Good morning, everybody. My name is Stuart Chambers. I'm the Non-Executive Chairman of Arm Holdings. I joined the board in January. Seems like ages ago, but it's not really. I became Chairman on the 1st of March, so been in the chair for three months. With that position, I get the privilege of welcoming you all to Arm's Analyst Investor Day today, and also to introduce the session. This is the customary cautionary statement. I think most, if not all, will be familiar with this. If you're not, please refer to it. I believe it's on page two of your booklet. This is our agenda for the morning.
Simon is going to kick off and, among other things, he's going to talk about how we gain share in our various end markets, how we build out our technology portfolio, and he's going to talk particularly as well about the embedded market. Lakshmi will take over. Lakshmi is responsible for our service strategy, and she'll be touching on our progress in the server market and also in enterprise infrastructure. You'll get a coffee break, and Pete Hutton stands up. Pete heads up our product development groups and therefore is responsible for developing and deploying most of our technology that we have in Arm. He's going to be talking, among other things, about winning in mobile computing, but also how we actually grow beyond mobile.
Tim will talk about our financial strategy and will also cover, in addition, our investment in R&D and growing profitability. Simon will chair our Q&A session, after which I suspect you'll all be very hungry, and we'll retire for some refreshment. Without any further ado, let me hand over to our Chief Executive, Simon.
Thanks, Stuart, and good morning, everyone. Thanks for joining us today for this year's Analyst Day. I've been the CEO of Arm now for almost a year, and I've had a great time in the last 12 months. It's been really interesting, really exciting, getting to talk to lots of different people in lots of different industries about lots of really cool stuff. The question I've got many times over the last 12 months has been around Arm's growth strategy. How is Arm going to continue to grow? We've had 20 years of fantastic performance. What's next? How is Arm going to continue to grow? I see that we've got many opportunities for growth. I'm really excited about what's going on in mobile and the potential for future growth in mobile, continued future growth, and the revenue opportunities that that generates for Arm.
Beyond mobile, we also see other opportunities for growth in high volume, high value markets. Our business is fundamentally about embedding intelligence. It's about designing energy efficient and cost-effective microprocessor technology and deploying that in pretty much anything we can think of, anything that moves, and most things that don't. We see many opportunities for deploying embedded intelligence and having that help drive growth in our revenues and driving Arm to be a bigger and even more profitable company. Today, what we're going to do is talk about some of those opportunities. We're going to go into detail about some of the end markets. You'll hear that from my colleagues. What I'm going to start with is a more holistic overview of where we see these opportunities for growth, and then we'll go into some of the detail later. Starting with mobile.
Over the last few years, we've seen a mass proliferation of mobile devices, and through the execution of our roadmap and our business model, we've achieved around a 95% market share in mobile, which is fantastic. You may think, well, 95%, how does that grow from there? Well, actually, I think we're just scratching the surface of where mobile is going to go and what mobile can do. Although mobile phones are very prevalent, there are still billions of people that don't have them. There's a big opportunity as cost is driven out of the end price of a handset to enable many, many more people to get access to phones and, in fact, mobile computers. The attributes that have made Arm so successful in mobile turn out to be useful in pretty much every other market.
Whether it's designing really small sensors or really high performance servers, low power, energy efficiency, the ecosystem that we've built up around our company is equally applicable in mobile as it is in all of those other markets. Through the execution of our roadmap, through the development of the products that we've achieved and built over the last 20 years, we've seen Arm processors used in more and more end devices. Everything from TVs to cars to motor controllers to disk drives. We're seeing Arm processors deployed in a massive array of end equipment. With that, there are new markets that are emerging, and we are leading the way in the new Internet of Things and wearable technologies, and I believe that Arm will be the most prevalent architecture in these new and growing markets.
The side effect of more mobile devices, of more embedded intelligence, is the generation of data. What we're going to see over the next few years is an exponential growth in the amount of data that is generated and processed and moved around. To deal with that data requires a network infrastructure that is capable of both dealing with the volume and dealing with the different types of data that will be moving around. It's not all about providing data into handsets. With the growth of the Internet of Things, it's about extracting very small amounts of data from sensors and broadcasting it in a very efficient way. That data is going to get processed up in a server.
It's going to get moved into the cloud, where the cloud computing will analyze the data, it will extract information, and then in many cases, provide that information back to the user via their mobile devices, in most cases, or in some cases, directly to other machinery in the industrial internet. To deal with this explosion of data, it's going to require new architectures in the network infrastructure, and it's going to require new architectures in the cloud. In all cases, to deal with the volume, we're going to need to enable all of that technology to be built in the most cost-effective and power-efficient way, just to deal with the volume. I believe that Arm technology and our focus on energy efficiency, cost effectiveness on ecosystem, makes us ideally suited to generate a solution for every point on that picture.
That's going to help us grow our business as we generate more and more high volumes of Arm-based silicon devices. Let's go back to mobile for a moment. Over the last 20 years or so, mobile's been a huge economic driver. It's been a huge driver of the technology industry. For us, the real discontinuity came, and the opportunity came about, with the advent of digital cellular, and particularly GSM, where Arm7, this processor that we designed many years ago now, turned out to be a great solution for building GSM modems. Over successive years, we've seen phones get much more sophisticated. From the basic Arm7-based processors, we've added more and more processing capability as phones became more sophisticated with color screens and cameras, and the first downloadable applications became available.
Around the Arm11 and Cortex-A8 period is when phones really became smartphones. We saw handsets with high performance, high bandwidth data connectivity into them. We saw color screens, touch screens, and we saw enough processing power for phones to start to become really quite useful devices. Building on from there, around the Cortex-A9 period, and particularly when we introduced multi-core Cortex-A9 devices, smartphones became really mobile computers. They had enough performance in them that you weren't just using them for sending text messages and playing basic games, but they started to become really useful mobile computer devices. The growth of mobile phones has been phenomenal.
You can see on the graph on the left there, that the number of subscriptions to cellular services has grown pretty much exponentially, and there are now about as many mobile subscriptions as there are people on the planet. The number of people on the planet is something that continues to grow. Although there are as many subscriptions as there are people, when you look at smartphones, there's only about 2 billion smartphones in circulation right now. Which creates a huge opportunity to provide smartphones and mobile computers to billions more people. We see that as a huge opportunity. It's an opportunity that comes about because devices are becoming lower cost. It'd be great if everybody on the planet bought an $800 smartphone, but frankly, they're not going to. That opportunity is only realized through the advent of low-cost handsets.
That in itself creates a big opportunity for Arm because of the Arm-based content that can go into all of these devices. This opportunity is enabled because not all phones are the same. Now, I mentioned Cortex-A9, and when Cortex-A9 launched, and when we delivered multi-core Cortex-A9, it was the pinnacle of high performance in smartphones. It was the top of the premium sector. Over the years, what we've seen, though, is smartphones really kind of stratify into different tiers of handsets, driven by different price points. What you see on the right there is how the market breaks down in 2013 between the premium, the mid-range, and the entry-level devices. In all of those different price points, what you see is a different mix of technologies.
Different mix of processing performance, different mix of graphics, different screens, different modem connectivity that enable these different price points. Across that entire range of technology, what we have is a processing solution that is matched for the performance level and matched with the price point. We have a technology lineup that is enabling these different tiers of device and enabling the volume growth that is going to come. Different tiers of this market will grow at different rates, but overall, this market is still going to grow at about 10% CAGR for the next five years. In most markets, that would be considered pretty healthy. Within each of these devices is what we refer to as a CPU subsystem.
It's got a microprocessor, it's got graphics, it's got connectivity that allows these devices to talk to each other and talk to external memory, and they're implemented in silicon. When we look at those subsystems across the entire smartphone lineup, what we see is an opportunity for a Cortex processor or a cluster of Cortex processors, combined with one of our Mali graphics processors. Then to implement that effectively in silicon requires our physical IP. We have a lineup of technology called POP IP. It's all about optimizing the performance and the power efficiency of the processor in silicon. Through the combination of these technologies, we can deliver these different price points and deliver the best performance with the lowest power.
What we saw with mobile is it was a growing market, and as we evolved our roadmap and our technologies, we've been able to grow with that growing market and help define the characteristics of that. As I said at the beginning, the attributes that we have in our products, this focus on low power and efficiency and building this ecosystem, has enabled us to grow into other markets. Those other markets see the benefits of our overall business model. If we consider networking. Historically, Arm has had a strong presence in consumer networking, in Wi-Fi hubs, in home gateways, home routers that allow you to connect your DSL modem or your cable modem. It's been very focused historically on consumer-level devices.
As the performance of our processors has increased and as we've added new technologies, with particularly the advent of Cortex-A9 and Cortex-A15, we find we're able to target technology into the enterprise infrastructure, into enterprise networking, and start to gain market share in what is a very valuable market from a silicon perspective. The architecture of the kind of chips that people build into these devices are quite different from what you see in a smartphone. What you see in an enterprise networking device are many, many CPU cores processing data in parallel. How those processors communicate with each other is very important. Doing that in an energy-efficient way is very important because there's so much data moving around the chip. To help address that need, we developed a technology that's referred to up here as CCN, which stands for Cache Coherent Networks.
I won't go into the details of what that means. If you're interested, we can talk to you later. It's a key technology for allowing these very high core count processors to be put together and delivered in an efficient way. By doing that, by building on our processor roadmap, by building features into our processors which allow them to communicate with each other, by building products like CCN that allow the subsystems to be put together, we've been able to expand the reach of the Arm architecture and target this new market, which I said is a very valuable space when you look at the silicon content. In embedded processing, again, we see something similar. Our licensees started building microcontrollers based on Arm7.
As people started to see the benefits of a high-performance 32-bit processor in microcontrollers, as they saw the benefits of the ecosystem that Arm has, where we can provide third-party IP, where we can provide third-party tools, where people who use microcontrollers can source their silicon from a number of different suppliers, we've seen the opportunity grow. As such, we've taken it upon ourselves to create a family of processors specifically targeted at the embedded microcontroller space. That is Cortex-M, it's been a really successful product line for us, I believe it's going to enable us to take a very high share of the embedded space. When we look at the top right here, up into the data center and servers. Again, some of our partners started developing server-based chips using our Cortex-A9.
