're expecting a tailwind as we go beyond this year and the next year. Highly diversified customer base and end markets.
We truly believe as an IP business model, the more diversified it is in our offering, the more we can scale it up and provide operational margin and leverage. Innovative mindset and winning portfolio. Again, it's about building the right portfolio. I was asked about, do we miss other assets and all that? Definitely, we keep looking for what assets, in order to make it more comprehensive and complete, as well as financial disciplines to drive that operating margin. As I came on board, we did lots of realignments of R&D and looking at what to invest, what to discontinue in terms of investments, and we expect to go and continue that in the journey. And very M&A mindset focus. We truly believe in the IP business, opportunities out there.
There's lots of consolidation that can happen, and we can take advantage of as we keep building our inorganic growth in addition to the organic growth. All that, we're very focused on the Smart Edge, diversified markets. We are diversified technologies and portfolio and building that success moving forward. With that, any, of course, questions to the whole team here, if you guys want to join me, of course.
Thanks to the team for some great presentations today. So maybe, you know, with Yaniv having presented all, I want to ask, re-ask this question-
I think you may need to come here for the microphone to work. Yeah. Yeah. Okay. Okay.
Okay. I'll start again. So, thanks, everybody, for the great presentations today. Very helpful. I want to re-ask the question perhaps, about the five-year growth, 1.5x, how that mix will be licensed versus royalty, and just to understand sort of, whether the licensing environment now is improved or, you know, is continually continue to be challenged in sort of the near term, flavoring the guidance you just gave.
You want me to take or you want?
Just yeah.
Go ahead then.
Okay. So, you know, we're not breaking out the licensing and royalties. I think we're looking at it as one bucket. Obviously, it's going to come from both. Obviously, both are planned to grow over the next four years. We don't know the magnitude of it and what we saw that last time, sometimes we were wrong, and the royalties were super much better in licensing. So we decided this time to give CAGR and the 50% growth that you mentioned, but not to separate the two. It's a combined business, and every... And then we were asked earlier to this morning, the business model hasn't changed. We're still, in every deal, we have a licensing aspect, we have a royalty aspect. In software, it's mainly royalties or only royalties.
That hasn't changed, and that's going to continue as part of the IP business model. So, that's what we, that's what we guided today.
Maybe a little bit just to clarify on that. If I summarize really the growth engine that we were talking, and I'll bucket that by royalty and licensing. Royalty, what we see three major growth within the so-called time period that we are talking for royalty, is the Wi-Fi ramp that we assured, is the overall IoT, where the so-called the initial investments happened in the last few years. It's now getting integrated into the platform, and we expect that royalty to grow. And our software business that will continue growing with the additional capabilities we added and the volume. Those will be the growth driver for royalty. For licensing, major growth driver is the AI to drive that growth, which is a whole new product line that we are coming to market and have significant growth. Again, it starts with licensing.
Royalty will come, of course, quite a bit later. And then in addition, is the 5G XC platform penetrating the different IoT board-based type of use cases from the licensing. We mentioned Satcom, we mentioned RedCap, software-defined radio for the different so-called integration of MCU solutions. And in addition, basically, the, the other new product, it will come overall in connectivity. We mentioned here Wi-Fi 7, Bluetooth 6.0 and all that. That will be the, so-called, the new generation of licensing. So we are not providing the clear breakone, the clear, breakdown, breakup, sorry. But you can see basically what are the key driver that for each of them that we drive.
Then my other question is really on the financial metrics. As we look ahead and try to track this five-year plan, I think one of the things you said is the IP content per device, but the other one seems to be cross-selling the products to the same customer. Would you contemplate metrics such as revenue per customer and how that's growing over five years, to understand how the cross-sell is working, or other metrics to help us understand these two key vectors of what you're trying to do?
Right. Right now, of course, we are not sharing that type of a breakdown. We already shared today quite a bit of a new perspective of how we look at the revenue by market segments and so on. We'll take that as an input, and over time, we can assess that.