In fact, the first ones were based on Cortex-A9, then Cortex-A15 as we delivered more performance. These are, of course, 32-bit processors. As we've added 64-bit capability into the Arm architecture with the development of version 8, we see a real opportunity here, we're starting to see new devices coming from our partners, which we believe are really going to revolutionize the data center and meet the needs of the modern data center. In all cases, we've started with existing technology, we've expanded our product roadmap, we're continuing to execute to develop new products, new technologies, which are going to enable us to further provide more value to our partners and allow them to target these high-growth markets in a very efficient way.
All right, where's the money? Everybody's used to the value of the smartphone application processor market. Where's the money here? There is a misconception, in my opinion, that no market is as valuable as application processors. When you look at this in broad categories, we tend to lump these markets together, group them together in terms of mobile applications, embedded intelligence, and enterprise infrastructure. When you look at the value of those markets, what you see is that today, each is worth about GBP 13 billion. Over the next 5 years, we anticipate roughly a 50% growth for these markets to be worth about GBP 20 billion in 2018. Each of these is a very high-value market when we look at the silicon TAM.
This is the revenues generated from every chip that is sold into these particular market segments. For Arm, there is an opportunity to generate a royalty stream from every chip that is sold into these markets. Based on the technologies that we have and the value of these markets, let's look at how and where there are some opportunities for Arm technology. Going back to mobile for a moment. As mobile continues to grow, I expect that we're going to maintain a very high market share in mobile. Much of that growth, as I said, will come from entry-level, low-cost devices. That is a good thing because there are many opportunities for Arm technology in those devices.
As they are adopted, as more and more people get to use them, I think that's going to have really quite a profound effect on society. If you give another couple of billion people on the planet smartphones, all the data that we've become accustomed to is made available to everyone, and that will change people's access to health information, to education, and will have quite a dramatic effect on commerce in some of the poorest places on the planet. I think the side effect of that is going to be beneficial to everyone. We're going to continue developing our roadmap. We're going to continue developing more and more high-performance parts, but with that focus on energy efficiency.
We expect to see mobile computing taking more and more different form factors, not just handsets, but we expect to see Arm processors used increasingly in clamshell form factors and hybrid devices, all enabled by energy-efficient processors and high data bandwidth connectivity, cellular connectivity, into these end products. Mobile is also very closely linked to the Internet of Things. There are many examples here where the smartphone is being used for the controller of the Internet of Things. Many of the examples here are very consumer-driven, but the same applies for industrial applications where data is going to be monitored in equipment and people want to make decisions based on what is happening, and the mobile device will become your window into the Internet of Things.
What's shown here are very first-world examples, Nest thermostats, parking sensors, which is that gray thing in the middle, Bluetooth-controlled light bulbs. Going back to the growth of smart devices, mobile computers in developing countries, I expect to see new applications which are specific to those markets. It may not be finding out where you left your car in a car park. It might be more about much more fundamental things about how you control the drip irrigation system that you have in your field. The Internet of Things, very linked to mobile, and I think one controlling the other is a key attribute that will make the Internet of Things a successful technology. In enterprise networking, we have today a very small share, but that is growing.
It's growing because of the design wins we have with pretty much every semiconductor company who is providing silicon devices into this market. You can see the logos here, all of whom have announced products, and there are many other companies who are working on products, and many design wins which are going on. As I said a moment ago, the chips are different. Lakshmi is going to go into this graphic in the middle here in a bit more detail, but you'll see high core count devices. For every one of those cores, that's a royalty-generating unit for Arm.
As I said, the technology that allows these devices to be put together is really important, and our customers are looking to us for more and more help for how to build these devices and get them to market quickly and allow them to differentiate around the technologies that we provide. The reason that Arm is being so successful in this space is because the investment that we are making in the Arm architecture, the investment that we're making in the ecosystem, in the third parties that are providing complementary IP. It's that investment that makes Arm the architecture of choice in many markets, and I believe it's why we will be successful in growing our share from what is today only about 5%, 5% in 2013, to a much larger share over the coming years.
First design wins we're seeing today are around LTE is obviously a very high-performance cellular technology, but it's in its infancy in terms of rollout on a global scale. We're starting at the radio end of LTE, and we expect more and more Arm technology to be used moving in towards the center of the networking core. You'll hear more about that later. We expect that 5% to grow, and we're really quite bullish about our prospects in this market. At the end of the network is a server. I think the Arm partnerships made great progress in the last 12 months in making Arm-based servers with all the benefits of energy efficiency that come, a reality. Again, we have many of our semiconductor partners building chips with announced designs. Some of them are starting to sample devices. You'll actually see some later.
There are many others which are in development. The chips are no good on their own. You actually need an OEM to build a product. There are a number of OEMs who are doing that. I'm expecting to see units shipped towards the end of this year. Real server boxes running real workloads in real data centers. Making that happen doesn't just require hardware. It requires a lot of software as well. We've put great investment over the last few years into building the ecosystem that's providing the software that will run on servers and make this low cost and low power advantage a reality. Again, you can see the logos here.
It's been fantastic effort from across the ecosystem in building all of this software and some great partnership activities going on, especially around Linaro, that is bringing these companies together to help make all of this software a reality. Together, these components are coming together, devices are being tested, and I think pretty soon we'll start to realize the benefits that we can see from revolutionizing the data center. Now I'm going to talk about the embedded market for a little while. I think what's interesting about embedded is just how diverse the end market is. The Internet of Things isn't one application. It's thousands of applications developed by thousands of different companies. Everything from big companies to individual inventors and innovators sitting in their garages soldering things together.
It's a market that's very, very diverse, and it's enabled by the fact that the microprocessors, the microcontrollers that our partners are building, are absolutely tiny. We have examples where our partners have built chips so small, they fit inside the dimple of a golf ball. You think about that, it's absolutely tiny. As a result of that size, they're very low power. They can be built at very low cost. That is a great thing because it's going to enable billions and billions of devices to be built and deployed in all sorts of applications. I believe pretty much any company can benefit from the growth of embedding intelligence and the growth of the Internet of Things.
We're going to see companies just understand more about their supply chains, understand more about how their products are put together, and have much greater control over their own infrastructure, their own factories. We're going to see companies developing new businesses, new services due to the insight that they're going to get through understanding how their customers actually use their product. I think the Internet of Things has the opportunity to benefit pretty much every company on the planet. Every company can become high tech in some way or another. A great example is the device shown here. This is British Gas's latest smart meter. Now, conventionally, you might not think of a gas company being particularly high tech. Here in the U.K., there's been a lot of advertising recently showing British Gas really trying to bring modern technologies to their customers.
The Internet of Things is one way in which they're doing this. Through smart meters, you can get much more information about how energy is being used, and you have the opportunity to control it and lower energy. That, of course, is something that everybody needs to be worrying about. This smart meter gives you information about use of your gas, use of your electricity from which you can make informed decisions. It's a great example of the kind of second-generation IoT devices that we're starting to see, which are people really thinking about how users interact with all of this data that's being generated and how you glean information from it. This is a great example, but maybe you want to see historic trends.
Maybe you want to anticipate what your energy use is going to be based on what the weather's going to do in a few weeks' time. To do that, you need to be able to combine data from different sources. When that happens, I think the real benefit of IoT will be realized, and it requires standards and interoperability between different providers and different gatherers of data to make that a reality. It's going to happen, and I think we'll see many new opportunities for IoT from companies like British Gas and, as I said, pretty much any other company you can think of. At the heart of the Internet of Things, or depending on how you look at it, at the extremities of the Internet of Things, are these really tiny sensors.
When you take any one of them apart, you're really going to see three things. There's an analog sensor, and there are a couple of digital components, which is where Arm has a big opportunity. One of those digital components is the microcontroller that's taking the data from the sensor and making some sense of it. Then there is some form of radio connectivity that is taking that data and broadcasting it up into the Internet where it can get processed. For Arm, there are opportunities in the MCU, opportunities in the radio where we already have quite large market shares, especially in the radio. As this market grows, again, we expect to see higher and higher volumes of chips based on Arm technologies. One thing that's driving that is the popularity of our Cortex-M series of processors.
We've licensed this now over 200 times to over 150 companies. Some of those companies have been long-term existing Arm licensees. Some of those companies have been new companies taking a license to Arm processors for the very first time. There's over 20 of them, I think, in the last 12 months alone. We're seeing a lot of design activity going on around Cortex-M to serve the needs of this very diverse market. I expect that Arm will be the number one architecture in the Internet of Things as all of these devices come to market and these companies start to sell their products. There are lots of examples of where Internet of Things devices can be used. You've seen the gas meter just earlier. There are examples in healthcare, in agriculture, in transportation.
Energy control is one area where people are spending a lot of time looking at the moment. I have another example of that here. This is a case, in this case in New York, where a new law was passed to try and reduce carbon emissions from all the buildings in New York, some of which are, of course, very old and have very old energy infrastructure. In order to make any progress with this, the first thing you got to do is start monitoring energy usage and then start controlling it. Doing that in very old infrastructure is very hard. There's a company here, Bonded Energy Solutions, who were commissioned to go and try and find a solution. They worked with a company called Electric Imp. They're a U.K.-based company who built a Wi-Fi platform based on Cortex-M3, I believe.
By deploying that, they're very low-cost devices. This is actually their development platform that's shown on this SD card. They've been able to deploy this Wi-Fi technology into these old buildings, start to gather data, start to control the sources of energy consumption, and they've exceeded the requirements that were set down by this New York law. It's a great example of old equipment being retrofitted with modern technology to help glean information and then get control of something which is otherwise basically open loop. There are lots of examples of this. The Internet of Things, as I said, it won't be the Internet of one thing. It's going to be the Internet of many things. There are many people, companies like Electric Imp, who are building IoT devices.
If you really want to get a flavor for the kind of diversity, the kind of innovation that is going on, then I recommend that you go and check out your local Maker Faire. This may be something you've never heard of. Maker is an organization who are dedicated to enabling people to go and experiment with technologies and just go and build stuff and see what interesting things result when you put inexpensive technology and software and cloud computing in the hands of pretty much anyone who wants to go and play with it. It's really quite exciting what's happening. This weekend just gone was one of their biggest events of the year. They run a Maker Faire in the San Mateo County Showgrounds, which is enormous place just south of San Francisco Airport. I went along.
Me and about 110,000 other people attended this for the weekend. I mean, vast event. Unbelievable. What's on display are things people are just building. They are taking low-cost development boards. They're writing software. They're soldering things together in their garages and building things. That sounds a bit hobbyist, and some of it is. Some of it turns into real product as well. While I was there, I took part in a product launch of this board that's shown bottom left there. This is an Arduino development board. Arduino is an organization that was set up a few years ago to make low-cost microcontroller boards available at very low cost. These sell for like GBP 30. The new generation of Arduino boards are all based on Arm processors, and they've sold about 1.5 million of these. This isn't low-volume stuff.
It's pretty respectable volumes. They're getting them out in the hands of people who go and experiment. They write software. They raise money off of websites like Kickstarter, and they make products. Now, this smartwatch I'm wearing from Pebble started off as an Arduino prototype. They used an LCD screen off an old cell phone. They hooked it up to one of these boards. They wrote some code, and they eventually turned it into a product that's gone into mass production using an Arm processor. It's a great example of how this experimentation turns into real companies, real products. There was just a ton of this on display. It's really fascinating to go and see. It's all about bringing low-cost semiconductor devices, software, and the cloud together. I hope I've given you an overview of where we see opportunities for growth.
I'm really excited about the growth potential in mobile, and I'm also excited about the potential we see for growth in other markets. I've outlined here a number of high-volume, high-value markets where Arm's technology, Arm's focus on energy efficiency, on our ecosystem, and cost-effectiveness can bring a great benefit to Arm as a company. I believe the technologies that we have are going to make Arm ideally suited to address the needs of everything that's on that picture. Everything from the smallest sensor up to the biggest server can be Arm-powered, and we can drive energy efficiency up as a result of that. Now, later on, Pete's going to talk a bit more about mobile. In fact, later on from that, Tim's going to talk about a TAM of another semiconductor market that we didn't even go into here. There is more to come as well.
Before we get to that, I'm going to hand over to Lakshmi, who's going to talk about enterprise infrastructure. Thank you.
Thank you. Morning. If you look at the word revolution in the title, you might think that we're being a little strong on that word. Actually, I fundamentally believe that there is a revolution unfolding in the infrastructure space, and Arm and partners are poised to enable the entire value chain to reap the benefits of this revolution. The cloud is transforming consumer behavior and driving new business models, and certainly driving technology. If you look at the clouds that are depicted here, you're showing public cloud companies like Facebook, Google, LinkedIn, and you also were showing some infrastructure as a service companies that offer cloud compute for rent. Consumers are completely dependent on these public cloud services. For example, my kids, they do their homework on Facebook and Google Hangouts because clearly meeting people face-to-face is so last century.
Businesses are also delivering services using infrastructure as a service provider. For example, Netflix and Comcast deliver services to a broad range of consumers using infrastructure as a service for their compute. A lot of these new businesses are actually dependent on the cloud for revenue and for their reputations. IT decision-makers are also under pressure to be able to deliver services at the cost model that their CFOs are seeing from these players like Amazon Web Services and others. CIOs are looking to transfer a lot of their end application services like storage and email and other things over to the cloud. Yet, in some cases, for example, in the financial industry, they're looking to be able to leverage the economies of scale that they see from some of these hyperscale cloud providers into a private cloud that they own and manage.
One thing is clear from all of this. There is a disruption coming in the ecosystem in the IT supply chain. A lot of the deployments that are happening are based on open source software running on Linux. They're driven by community projects like Hadoop or Redis as examples. From a supply chain perspective on the hardware side, a lot of the cloud providers go directly to the ODMs in Taiwan, where they specify exactly what they want to see in a server. Clearly, they're driving technology into new places. Most IT decision-makers plan to move a lot of their storage or web and email services to the cloud. End-user spending on public cloud services is expected to exceed $250 billion in 2018. The resulting equipment spend on the cloud is expected to be around $80 billion in 2018.
You can see that there's so much revenue and services that are being generated from the cloud that there's a lot of innovation going on in this space. Let's talk about the network infrastructure that's needed to be able to deliver these kinds of cloud services to end users. Believe it or not, this is actually a simplified representation of network infrastructure. Focusing on mobile infrastructure, actually, a lot of end users are accessing services like Facebook on their mobile devices. If you were at home, you're probably covered by your local macro base station that's been deployed in your suburb. If you were in a train, you would actually be covered by a combination of small cells along the train route and the local macro cell that's covering that area.
All of this data gets aggregated and then transmitted over a microwave link into an evolved packet core network, and then eventually gets to the cloud. From a broadband access perspective, a lot of folks are accessing content like Netflix at home on a tablet or, say, on a Wi-Fi-enabled TV, and typically that gets routed through a Wi-Fi router in your home to a cable box or a DSL modem, which then gets aggregated at a fiber termination point or a DSLAM and sent into a Carrier Ethernet network. On the enterprise side, we talked about a lot of enterprise CIOs looking to move some of their services to the cloud. Today, a lot of enterprises maintain their own data center. Users are typically accessing services within the enterprise campus using a wireless LAN infrastructure. There is storage.
There are also specific appliances, like security appliances that do intrusion prevention or content filtering, as an example. Those are then connected using a local LAN network, which then gets connected into a provider into the cloud. As you can see, this is quite a complicated process or complicated deployment. In order to be able to address the service requirements and address the complexity here, people are looking more and more to adding intelligence at the edge. Simon talked about the exponential growth in data, and this presents some challenges for network operators and providers. We also talked about the billions of internet connections or IoT connections that are coming, and it really drives the network in terms of requirements for high-performance networking equipment. A lot of the operators are looking to deliver a better user experience for their subscribers and end users.
What we've seen happen also is that a lot of the cloud services are now very much dependent on network performance for their end-user experience. Cloud services are also increasing the amount of low-latency traffic that needs to be serviced in the network. Again, that's creating a demand for high-performance networking equipment. Cloud services also require a lot more control processing. For example, if there are contextual services, they need a lot of real-time decision-making. For example, if I'm in Cambridge and I want to find my favorite pub or curry house, I look it up on my phone using Yelp, but there's a lot of data, disparate pieces of data that have to be processed, and decisions have to be made in terms of what data gets presented to me, the end user. You require more real-time network management and decision-making closer to the user.
Operators are also under significant pressure from a CapEx and OpEx perspective. They're looking for deployments that can scale to meet their needs but also allow them headroom in terms of processing capability. They're looking for flexible and yet standard processing architectures. Operators are also exploring initiatives like Software-Defined Networking and Network Functions Virtualization to be able to get to a more scalable and manageable solution. In fact, Arm is a part of the ETSI working group for NFV, and we're working with operators and the entire ecosystem on NFV. These are all still quite early. Still, NFV and SDN are in the exploration phase, and operators are planning for it and exploring it with various proof of concepts. Operators also want to be able to rapidly deploy new services to their end customers and provide ubiquitous access to their end users.
They also want to deploy applications as a service. For example, in a specific geography, if they want to be able to spin up traffic management services or security. Historically, when they had to do that, they had to deploy specialized boxes, which by the time they deployed those, they had already outgrown the capacity that those were servicing. They're really looking for a flexible architecture that can get them to the next level. What I really want you to take from this is that you're starting to see a requirement for more intelligence at the edge of the network. Operators are looking for scalable, but yet standardized processing architectures. You're going to see more and more demand in terms of new applications and new services that are being driven a lot from cloud-based applications.
I want to focus a little bit on the mobile infrastructure opportunity and specifically wireless access. Operators need to be able to deploy solutions that range from small cell base stations or femtocells all the way through macrocell base stations, and then centralized radio access networks, which are coming in the future. They also need to be able to deploy different combinations of this equipment to be able to provide a better coverage across the network. In order to be able to do that, they need unified approaches to software management, and they need instruction set compatibility all the way from the small cell up to centralized radio access networks. There are a lot of complex protocols that need to be processed. In order to be able to do that, they need heterogeneous processing capability.
They also want to preserve their software investment going from platform to platform. We're also starting to see them deploy virtualized applications. For example, in LTE deployments, they're encapsulating 2G protocol packets to be able to service legacy requirements. Operators need a variety of form factors. They need to be able to have energy-efficient platforms to meet their CapEx and to meet their OpEx goals. They need optimal architectures for these different kinds of processing that are happening. All of that has to be underpinned by scalability. The two major processing types that we see here are baseband processors and digital radio, and these are typically different silicon on chip form factors or SoCs. In a small cell or a femtocell, you'll see 1 to 2 chips going into each of these femtocells or small cells.
When you go to a macrocell base station, you start to see tens of chips being deployed in each of these base stations that's going out there. When you get to centralized to C-RAN, it's going to be hundreds of chips. The kinds of cores that we see and the configurations that we see being deployed in this space, we'll see 32-bit Cortex-A15 and A9 in the small cell. As you start to migrate up in terms of performance requirements, we see various multi-core configurations. We'll see Cortex-A53 coupled with specialized processing elements to do packet processing and other things. We'll see Cortex-A57 doing the heavy lifting control plane processing. In this space, we see it evolving all the way to 32-core and beyond, especially when you start to get to the C-RAN space.
Looking at the enterprise networking opportunity, there are a wide range of design points that are needed to cover the range of performance in these applications. You have switching and routing, security appliances, like I mentioned in terms of intrusion prevention, intrusion detection, and other specific appliances. Of course, wireless LAN infrastructure. There are 3 kinds of processing that typically happens in these applications. You have packet processing, which as the name implies, is processing the packets and the lower layers of the protocol stack. Network processing, that's doing applications like network address translation. Management and application processing also sometimes known as control processing. There are various combinations of these types of processing depending on the box that's being deployed.
In a switching and routing platform, you'll have pretty heavy on the packet processing and the network processing, but there will be some management and control plane processing. In security and appliances, it typically tends to be more management and application processing. In wireless LAN, you kind of see a combination of all of these. Here, the SoC platforms that we see being deployed are going to be based on Cortex-A57, several Cortex-A57s and multi-core configurations to do some of the control plane functions. Cortex-A53, again, coupled with specialized processing to do some of the data plane functions. Cortex-A57 in the security and appliances. These are going to be pretty heavy in terms of the core counts to be able to do security and appliances.