Maybe the next, LSD, right? Leave something for next time.
Yeah, even need to track it.
Yes.
Even on earnings calls, we do. You know, the last two earnings calls, we pointed out that we did three combo Bluetooth, Wi-Fi deals on each quarter. So when there's use cases and applications where we can say there's a combination of IPs, we'll do it to help you understand that we're layering it in. But as Yaniv and we talked about earlier as well, you know, the revenues are so much more diversified now in terms of end customers, with just about 50%. The top five royalties only coming from 50% of the customers these days.
To track all the different customers and how much they're contributing, it's gonna be a huge, big spreadsheet that I know some of you here would love to manage, but, as Amir says, it's difficult internally to keep an eye on all of that and share it around.
A question about software and how you see that growing. Do you contemplate monetizing software, or just doing more network related or AI related? And if you make a software stack or an AI engine or that you license, can you then create other software stacks that can run on the same AI core that might do a different function, but on using the same core? And does that open up a whole new opportunity for the company on licensing software and also growing a separate royalty stream?
You want to take the software, maybe Michael, the software packages?
No problem.
So first, when we talk about our software licensing business model today, we are really talking about where we provide it as a complete package, as a product that provides a user experience, whether it's 3D spatial audio, voice activation, voice enhancement, and those type of things. When you were alluding to the software stack that goes on top of our silicon or silicon IP, today, we are also licensing that, and in some cases, to some degree, so-called, as a separate licensing license agreements. But this, this is really to enable our penetration and stickiness within our silicon capabilities in those domains. Over time, there is a potential to take it just that alone, so-called, as a solution to offer in the marketplace.
When we do that, we want to come up with a complete experience, not just here is a software stack that enable a silicon. Okay?
Are you charging up or is that just part of the initial license, that software? Is it like or is it truly a separate royalty, or is it when you sell that IP core, it's already bundled in there and that's just the price, but there happens to be software as part of that now?
Yeah. So this is what Amir mentioned exactly. Some of those software pieces, especially around AI, are licensable. The idea right now around AI, because it's still a new market for us, is to enable that customer to pay royalties per the chip itself. So that's an enabler. You need to buy it separately, you license it separately, but the royalties will pay be part of the chip as well. On top of that, what Amir said, is that we may think in the future to license, if we have a strong enough customer base of AI, chips out there, we wanna come up with different software version. Maybe we could also charge royalties in the future on those specific software packages. Right now, it's a licensing piece on top of the AI license for the core itself.
Let me give some- about a bit, sorry, a bit more color on that. So if you look, for example, our software sensing technology that I referred before, is a separate software package, but we can just take royalties on this piece of software. So today, we already exercise this business model to take our voice technology with spatial audio and the three pillars that I explained before on software and sensing, and we can charge royalties just on that.
I find your side-by-side comparison between Bluetooth and Wi-Fi quite interesting. So I think in the past, maybe products such as TWS headsets has contributed nicely to the Bluetooth growth. Do you expect to see any products similarly in Wi-Fi that maybe helped you to jumpstart the growth? That's the first question. Second question is: Can you comment on the competition you face, in the respective markets in Bluetooth and Wi-Fi? Thank you.
Yeah, in terms of, I'll start with the Wi-Fi. In terms of so-called what drive the demand in the market, I would say overall, the needs for the next generation of throughputs of lower latency and higher capacity drive the needs. It's not necessarily there is one use case or application that drives the Wi-Fi needs. Wi-Fi market is already very established and large. The, the new things for us in the last several years is that we came with a very capable IP solution that lots of different companies decided to go and utilize. 'Cause they moved from, let's say, being just an MCU, and they needed to connect the Wi-Fi. They wanted to go and build a new b- solution based on Wi-Fi. They came to us.