In the wireless LAN space, we see our Cortex-A15 coupled with Cortex-R and Cortex-M processors for the processing functions. In the wireless LAN space, we see about 3 to 4 chips typically in each of the wireless LAN devices. In security and appliances, it's going to be a couple of chips. In switching and routing, depending on the throughput that you're trying to service here, it'll range from a couple of chips all the way to tens of chips per platform. What you should take away from this is that there is a lot of intelligence being deployed in these systems, lots of complex processing. Clearly, intelligence at the edge is going to be needed to be able to service these demanding requirements. What does this mean for the data center?
We've talked about some of the cloud workloads that are emerging in terms of hyperscale deployments. This is really shifting the characteristics of the data center. Most of the bottlenecks that are seen in the data center are typically on the networking and IO side, not on the CPU side. Historically, the way data center operators were trying to solve these challenges was to scale up, really go for performance at any cost. Now they're looking more at throughput, they're looking at workload optimization, and it's really about performance per watt, per dollar, per cubic foot. They're looking at more scale-out deployments with smaller and smaller cores. A lot of these workloads are also based on Linux or open source applications, it really opens up their ability to look at new and innovative architectures to help solve some of these challenges that they're facing.
Let's talk about some of the workloads that are being deployed on our, or planned to be deployed on Arm-based SoCs. Most of the workloads that we're looking at from an Arm perspective are typically about balancing CPU performance versus IO and memory performance. Historically, that wasn't the case. It was out of whack. A lot of the applications that are going into the data center, they're looking at how do you balance that performance to achieve better overall throughput. For example, one of the applications that we're showing up there, cold storage, typically in that application, the CPU is idle 80% to 90% of the time. You really need to spend more of the processing on the IO and the memory side.
If you had an integrated solution that had a storage controller integrated onto the SoC, you would actually achieve better throughput and better performance per watt, per GBP. Some of these other applications, like big data scale-out applications, for example, you would look at a different kind of accelerator to be able to process some of those applications. There are a variety of different applications where you might need different heterogeneous processing capabilities to get the scale and performance that you need. Also think about the fact that a lot of these mega data centers, as they're called, are actually deploying hundreds, thousands, and in some case, millions of these servers.
The effect of optimizing for a given workload, and by the way, a lot of times it's a print and repeat in terms of the workloads that are being deployed across these thousands of servers. Think about if you optimize for a given workload and you're actually deploying it across thousands of servers, it really amplifies the benefits that you can get from workload optimization and total cost of ownership. Let's think about how Arm is going to address the various workloads over a period of time over the next 5 years. Today, there are several 8 and 16-core 32-bit devices using our Cortex-A15 as an example that are shipping in the market. We're starting to see 8 and 16-core 64-bit devices coming to the market focused on the service space, and our architecture can actually scale to 64 cores today.
Over a period of time, I think through 2016, we start to see core counts increase to where it'll be 16-48 cores. By 2018, we expect to see 64 cores as this evolving sweet spot. The workloads that are suitable for the Arm solutions in the market today will be web tier applications like static web hosting or dynamic web hosting, cloud applications as a service, I already talked about storage, big data scale-out applications, and offline analytics. As we start to get 64-bit software maturity, we will be able to address a broader set of applications. It's not just about software maturity, but it's also about our partners delivering higher and higher performance solutions. We'll be able to address applications like HPC, telecom management, as some examples.
Out in time, by the 2018 timeframe, we see a path to where we would be able to address enterprise server applications, online transaction processing, and applications like desktop virtualization. SoC platforms are really enabling this infrastructure revolution that I talked about. OEMs and end users are really wanting to deploy software-defined platforms. For that, you need higher levels of innovation. You need flexibility, manageability, scalability, and these need to be energy-efficient platforms to be able to integrate as much as possible onto the SoC. From the server space, what we're starting to see is that people are moving away from this one-size-fits-all model and looking to add more specific IPs. I talked about some of the accelerators that are possible with storage, networking. We're also looking at, from a server perspective, people thinking about adding cloud workload accelerators, as an example.
From the networking space, they need diverse processing capability to handle the complex protocol processing that needs to occur. Because they're wanting to move to a more software-defined world, they're looking to the Arm CPU as an example of a standardized processing architecture to where they can marry the benefits of server-class processing capability with these accelerators. Clearly they're looking for scalable solutions. They need rapid integration of new IP to be able to deploy these as quickly as possible. Arm provides the framework for integrated SoCs. We have our compute clusters, which scale to more than 64 cores. We have system IP, like the Cache Coherent Network that Simon talked about earlier, and of course we have high performance memory interfaces and memory controllers that we provide. Our partners are really able to rapidly integrate specific application-focused accelerators.
It could be network packet acceleration, things like crypto functions, as an example. Or they could include GPU capability or DSP capability, and they're really able to innovate around this Arm SoC platform. One of the things I want to focus on is the fact that the Arm SoC bus standard, the AMBA bus standard, is the industry standard for Arm-based processors. A lot of third-party IP companies develop IP for the AMBA bus standard first because their market, in terms of hundreds of semiconductor companies developing IP for the AMBA bus standard, is quite large. It really accelerates the availability of IP for the partnership. What does this really mean for the Arm partnership?
AMD has been talking quite a bit about their ambidextrous approach, and Andrew Feldman, who's the GM and VP of their server business unit, has been actually quite vocal about the benefits of developing solutions with Arm. He talks about how it's a tenth of the cost and less than half of the time, but really, the quote says it all. I want to talk about some progress since I was here last year. Actually, last fall, we had several companies announce plans for 64-bit Arm products, and actually moving to Arm in general from some legacy architectures. A lot of these folks were historically developing on MIPS and PowerPC, but they're now offering solutions based on the Arm architecture. Altera, Broadcom, Freescale, and IBM announced their usage of the Arm product for networking and infrastructure applications.
AMD announced their Hierofalcon project, which is for embedded and networking applications, and their Seattle processor, which I happen to be holding in the palm of my hand, for server applications. We also announced the creation of OpenDataPlane. If you look at the networking space, OEMs are striving to have a common software base, and they want to be able to migrate from their legacy architecture. OpenDataPlane provides a common platform for them to be able to do that. At Mobile World Congress this year, we had several partners that were demonstrating virtualized network applications. We had a TI cartridge in a HP Moonshot system that was actually coupled with software from our partner Enea that was actually demonstrating a carrier-grade management system for IoT endpoints.
We also had Avago LSI's 16-core Cortex-A15 based Axxia solution that was demonstrating a virtualized evolved packet core application. Earlier this year, we announced the server-based system architecture at the Open Compute Summit, where we had 14 companies, including software companies like Microsoft, Red Hat, Canonical, SUSE, and Citrix, and OEMs like HP and Dell, that collaborated with our silicon partners to develop a standard that allows them to simplify their software development and deployment on Arm server-based systems. A couple of weeks ago, we actually had an event in Austin where Canonical came and demonstrated their long-term server product, Ubuntu, on the Arm 64-bit architecture. Actually what was really exciting was that they demonstrated that 40,000 packages had been ported over to the Arm 64-bit architecture, and there were hundreds of applications that they demonstrated running on Arm servers.
The even more exciting thing was it wasn't just about the applications, but it was about how easy it was for them to provision and spin up server instances using their Juju MAAS tools. A couple of weeks ago, Red Hat actually demonstrated on the AMD Seattle-based processor, which is the one here. They demonstrated their server developer preview on this product, and that was also very exciting. We have partners like HP, Dell, and MiTAC that are poised with systems, and we expect to see additional proof points on things like NFV in the coming months. Of course, more news from our partners on their 64-bit efforts. As I talked about, there are several partners. You can see the number of companies that are investing with Arm for infrastructure. We have a vibrant silicon ecosystem with a number of partners that you can see there.
We have OEMs like Cisco, Nokia Solutions Networks, Huawei, and we have end users like Facebook that are developing on Arm. From a software perspective, you can see we actually have quite a vibrant software ecosystem as well with some of the partners that I mentioned earlier. All of these folks are working together to realize a really large opportunity from a silicon perspective, which is GBP 20 billion in the year 2018. The Arm ecosystem is innovating on multiple fronts. There are demanding requirements from networking and data center infrastructure, and you need solutions that can scale across a variety of design points. In order to be able to address these needs, you need an innovative ecosystem that has a lot of choices. Arm's vibrant and vast ecosystem is really focused on enabling our partners to deliver to the demands of the cloud economy. Thank you.
We will have our coffee break now, which is coffee is next door. Please be back in 15 minutes to listen to Pete.
I'm not going to smile. Okay, we ready? Welcome back. I'm going to spend the next half an hour talking you through how we're doing in mobile and how we're taking the momentum we have in mobile and moving it into other areas. I'll wait until everyone Okay. Our business model has enabled waves of innovation in the mobile space. What this chart shows you is the ramp of, Simon talks about the early Arm7 going into voice-only phones. Then we had the ARM9s, the Arm11s going into feature phones, and then recently the rise of the Cortex in the smartphone. You can see that we have moved gradually over time, people from the ARMv4, ARMv5, ARMv6 architecture to the Cortex-ARMv7 architecture. That's what's enabling the smartphone growth.
We have shipped over 30 billion devices, or our partners have shipped over 30 billion devices into the mobile space, and we've had about 100 partners shipping silicon. A very good success story in the mobile space. That continues. Cortex-M is getting widely deployed into IoT-type devices, and some studies out there say there are going to be tens of IoT devices per person, 100 IoT devices per person. I've actually seen one out there which says that there are going to be 1,000 IoT devices per person. That's quite nice in terms of royalty growth. I think it's also very nice in terms of the kind of opportunities it's going to open up for us to control our own lives, get more data, see what's going on out there. We also have Cortex-R deployed in baseline modem-type technologies.
More and more we're seeing integrated application processors consisting of a range of A-class processors, R-class processors, and M-class processors. There's multiple processors per application processor chip from our customers. Of course, we have ARMv8-A starting to deploy out there. I'll show you in a minute how that deploys and how we see that deploying throughout the partnership, and GPU compute. Both of those are enabling more and more compute capabilities, heterogeneous compute capabilities to go to mobile devices to enable new use cases. As we discussed earlier, smartphones are segmenting. With any market, they're going to segment. The high-end premium mobiles, you can see about a 5% CAGR over the period. Mid-range mobiles taking off quite nicely. Then just explosive innovation at the entry-level space. Simon talked earlier about the fact that there are 2 billion smartphones out there.
There are 6 billion mobile subscribers out there. We have another 4 billion people to move to smartphones, and we can see that happening now. We can see the less than GBP 150 area taking off. We can see mass deployment out there, and we can see some amazingly innovative technology in that space. At Mobile World Congress a few months ago, we saw the Firefox OS phone, GBP 25. GBP 25 in your hands, and you have a full smartphone capability. That's going to markets such as Latin America, Indonesia, and Africa. These markets or these new markets are not only mobile first, they're going to be mobile-only type markets. The mobile phone is going to be the only phone these people use. Wired wireline networks aren't going to get there. They are going to be mobile-only.
What's happening is these are enabling new and innovative services on top of the mobiles, on top of the smartphones. Anything from improving disaster response through to mobile banking. I read one article the other day where it said that in some sub-Saharan countries mobile banking outstrips non-mobile banking. Again, these are mobile first economies. Even banking in those cases is mobile-centric. My personal favorite, I think Simon talked about the new and innovative applications these types of devices are going to enable. My personal favorite is the medical tools in Africa. A great example of that is using the camera on the smartphone to detect things like glaucoma and cataracts in people's eyes. Really cool. Forget the money. Well, I shouldn't say that here, but really cool in terms of deployment and the benefit that it's bringing to people's lives.
Now, a great example of where the diversity that we offer with our business model has been exploited is China. China, we have about 10 application processor partners delivering silicon into China. We've got about actually 70 licensees in China, but I'm just going to focus on the application processor guys. That's 10 partners delivering silicon into China. What that's enabled is it's enabled local OEMs to have the majority of the market. If you see the breakdown of the chart in the middle, it's broken down in terms of the various suppliers, and you can see that local Chinese suppliers are the majority of the suppliers into that market. People like Xiaomi have gone up to third in terms of smartphones, or the provider of smartphones in that segment, and they are expanding into places like Latin America.
They're expanding into places like Singapore, Hong Kong, and the U.K. Our technology going in, enabling local manufacturers, local OEMs to take control of their market. A range of price points and form factors coming from that. One particular semiconductor manufacturer, or one specific semiconductor manufacturer, based on one chip, we've seen 150 different price points and form factors coming from that one chip. Again, that's one chip from one semiconductor manufacturer. We have lots of chips from each of those guys, and we have more than 10 going into that space. You can see a vast explosion of innovation in that space, leading to tablets and smartphones at all different price points. In China, that's mirrored by the ecosystem. There are some logos up here. We have hundreds of ecosystem partners in China.
Ecosystem partners ranging from game engine developers who are working very tightly with them to optimize their game engines to run on our CPUs and GPUs. Middleware providers where, again, my engineers are out working with them to optimize their solutions for the Arm architectures. App stores. I think a pretty interesting point here is there are over 200 different app stores in China. Just think about that. 200 different app stores, a million apps per app store, all serving the local market. All of these, we're working with them to optimize that content for our architecture. Another example of how software optimization and software innovation is helping expand our reach is premium mobile computers. You've seen recently things like Microsoft Office and Photoshop being ported to Arm-based devices.
What this is doing is this is increasing people's usage or increasing business usage of those mobile devices. Over time, those will become the main compute platform. In fact, I think it wouldn't be stretching things too much to say that the PC you own now is the last PC you'll ever own. You can see that in some cases, 70% of Salesforce users are spending all their time on tablets. With the advent of these productivity applications, I think that just continues. These productivity applications are enabled by the new technology that we're putting out, new desktop-class CPU and GPU technology that we're putting out. I think the graph at the top right shows you that internet-connected mobile devices outstripped PCs and tablets in about 2013. It's already taking off.
It's already starting to dominate that particular space, I think that just continues. These features are being introduced firstly at the premium end of the smartphone and mobile space. If I take you through a little journey, I'm going to take you through a little journey from 2011 to 2018. Past to the future. 2011, which is only three years ago. If you look at the premium devices there, they were dual-core ARM9. They were GPU-enabled. They had color screens. They had reasonable connectivity, and okay multimedia-type features. Pretty high premium-class devices. The mid-range, we had the Arm11, we had the ARMv6, Arm11 devices. No GPU, really no color screens at that point. The entry level, we had the ARMv5, ARM9 devices. Again, no GPU. That's only three years ago, and just think about that.
I mean, three years ago, we had a mix of architectures. Last year, we actually had moved everyone over to ARMv7. Everyone had moved over from the existing architectures. We were consolidating on ARMv7. We'd done that by a mixture of technology we developed and also extensive work with the ecosystem. We have a lot of people going out and working with software developers, explaining why the technology is great, what they can do with it, providing them with software enhancements so that they can move quickly. This year, 2014, we have ARMv7-A octa-core and hexacore devices out there based on A15 and A7. We have high-end GPUs based on Mali-T760, high-end connectivity, and really powerful multimedia features. You can see that the features from the high end have migrated down. The mid-range smartphones, again, this year they're based on ARMv7-A.
Towards the end of the year, even in the mid-range, they'll be based on ARMv8. High end at the end of this year will start to deploy ARMv8, both Arm Cortex-A57 and Cortex-A53 designs. The same happens in the mid-range. That happens this year. ARMv8 applies high and mid-range this year. At the entry level, we're on ARMv7. We're still on Cortex-A7 devices and Mali-400. You can see the entry-level smartphone of this year has more features and more capabilities than the high end from 2011. During this period, because we're adding more and more technology, because our users are adding more and more technology, the ASPs are remaining broadly flat. The premium space SoCs are a GBP 15, GBP 20 range. Mid-range, we're talking GBP 10-GBP 15. At the entry level, they are GBP 5 or less.
ASPs are remaining broadly flat because we're adding more and more technology to those spaces. Because we're adding additional Arm technology to each of these spaces, our royalty percentage of those ASPs is going up. If we forward to 2018, we see ARMv8 going everywhere. ARMv8 will go everywhere from premium-end, mid-range, entry-level. By 2018, the entire smartphone and tablet market will have gone ARMv8. That's enabled by a couple of things. That's enabled by the fact that there is the demand out there for that. For example, we can already see high-end mobile games requiring 2 gigabytes of memory. The big thing about 64-bit is it allows you to break through the 4 gigabyte barrier. Right now, people on high-end games in mobile are on 2 gig. The gaming guys are demanding 8 gig out of a smartphone.
Gaming is already driving a move to 64-bit. That just continues. The other thing that helped with the move is we're offering 64-bit at various price points. We have the Cortex-A57 ultra-high performance core, and we have the Cortex-A53, which is a very capable core, but it's offered at a lower price point. There are numerous 64-bit price points being offered. The entire market, I think, will go 64-bit in this period. It's not just smartphone and mobile. As I said, gaming guys are demanding more and more memory. They're already bursting through the 4 gig barrier. That's going to demand 64-bit deployment into things like DTV, into set-top box, and into over-the-top boxes. We're also seeing users in those spaces demanding more performance, and 64-bit gives you that. We see 64-bit going everywhere across the mobile space and into the home segment.
This rapid growth in terms of mobile, in terms of deployment, is really enabled by mobile SoCs. I'm going to spend 2 minutes on those. I won't go into too much detail on those, but mobile SoCs are the most complex systems ever built, other than humans, of course. They have billions and billions of transistors in them, and literally billions of transistors. They have tens of millions of lines of software in them. If that wasn't difficult enough, the entire market is now starting to move at China speed. By China speed, I mean that the timescales for concept to SoC to product deployments are getting shorter and shorter. The SoCs are getting more and more complex. The software that goes on top of them is getting more and more complex, and their timescales are getting shorter. It's just more interesting.
What that means is we're being requested by our partners not just to give them CPUs and GPUs and tell them to integrate it themselves. We're being asked to provide all the rest of the IP around it. All the interconnect IP, all the memory controller IP, video display, and to make sure that all of that is tested, is optimized, and is delivered to them fully. They're relying on us. We're getting demands back, particularly actually from some of the Far East customers, to deploy more integrated solutions. We're doing that. At the same time, we're making sure that all our stuff works together, but it also has to work with all the IP the customers are going to put around it to differentiate. All the hardware IP they're going to add, all the software IPs that they're going to add.
Again, we have to make sure it works with that. Then all the third-party ecosystem. We have the widest and most effective ecosystem in the industry, and that's because we work at it. We have guys out there working with the semiconductor manufacturers. We have people working with the EDA and third-party IP guys. We have a ton of people working on the software ecosystem side of things, right? Hundreds of engineers literally working on this ecosystem. It doesn't just come free. We have to work at it. I am not going to go through this slide in detail. I will ask questions later on what DVFS means, but I won't go through it right now. This is trying to show is as we deploy more and more technology, performance goes up, and that's great. There isn't much more battery power left.
We have to keep energy usage flat. We do that through a range of techniques, a range of techniques at the semiconductor side, a range of techniques in the IP itself, and then a range of techniques at the software side. The only thing I'm going to talk about here is our big.LITTLE technology. Our big.LITTLE technology has been deployed widely across the partnership, and what that enables people to do is use a mixture of big CPUs and little CPUs. Now we're starting to extend that to also include things like the GPU. You can now balance workloads and usage across big CPUs, little CPUs, and the GPUs. This just continues. All the techniques we're developing have to keep continuing if we're going to deliver that increased performance to people.
Hopefully I've shown you how the smartphone and tablet market is large and growing. It's going to go from GBP 13 billion-GBP 20 billion over the period. The new technology that we're adding, the migration of features from premium to mid-range to entry-level, keeps the ASP stable within each category. The fact that we're adding new technology, new technology in terms of ARMv8 over ARMv7, in terms of graphics, in terms of video, in terms of POP, means that our royalty per chip increases over that period. Combining those factors, we can grow royalty revenue from the mobile space 15%-25% CAGR. That was mobile. As Simon pointed out, in the mobile space, we have added newer technology.
We have everything from Cortex-A7, behind that there's a Cortex-A5, Cortex-A9, Cortex-A17, Cortex-A53, et cetera, we are continuing to add more and more technology in that space to serve those particular markets. We also have the same in the other markets. Enterprise infrastructure and embedded intelligence, we already have a range of technology there. Things like the Cortex-A57. Cortex-A57 was designed from the ground up to serve both the mobile application market and the enterprise infrastructure market. It has a number of features in it which are incredibly useful for enterprise infrastructure and actually are being used in the mobile space now. We continue to add new technologies in those spaces. Lakshmi talked about our CCN networking. We have new technology coming in enterprise infrastructure as we engage more and more in that space.