The other thing that we have done much more aggressively in the last few years and successfully is penetrating the Wi-Fi access point. A lot of the new access points players today in the market and the newcomers that came in the last few years are using or are going to use in high volume our Wi-Fi AP IP. So that's what drives significant. And with that, you need to understand that as you look at the accumulation of the licenses that we have had, in Bluetooth, the early days, it took time to accumulate, let's say, to 30, 40 type of licenses. Once we reached that, and we kept accumulating the same level every year, that volume start ramping up very nicely. We see the same thing is happening and will continue in Wi-Fi, just with that shift in years.
But much higher ASP, the license was bigger. If you look at also the revenue that we generate on Wi-Fi in the last five years, where the royalty was much, much lower than Bluetooth, still it was very significant number versus the Bluetooth, 'cause the value per license and per royalty-bearing unit is much higher than Bluetooth. So that's why we're expecting significant growth there. And both station and access point, in Bluetooth, you don't have access point. It's that helps a lot, and also the combination of the two. If you had an-
About competition.
Yeah, about competition. I would say overall, we are clearly, clearly the number one in the marketplace as an IP supplier. I would say overall, we have a dominant position versus competition. We see here and there a little bit solutions coming from potential competitors. I would say in Bluetooth, maybe a little bit, but no one has that so-called backward compatibility and future-looking in terms of the standards, plus BLE and Bluetooth dual mode, plus the software stack on top of that, just the whole completeness, we really don't have competition with... from that perspective. And then Wi-Fi, we are clearly the solution out there as an IP supplier, provider, right?
I'll just add one more. Martin, you asked about the markets and potential drivers for volume. The Wi-Fi 6 for IoT in the smart home is a good example. So most devices in the home today, you're talking about smart plugs and switches and so on, that Amir talked about earlier, they're predominantly still using Wi-Fi 4, the old 11n technology you remember from a decade ago. Wi-Fi 5 was not built to address IoT, and then Wi-Fi 6 was architected from the ground up to have this very low power, ultra-low power profile, that suits these types of connected devices around the home. The other thing to remember is now we're connecting 50 or 100 different devices in the house. So if you keep adding Wi-Fi 4 devices, you're gonna have a lot of interference and issues with your Wi-Fi performance.
Wi-Fi 6 can handle many more connections on the same network. So this is why we're seeing a lot of our IoT-focused customers who've done really well in Bluetooth, going after that now. They see there's a big opportunity to scale up into the connected home type of application for many devices which never had Wi-Fi or were using the older Wi-Fi. So that's a volume driver.
Yeah, a quick question on NPU. Beyond the two automotive customers you have, can you talk about, you know, market validation with other end markets and sort of what those end markets might be? You know, PC, auto, mobile, you know, any color there would be helpful.
You want to start with that or?
You want to?
Okay.
Go for it.
I think the question was about NPUs, right?
Mm-hmm.
Penetration in the market. So definitely we talk about automotive, where we see, and it was also in Michael's slides, the infrastructure type of customers. They are moving from so-called cellular communication to generative AI as part of the overall capacity optimization. So that's another opportunities for us with NPU to come along with our wireless communication DSP technologies. We see the NPUs going to client device. You mentioned PC, it's PC, tablets, those type of devices as well. We see it's also in, let's call it, high-speed wired optical communication, where you need to and you want to use neural processing unit with basically running a network in order to improve the or do noise cancellation in a better way. And we really see it across many, many different type of implementation, implementation or applications. This is the high end.
Also, slightly lower end, you will see NPUs or AI capabilities going in the broad base of the different devices, so-called the Smart Edge devices, from smart speakers to other connected home devices. Enabling audio type of use cases related to AI, as well, different type of assistants, audio assistants, as well as more low-level video recognition, image recognition for video applications, like camera door cameras and so on and so on. So we see that really propagating all across.
My question was along the same lines, but maybe just to go into a little more detail on that. Would this be a companion chip with someone that's already has a full CPU, or is it integrated with the CPU? And then, if they license from you today, when would they have a product in production? And I'm thinking for the, say, even the home appliance market.