Embedded intelligence, there's just so much to talk about. I think we'll do that at the break. We have a range of microcontrollers in that space. A5 and A7 are becoming the de facto standard in wearables, and clearly, we have all the real-time controllers as well. What this means. The new markets we're entering and the new technology we're deploying extends our license opportunity. What you can see here is gradually over time, the Cortex and Mali licensing dominating the chart. A fall off in our classic processors, but Cortex sweeping everything before it. In the new market areas, last year we had a number of companies in mobile computing taking both our Cortex-A50 series and Mali-T700 for mobile computing. We had 10 licensees of Cortex-A just in the enterprise networking and server space.
Not mobile licensees, but guys who are taking our technology specifically for enterprise networking and server. Embedded intelligence, we have a whole bunch of existing customers, but 22 new customers in that space. Customers that were new to Arm, taking M-profile largely for MCU and IoT devices and for low-end wearables as well. License opportunity extending. What you can see along the bottom is ARMv8-A licensing, and it's gone pretty well. Given the amount of time and effort my boys and girls invested in it, I'm glad it went very well. We're up to about 44 licenses just now, which is great. I think we have a lot untapped. You can see our ARMv7-A, we're about 150 licenses. A lot of additional potential licensing to go in ARMv8. That's existing markets, let alone new markets.
I'm going to spend a little bit of time just talking through the licensing options we have. We have a range of licensing options, largely because we have a range of customers. We have everything from the very big multinationals with multiple divisions, all the way down to new startups. Simon was talking about the guys at the Maker Faire, so very small startups. We have a range of licensing options. I'll cover architecture licensing in a minute. Ditto for subscriptions. Perpetuals are largely for large companies who want access to the technology for multiple designs. They get access to technology. They can have a Cortex-A57 perpetual license. They can do as many designs as they want forever. We do term licenses largely for mid-range companies, so companies who want access to the technology for three years or five years.
Again, they want to do a number of designs around that. We do quite a lot of single-use licenses. Those are, as I said, the startups. The guys who go, "I know I can do one chip. I want access to technology, and I want to build it around Arm. I want access to technology for that one single design." Quite a lot of single-use licenses, and you can see over time, a lot of those guys extend up as they get successful. If I cover two of those license types in particular, if I go through architecture and subscriptions, these are both examples of where partners want long-term relationships with us. They don't want transactional relationships. They want very long-term relationships. Architecture license is a great example of that.
That's where customers get a license to our technology so they can build an Arm-compatible processor. The reason they're doing that is they want access to that ecosystem, the ecosystem I talked about. All the software, all the tooling, all the third-party software that runs on an Arm processor, they want access to that, and they want to build an Arm-compatible CPU that really goes at the extremes of our roadmap. They want to do something which we'll probably get to in three to five years when the market develops. They want to do something like that now. I should note that all of these guys have to pass Arm certification. They're getting access to the Arm ecosystem. That's the value they get. We have to make sure that they enable the Arm ecosystem when they run on the Arm ecosystem.
This is really only for very large companies. The license fee itself is large enough. The engineering investment they have to make on their side is enormous. They have to put up complete CPU design teams, validation teams, and pass all our qualifications. It's a very large investment for the licensee. I think the other thing that's worth noting about this is the royalties that we charge architecture licensees are broadly the same as if they took our own CPUs. You have to do a very large investment, and your royalties are still going to be the same. I think it's also worth noting that most architecture licensees do deploy Arm CPUs around their own architecture core as well. You'll have a large architecture licensee core maybe in the middle of the SoC, and then Arm implementations around it as well.
The other area where people are engaging for a long-term relationship is the subscription model. Subscription models enable customers to take access to an entire range of Arm technology over, say, a five-year period. We segment it so we can say you can have all ARMv8 high-end processors and anything that we come out with in that period, or all M-profile processors, or all Mali GPUs. You can see here there's a range of customers we have up here who have taken subscription access to our CPUs. Quite a few of them have also taken access to our Mali GPUs, and increasingly now, some are taking access to our POP IP. It's a very good model to get long-term access to technology.
It gets past the project going, "Well, I need an Arm core, but I'm going to have to go and talk to purchasing, and it's going to all be very difficult." Well, actually, you have everything. Once you sign this, you get access to all the Arm technologies. The design-ins are really easy. Our relationship with the customer, the engineering teams with the customer just changes dramatically. They're really good things. Our licensing base continues to expand. We've added one to two subscription licenses per year. Of course, we have renewals that go along with that. I think the other point that's of particular interest to me is that 25% of our processor licenses are with new customers. That new customers to Arm, a quarter of our licenses are with them. That's quite interesting, and that's because we're licensing non-traditional customers.
Not just the semiconductor guys, but OEMs, software companies, service operators, et cetera. Licensing, I think, for us, is a precursor for future royalty growth. This chart is really showing you how cumulative licenses of Cortex have taken off, and then a few years later, you can see the unit shipments. We really see licensing as a precursor for future royalty growth. You license it, a few years later, you're going to get the unit. In terms of v7a deployment and v8a, you can see that they're moving at more or less the same cadence. From processor licensing to 15 licenses signed to 1 million reported shipments, about the same in v8 as v7. I personally think by 2018, v8a shipments will have exceeded the 2 billion units. That's what we did with v7a.
We have those 4 billion other smartphone guys to turn over as well. Hopefully, in that period, we'll have turned them over. We talk a lot about licenses. We do 20 to 25 licenses each quarter. What we try and do is show you in the analyst community how important some of those licenses are, and we've been doing these charts for the last 3, 4 years. What we do is we say, here are the key customers or key divisions of customers that we have to enable with Arm technology in order for us to get an 80% market share. We do these at each earnings announcement. This is what we had in 2010. In 2010, we had 37 guys shipping mainly Arm-based chips, 15 shipping some Arm-based chips. The rest were largely based on their own CPUs.
Then 23 guys who didn't have any design wins weren't public, didn't want to use our technology at all. That lot of customers got us a 25% market share in the CPU space. Since then, we've added new customers because we have enterprise networking, we have IoT, we have wearables. We've added a number of customers that we're tracking in that space. We're now up to 100 different entities. Either companies or divisions of companies that we're tracking. You can see now we have about 40 shipping mainly Arm-based chips, and only 17 who don't have an Arm design win or aren't public. That set of customers gives us a 35% market share. You can see that in Q1, we announced that some of those reds turned color. AMD and Freescale announced their plans.
We do also have another two who we know are migrating to Arm-based designs, it's their secret, I can't share it with you. We've got, and I worked it out in my head, 87 companies now out of that 100 moving to Arm. If all 87 customers executed on their plans or their plans to deploy our technology, over the period, we'd get a 70% market share. We'd grow from that 35% I talked about in 2013 to 70% over the period. Of course, those 13 are very important. They're very important to me. They are 10% of the market. They are very important, and my job is to turn those guys over. I'm working very closely with them to determine the right time for them to adopt Arm technology.
I don't think there ever is not a right time for them to adopt Arm technology. It's just a matter of time and me doing my job properly. Let me just summarize. We expect mobile devices to continue as a major revenue driver for us. We've got a high share of a growing market. ASPs, due to deployment of new technology, are staying stable. Our royalty percentage, because we're deploying additional technology, is growing over that period. We're well set up for markets beyond mobile. Lakshmi talked about enterprise infrastructure and service, but also consumer electronics and embedded intelligence everywhere. We are getting a lot of partners who want longer-term, deeper relationships with us. I think that's great, and we have the models to enable that. The new IP that we have continues to attract new customers and new types of customers. It's all good.
Well done. Okay. Thank you for listening. I think I'd like to invite Tim up for his last investor meeting.
Okay.
Okay.
Good morning, everybody. The good and bad news, depending on how you do it, is this is definitively not my last investor meeting. In fact, the announcement of my demise in 12 months' time, which was actually made on the 1st of May, already feels like a very long time ago. I can assure you that we are deeply embedded in business as usual from now until early next year. You'll be seeing a little bit more of me yet. We've heard a lot this morning from my colleagues about Arm's extensive opportunity, both in mobile and in new markets. We've been introduced to a little bit of depth on some of the technology.
I think what I'd like to do now is just sort of pull the threads of that together from a financial sense, because I think I know that that is something that interests quite a large proportion of this audience. That is what I will seek to do. We'll have a quick reminder of sort of where we are on our financial journey, 23 years into the project, if you like. Then we'll look at drivers of license growth, drivers of royalty growth. Not going into huge detail because I think Pete's touched that, and we've looked at a lot of detail of the markets. How our investment needs to change or not to access this opportunity and what that will mean for profitability, margins, cash generation, returns to shareholders in due course.
That's what we'll look at over the next 20 minutes or so before we go over to questions. A very quick reminder. This is the 2013 P&L. It's a snapshot. I don't want to go through it in detail, but I think suffice to say, license revenue makes up 40% today, 23 years in, of Arm's total revenue. Royalty is 50%. I personally find it very encouraging for the reasons Pete was saying around licensing is a precursor to royalty. I find it very encouraging that license revenue has grown more strongly than royalty revenue in the last three to four years. I think we will, in due course, revert to the more normal relationship of growth between licensing and royalty as we look forward, and we'll touch on that later.
I think this bodes very well for our medium to long-term royalty opportunity that license revenue has grown so strongly. If you looked at this P&L about three years ago, actually royalty was a higher percentage of total revenue then than it is now, which is very good news. From a margin standpoint, the business is obviously moving close to 50%. We'll look at how that's developed over the years, but suffice to say, three or four years ago, it was in the early 30s. I'm sure most of you are aware we've gone through quite a long period of investment, probably higher than average investment by Arm standards over the last three to four years, and yet the margin has increased significantly in that time.
We all understand, I think, the Arm business model in here, the drop-through from profit to cash is fast, and it's pretty much one to one. We'll talk a little bit about how we plan to manage the cash going forward as well. Again, a quick reminder on where we are on the sort of market share gains. We've looked at a lot of the individual market opportunities. This is a chart that we share a lot with investors when we're on the road. It's a good way, I think, of gauging how Arm is building market share over time across the whole spectrum. What you can see on the bottom right there, in 2013, our licensees shipped a total of just over 10 billion Arm-based chips. That represented about 35% of the total embedded processor opportunity we think last year.
You can see from the little table on the right that that 35% has been growing fairly relentlessly at 2%, 3%, sometimes 4% a year over the last six or seven years. The share has broadly doubled. That's kind of where we are based on what we've done so far. We have an installed license base of broadly 350 semiconductor companies. To those companies, we have sold over 1,000 licenses. That's where we are today. Most of this morning, we have concentrated on, if you like, the big three mobile application processors, enterprise infrastructure, and embedded. I think you can see that on the mobile side, it's high share. We expect to maintain a high share in that space. In enterprise, it's a big new opportunity for us. We have a low share currently, but lots of room to grow.
Similarly, with embedded, a relatively low share, just over 20%, but lots of opportunity to grow. There are other markets. Simon mentioned this. Arm is about layers of growth. It's not just all about application processors in mobile. It's not just about the emerging opportunity in enterprise. We have been working very hard now for multiple years in segments like automotive and making steady progress, which we think will continue. Similarly, in the wireless connectivity space, there are apps processors in many other end markets, not just smartphones and tablets. There are a series of other markets. It's layers of growth. These markets shouldn't be ignored, but we focused clearly on the three on the previous slide. Thinking about licensing. If you look back over the last 10 years, there's kind of been two phases.
There was a phase before the downturn, where Arm's license revenue was growing at high single digits, 9% compound, up until 2008. Since the downturn, it's grown at 19% CAGR. In fact, if you took that number from 2009, where we know the industry went down and Arm's licensing went down a little bit as well, it's been even stronger than that. What we've seen in the last few years is much higher than trend growth. I think that's been driven primarily by the fact that Arm's addressable market, as we've heard this morning, has been broadening.
This has given the opportunity for our licensees, not only our existing licensees, to deploy Arm in more markets, but it's been given an opportunity to more semiconductor companies, both at the low end, if you like, in the embedded space and towards the high end to license Arm technology for the first time. Arm is, at one level, an outsourcing business. The higher the cost, the more the complexity that our partners would have to be spending themselves. The more fixed cost we are saving them, we're replacing it with lower variable cost. They recognize that as the cost and complexity grows, our pricing will go up. Our license fees go up over time. The royalty percentage per chip goes up over time. That has also been driving the licensing growth in recent years.
As Pete said, ARMv8 has been a big growth driver, but it's in the early stages of its overall licensing cycle, much more to come. Those of you who have talked to me about the Arm opportunity over the years would know that I consistently say that in the medium to longer term, we expect licensed revenue growth to normalize back to the rate that we saw before the downturn. We've had an acceleration. We've had a rapid broadening of our addressable market, because I think as we move to steady state and we update the next generation buy-in market, we think it trends back to about 10% growth for licensing. Looking in a little bit more detail at what's happened in the last few years.
The very good news about Arm's licensing is that the backlog has grown much faster over the last 5 years than licensed revenue. Backlog in Arm world is contracted revenue. This is not some sort of discretionary pull down that we may or may not take. This is licensed revenue that's been contracted but has not yet been recognized as revenue. Backlog finds its way into the P&L over time. What's driven the growth in the backlog over the last few years is a combination of the subscription licenses and the architecture licenses that Pete referred to, which are by and large recognized on a ratable linear subscription accounting basis. When we sign one, the backlog shoots up.
As that license is recognized over the duration of the license, the backlog eases down, and then most of our licensees who use this type of model will then renew. They'll either take an architecture license to the next architecture, or they will renew their subscription, and the backlog goes back up for that customer and then eases down again. That's been a big driver. Also version 8 of the architecture has also been a big driver, because big semiconductor companies have been licensing that technology whilst it's been in development, normal Arm lead partner licensing. That revenue goes into the P&L over multiple quarters. That's what we've seen. One of the sort of underpins, if you like, of licensed revenue over the next few years is this major growth in backlog unwinding into the P&L.
Looking forward, as we transition from the 19% license growth that I showed you earlier to the 10% that we think in the future. As we transition there, what we would expect is that the backlog stays broadly around current levels. Clearly, in some quarters, it's going to go down. In some quarters, it's going to go up. We would broadly expect the backlog to stay around current levels as we look forward. Switching to royalty. What is it that's been driving Arm's royalty growth over the last 5 years? Arm's royalty revenue has broadly doubled over 5 years. It's a combination of some industry growth. I think in this period, which obviously includes the downturn, industry growth has been very modest at a 1.5% CAGR.
Most of you in the room would know that, in a decent year, the semiconductor industry is growing maybe at 5%. That's a pretty decent year these days. Typically, it would be in the 3%-5% range. Over this period, if you like, the tailwind that Arm's royalties have had from the industry is about 1.5%. Layered on top of that, we've obviously had the share gains that I showed you earlier. We've also had the increasing percentage per chip that colleagues have talked about as more Arm technology has been introduced into these devices. What that means is that over the last 5 years, Arm's royalty has grown at around 15% higher than the industry. For all the reasons that we've been speaking about this morning, we envisage a world where that type of outperformance continues.
As I say, if the industry goes through a period without major cycles, it could be a 3%-5% growth for the industry. We would expect Arm's royalties to grow at about 15 percentage points ahead of the industry looking forward. One subject that we talk a lot about to investors and, to some extent, to analysts is the average royalty per chip. There is a sort of a general sense that a number like average royalty per chip going down is a bad thing. That may be the case, it may not be the case. I think if you think about the broad spectrum of end markets and chips and chip prices that Arm is being designed into, I think you can quite easily quickly understand that this is really all about the mix of the different growth rates of Arm technology.
You can see in the last four years that the average royalty per chip has been flat to slightly up as the positive impact of, as it says there, ARMv8 higher royalty, multiple processors, increased Mali penetration, driving the average up. Driving the average down would obviously be Arm's penetration of very high volume markets like embedded processing, which are characterized by lower chip prices. What is likely to happen going forward as we break into the markets that we have been talking about? Let's look at an example. If all Cortex-A family processors last year had been version 8, i.e., same number of shipments, but they had all been version 8, then the average royalty per chip would have been about GBP 0.06 across the piece, which would have been good news financially.
If Arm's processor share in microcontrollers had doubled in 2013, i.e., more Arm in lower priced chips, the average would have been about GBP 0.04. Not a bad thing. All royalty revenues are 100% margin, don't forget. If both of those things had happened, lo and behold, the average would have been GBP 0.048, but the royalty revenues would have been much higher. I think we would have all been delighted at many more royalties, same average per. I think we need to be very careful when we sort of assume that average royalty per chip going up is good and average royalty per chip going down is bad. That's not necessarily the case. We are in business here to grow our royalty revenues as fast as we can. Let's switch a little bit to investment. I think it is no coincidence if you look back over the last few years.
We have been going through, as I said at the beginning, an investment phase. We have been increasing our headcount to Arm, both in our R&D teams and in our business infrastructure more quickly in the last three or four years than we have in the periods before that. I think part of that is growth. A lot of it is investing in the technology to seize this market opportunity. I think the fact that license revenues and investment of our R&D have grown at a similar rate, I think is no coincidence. I think when we look forward to a world that I am painting where license revenue grows at about 10%, I think what we will see is our investment in R&D, the growth rate will also moderate as well. The underlying operating leverage in the model continues to play out.
In summary terms, looking back over the last few years, as I said, margins early 30s-late 40s, our earnings over the same period have grown at 29% compound, the cash flow has grown from there. Normalized cash generation have grown at the same rate as earnings. Like many tech companies until 2003, 2004, we were a cash collector. The more cash we had, the better we were. We started to think about balance sheet management, capital structure management. We introduced a dividend back in 2004 for 2003. We introduced a buyback program, we deliberately managed the cash of the business down to about GBP 50 million by the end of 2007. When the world started looking over the precipice in 2008, 2009, we very deliberately allowed the cash to build.
Apart from anything else, we wanted the business, even in a really difficult cycle, to look very robust in the eyes of our customers and our investors. We haven't done the buybacks until recently, since the fourth quarter of 2008. We've been focused on the dividend, which has grown through the cycles, including through the downturn, when it grew 10% per annum, has grown at mid-20s and sometimes higher in recent years. The result of all that, we confirmed again in February that our intention was to keep the share count flat over time. As it happens, because of that buyback activity in the 2005, 2008 period, share count is flat now relative to when we started that buyback in the middle of 2005. We said in February that we intended to continue to maintain that flat share count.
Eagle-eyed observers will have seen us in the market in the last week or so buying back some Arm shares. This is not going to be a major buyback program. It's going to be a limited buyback program with a view to maintaining a flat share count over time. Clearly, given that it is a fairly limited buyback, we don't need to be doing that particularly fast, we can pick our moments. In summary then, before we move to questions, what does this mean for the future P&L? If you look back over the last five years, what has happened? License revenues have grown at 19%, royalties at 17%, overall revenue at 15%. OpEx, as we invest to seize the opportunity, has grown at 14%, the margins are, as you see, up close to 50% now. Earnings and cash we've talked about, and dividends.
Where does this go? We see a world of 10% license revenue growth. We see a world of royalties growing at 15 percentage points above the industry. We see a world where operating leverage continues to be part of the model. We don't manage the business, you've heard me say this many times, for short-term margin. We want to seize the growth opportunities we have. We're in business to grow profits, to grow cash, the margin in a sense will be what it will be. In this model, you've seen what's happened to it in the past, we see no reason why operating margins don't continue to expand, notwithstanding investment opportunity over the next few years.
I think a combination of the revenue growth, the margin expansion, and the reducing tax rate that we've talked about as a result of initially R&D tax credits, more recently, the Patent Box tax regime. These are all quite strong tailwinds for Arm's earnings growth. I think that is going to lead us to growing profits, growing cash. I think you can expect Arm to increase the payout ratio of the dividend over time. It's been about 30% historically. I think there is room for this business both to increase that without in any way inhibiting our ability to access the growth opportunity. I think you'll see that in combination with maintaining the flat share count is how we manage our cash.
We don't plan to build a huge cash pile, and therefore, if a combination of those two factors is not sufficient, we will do some other form of return in due course, as and when appropriate. With that, I'll hand over to Simon to chair some Q&A. Thank you.