Yeah. So the NPU, typically the, the, the large NPU, the NPU that we talked about, is typically a companion to some processor, internal processing that you need to run, you know, just what you need to run on, on that machine in terms of just control and understanding of an interface with operating system or, or the touch sensors or the touch display and so on. And so the NPU is accelerator. It's accelerating to run the neural networks in addition to some processor that is there. The opportunities, and I won't go into that too much into details, to look at the different type of combination with our different technologies into one single solution. Which we may talk about it in the future.
The other question was time to market, like, if they licensed today, when would they have it?
Ah, yeah. So as, as we showed, also with all the other silicon IP technologies, from the time that we license, it's typically on average about two years until they go to production. If it's simpler system or already they are quite advanced with their own developments, it can be even a year. If it's very complicated system and they are very early in their developments, when they license from us the IP, it can take three years. It also depends on, really when they license from us the technology, is that just before they start the project or where they are made already some progress. But it's anywhere between one and three years, typically two years. Okay?
I have a couple of questions here that I'll just take that were sent into us. The first says, it might be a question for Yaniv, but of the 250 deals done in the last five years, can you estimate how many are producing royalties today? And the second question is: When a customer licenses an additional IP, is it faster for that customer to get to market when they're coming with the second IP as well?
Let's start with the second one. You know, it's the same question and answer we just gave right now. The time of designing a chip is more or less the same. If you have the capabilities and the knowledge in-house because you've done a first generation and you need to add feature sets or a new standard, and you already have done it once or twice, then the next use case will probably take as much, but it may be a bit more efficient from the size of the team you need for that next generation chip. But it does not change much, the timing to create royalties or for the chip to go into production. And the first question of how many people or customers out of the 250, it's a good one.
We need to do an analysis. I'm not sure if I have a number off the top of my head, but we could look at the delta of number of royalty shippers. As we presented earlier, we had about 60. Not all of them are the same, 'cause some have changed. Today we have about 80-ish. So those 20 for sure came as part of the 250 deal. But I think the number should be larger because we did have some old-timers that finished their production and went. And as the markets changed, the markets also changed for us, and others got in from in the last four years, mainly around the Smart Edge type of devices.
So I would say probably anywhere between 20 to maybe double that is the right number over the last five years, out of that 250 deals.
Thanks. Amir, I just wanted to get a question into you about the framework for thinking about the competitive landscape in AI and NPUs as you start going out there. Is it really you guys as, as a co-processor accelerator versus this, an AI capability in the native chip there, or are there people building co-processors, competitive? Just to understand. I just want to get a framework for how to think about your edge AI NPU and what the competitive landscape there looks like.
Yeah, just to clarify, also, I would say to the most part in the marketplace, and as well as our competitors, when you go to very high performance level as a requirement, and when you have that, what we showed, like neural processing unit, that needs to go to hundreds and thousands of TOPS, this is a separate accelerators that goes with some kind of a co-processor that is there to run the rest of the processing needs of the system. So it's also the overall landscape is moving there. So I'm not sure if I answer your questions, but... I think that's good? Okay.
Hi. I just want to make sure that your revenue-
... long-term revenue guidance didn't assume any M&A. That was just organic, correct?
Correct. This is the organic model, no M&A, and no, no other fruit.
Okay. And as you look at the market, in terms of these pretty big growth markets you identified, what type of opportunities are there for getting significant market share or going into additional markets? Like, how do you look at M&A?
Can we? Can you hear me? Can you hear me now?
Yes.
Yeah. So I think Amir, Amir mentioned that earlier today. I think we're gonna look at two type of deals. One on augmenting our technical capabilities, bringing more features, but at the same time, growth. Growth can be into expanding our capabilities with a scalable business into areas we play in. And I think we're gonna still play in those six end markets. Probably, we're gonna add additional use cases that can vary from everything. You know, I don't wanna talk about specific one, but there can be multiple use cases we can go and address later on, and by that, decreasing our SAM across the end markets.
And where are we in sort of identifying what's attractive, what's not attractive? I know in the past you've looked at going into, other areas that-
I think we've. You know, we have a strategy and a plan. I think there is a game board, but this is what I can share right now. We're looking at a lot of things. It doesn't mean that we're gonna act on something really fast.
Right. And do you think this is a, just looking at the macro, it's probably a better time to be affecting a strategy in terms of valuations, given what's out there?
I think it's.
Are you seeing more opportunities from that perspective, more companies are looking at divesting initiatives that maybe you can productize?
Yeah. So there is definitely the divestment from larger companies that may fit our product portfolio very well. I think from a valuation, it goes both ways. You know, it really depends on our share stock as well. But we do have a lot of, you know, our balance sheet is very strong. So when we look at a deal, and I'll let Yaniv comment around that, there are multiple ways to do that overall.
Some mobile OEMs are trying to design their own 5G, you know, basebands and haven't been successful or been able to get past 4G. You know, is there any possibility, you know, any time before, you know, in the next, you know, few years that that actually could happen?
Yeah, I would say first there was a beyond the so-called, the Cupertino folks, okay? Beyond that, the discussion, I won't go there. But just recently, there was a company that basically decided in China to stop doing their own SoC modem technology, and they were actually using some of our IP and technology. So definitely we see that so-called landscape change. But also what we see from the assets of that team and capabilities, we see that now potentially propagating such that maybe others will try to internalize. So within the smartphone OEM, there is the Apple, the Samsung, and several in China. There is a potential that one of them in the future will try to go and build or start building internal capabilities for cellular technology.
Time will tell if they are going to be successful and whether they will really take it all the way. I'm not talking about Apple, okay? As, as everyone have seen, this is a major undertaking and not easy to overcome. Yeah. There is potential.
I have two questions on Gen AI. So first is about, you know, market forecasting 2.7 billion in Gen AI devices by 2027. I think smartphone is a very obvious device that account for a lot of that forecast. Is there any other edge devices you feel like a very high potential device that counts towards that forecast? My second question is: you talked about the scale you're capable from, you know, single TOPS to 1,000 TOPS, and then you also talk about the efficiencies Ceva has over competitor. Does that mean you would maybe favor more low-power devices as the first end markets or end customers to utilize your Gen AI or NPU products in the future?
Yeah.
You want to take or should I? Yeah. So for the second question, I think, you know, AGI as technology is ubiquitous. As Yaniv, as Amir explained, different use cases can be in digital cameras, drones, surveillance, ADAS system and mobile phone, notebooks, ubiquitous. So that's the reason we don't want to bet on one single use case. Strategically, we constructed the architecture to be super scalable in terms of how many engines, because this architecture is very, very modular. How many engines you can integrate in our in different solution to start from a very low end, and that will enable us to go to the specific use case of consumer.
Going, adding more and more engines and going to cluster architecture journey with many engines, many cores, all connected together to achieve the 1000 TOPS that I referred before. That's for more high-end use cases like infrastructure use cases, ADAS, ADAS technology, and more. So strategically, we invest to create modularity, scalability, not to target one specific use cases, but to give all our possibilities, and it's a real asset to go to the market. Talk about edge?
And about the devices, I think you may, you know, mobile for sure, PC, tablet, and some other devices you interact with. It's not gonna go for your earbuds, most likely, but they have the compute full compute; they're gonna have some aspect of it. But again, those are more of the more compute-intensive devices that can do everything locally. Again, as Michael mentioned in second, there's gonna be a lot of hybrid activity, so you're gonna do an infer something on your small device, move it to somewhere else. So we're gonna see a lot more around that. But those who can run it fully, completely with the data sets attached to it, those are gonna be probably the type of devices based on the requirements of memory and everything else.
Okay, so if we've no more questions, I think we can call an end to the day, and thank you all for coming. We appreciate all of your attention for the last few hours and also to everybody on the webcast, thank you. Good luck and, see you down the road. Thank you. Bye-bye.
Thank you.
Thank you.