Talking of chairs, we're just going to reconfigure the stage here. I'm going to ask my colleagues to come up on stage. Tim, don't go too far. If I could ask when we go around the room, if people could just ask one question to start with, and then we'll get around everyone, hopefully, and then be able to come back for more. Here we go. We'll start down here, Francois. If I can ask you just to state your name and where you're from. That'd be great.
Yes. I'm Francois from Morgan Stanley. There was a slide showing that in 2018, all smartphones would be V8, so 64-bit. Very interesting to have this target now. Do you expect the royalty rate to be like a linear progression overall from the 1.5% today to actually what could it be? That's my question as well. What's going to be the royalty rate for everything going to V8 or 64-bit in 2018? That's my question. Thank you.
Well, I guess the total royalty per chip is going to be based on the number of cores, whether or not there is Mali attached, whether or not our physical IP is in there. There's going to be some spread of that. In terms of the linearity of that, it's a bit hard to say. I would think we'll start to see slowly through next year growth of more devices becoming V8. You're probably right, probably quite a linear progression towards all phones being V8. It's going to trickle down from the top. Pete showed in his slides how technology moves down. I think as
More and more software gets written for the top of the market, it is going to drive the acceleration into the bottom of the market as well.
What about the other Cortex-A users, like set-top boxes or TVs? Because there is probably less need to have V8 for power consumption. Is that 2018 as well, or is it going to take longer? What's your best guess? I know no one knows, but it's really a guess.
I think there's going to be a very long-term market for 32-bit Arm processors. In the same way that eight-bit micros still ship today and will probably still be shipping when we've all gone to our graves. I think 32-bit micros are going to be shipping for a very long time. Not all markets need the power of 64-bit processing. Not all markets need more memory than four gigabytes. In some cases, it'll live on for a long time.
I'd agree with that, but I'd also say things like home, they are going 64-bit. Things like DTV, set-top box, over-the-top boxes are going 64-bit, and they're going 64-bit fast. They're going 64-bit because they need access to more memory and because they need access to more performance. We can already see partners supplying 64-bit specifically into those markets.
Thank you.
Yeah, Krish.
Thanks. Anshul Gupta from Credit Suisse. One of the key drivers for your royalty rate increase is going to be adoption for big.LITTLE technology.
Right.
When I look at the chipset vendors for smartphones, two of the biggest vendors aren't using big.LITTLE right now. What gives you the confidence that those guys are going to use big.LITTLE at some point, whether it's a question of those guys hitting some sort of a barrier in terms of performance at some point, or whether it's just a question of you improving further on your existing big.LITTLE technology?
Well, big.LITTLE is an example of multi-core. One of the things that helps drive the total royalty per chip is the number of cores that are in it. Big.LITTLE is one way in which more cores can be used in chips, but it's not the only way. I think we'll continue to see more cores being deployed, or many cores rather, being deployed in these devices. Now, the overall uptake of big.LITTLE has been quite strong. More people are using it now than there were 12 months ago when we were sat here. I think we'll see continued upgrades of that or increased use of big.LITTLE driven by the needs to lower power, and the adoption of the technology will be driven by our execution on the roadmap of CPUs and also on the software that helps really maximize the benefit of big.LITTLE. Darren, anything on that?
No. That's exactly right. The big challenge has been getting the software there. The software has been deployed. Quite a few partners are deploying it. Even I think some of the big smartphone guys you talked about are deploying it now.
Thank you.
Okay.
Morning, it's Nick James from Numis. Just thinking about the embedded processor opportunity and the Internet of Things and comparing that to the processor opportunity in mobile, where you had a lot of performance demands that kept accelerating in the mobile market, and that meant that there was opportunities to sell new cores and then keep developing. In this Internet of Things, is there going to be a similar path to more innovation in the terms of the processor core that you put in a chip for an Internet of Things type of application?
Yeah. There are many applications that need more processing. Pete and I were in a customer the other day where we were talking about technologies that are required for some of these deeply embedded markets, things like floating point, doing some mathematical functions in a very effective way, a very efficient way, doing cryptography in a very efficient way. This isn't just very basic integer number crunching. Some of the algorithms that these deeply embedded devices run are quite complex, and so I can see a roadmap of additional technologies that we'll develop over the years and deploy.
Great. Thank you.
Just there.
Thanks. It's Vijay Anand from Espírito Santo. A question on the licensing backlog. Tim, as you said, most of the growth has been driven by the subscription and architecture licenses. Given that you expect backlog to be flat over the medium term, does it mean that Arm has reached a level of peak penetration with respect to subscription architecture licenses? Would you say the new type of Arm licensees, the software guys and the OEMs and operators, why wouldn't they go for the subscription or architecture licenses? Thanks.
Well, they might do. I think you're going to continue to see probably more subscriptions, and most of the subscriptions will also renew. Out in time, you're probably going to see more architecture licenses as well. I don't think it's capped out. It's a part of the model. It works well for some of our licensees. It is not the most common license model, as Pete showed. Most of our semiconductor companies will still continue to take implementation licenses from us. I think it's going to stay an integral part of the action.
Thanks.
Just there.
Hi, it's David Mulholland from UBS. Just one question. You put up the slide showing the different areas of IP that you have in different devices, and one of the things that's driven your business has been moving into things like Mali, into graphics, and moving into core network processing as well. Over the next 5 to 10 years, should we be thinking of there being further IP blocks coming through from Arm? Obviously, you've got an investment in connectivity, is that something we should look at as being a driver that maybe isn't so obvious today?
Well, yes. I think I covered in the application processor space, we are being pushed to add more and more IP, just so we can make sure it works together and people can get it out faster. In things like enterprise networking, we have added specific IP that customers have asked for. As we engage more and more, I'm sure there are going to be more demands. That tends to be where a lot of our product offerings come from. It's working with a partnership, getting their input, and then deciding we're going to build IP. Enterprise networking, I would have thought that there'll be more that we deploy in that space.
Hi. Janardan Menon from Liberum Capital, Liberum. I'm just going to your slide on potential market shares in 2018 in infrastructure division. In servers, you're saying it's going to be 10%-15% market share. In all the other areas, including mobile infrastructure, 60%, 30%, et cetera. It's relatively very high market shares that you expect to get over the next 5 years in the infrastructure segment, and a little bit more cautious on the server side. I was just trying to understand why that is, because is it that the ecosystem in infrastructure is easier to penetrate, or is it that the competition is weaker in those segments that gives you the confidence? From my understanding, infrastructure is something where the timing is quite difficult to predict because these guys move.
There are 2 or 3 very large animals which move quite slowly and at their own pace. With things like the LTE development for many of the base station manufacturers already completed. What gives you the confidence that you can achieve the kind of targets that you've set for yourself by 2018?
Is this your question? Sorry, it was an extensive question.
Yeah. One is.
I'm not sure.
I'll divide it into two parts. One is. Why do you think infrastructure is going to be so much easier than server market, in terms of getting to very high levels of market share? Because the starting point is similar. Secondly, the timing could be unpredictable. Would you agree that timing could be unpredictable at the rate at which an Alcatel or an Ericsson or a Juniper, whoever it is, moves over compared to consumer markets or even the server market?
Yeah. If you look at infrastructure, there are a diverse range of players that were already existing in that market using alternative architectures, and they have made a commitment to move over to the Arm architecture. They have an established footprint already in the networking space. They were deploying there using alternative architectures. They've chosen to go with Arm, and so the growth that we see from an Arm perspective is kind of an expansion of our footprint in that space. A lot of the players already had an extensive footprint in that space. I think that's why we can feel confident that the innovation that they're bringing in other areas like the heterogeneous computing and all the other things that I talked about, is going to serve them well as they continue in that market.
Hence, the confidence that the market shares that we're talking about are going to grow. That being said, as you highlighted, the lead times in terms of people adopting and deploying on the infrastructure space is a little longer than what you're used to on the consumer side, but time in market is longer, 5-7 years, as an example. Now in the server space, obviously, there's been one very large incumbent, it takes time.
Adi Metuku from Bank of America Merrill Lynch. Had a couple of questions. I'll ask the first one first. Just looking at your slide on servers. You're saying the enterprise server opportunity and the high-performance compute opportunity is going to be in 2018. If you take what your customers are saying, AMCC is saying that their offering that's going to ship in the second half of this year is essentially going to be a competitor to Xeon, [Rosslyn] Avoton. AMD seems to be planning a similar thing in 2016 by doing their own custom core. It looks like your targets are a bit conservative. Maybe you could.
Very happy for our customers to have bullish targets, certainly when we look at the types of chip they're building, they are delivering very high performance. They've got to get those built, they've got to get them deployed, they've got to get them in boxes, they've got to get them in server rooms. These things do take time. Certainly they are putting a lot of emphasis behind that, have very aggressive goals. All power to them.
Just a quick follow-on on that. Is there a chance that you can actually end up having more than your 10%-15% share in 2017 rather than 2018, as you're saying at the moment?
Absolutely, there's a chance that that could come through. We think that is a sensible prediction for what might happen. Of course, there's opportunity for more, there's opportunity for less. We are making a forecast out in time, forecasting is unfortunately not an exact science.
Hi, Matt Ramsay from Canaccord Genuity. A couple of things on the enterprise market, though, Lakshmi, you spoke a bit about the increase in core counts that's coming over time. In a heterogeneous design in the mobile space, where there's additions of two types of cores, GPU, et cetera, it's pretty clear how the royalty rate can be additive in that case, you guys have laid that out for a while. In markets where there's big multi-core designs that are homogeneous from a core perspective, how does the royalty rate trend in that case, as you go from, say, a 16 core to a 48 or 64 core? How do you monetize the IP that you have around coherent interconnect? That's a big part of the value I would expect in that type of scenario.
Yeah. Fundamentally, our royalty model is based on a percentage of the selling price of the chip. A chip with 48 Cortex-A57s in it is going to be a fairly sizable piece of silicon that's going to sell for quite a high rate. You get that help in terms of our royalty on that chip. Plus, our royalties are based on how much Arm IP is in there, whether it's a heterogeneous collection of cores or a homogeneous collection of cores. Fundamentally, the royalty increases with the number of processors in there. You get a multiplying factor between those two. Technology such as CCN is licensed. It doesn't command its own royalty. That's the technology right now. That may change over time. That's how the model for that sort of technology has grown. Did you want to add?
Just a quick follow-up to that. I'd be interested if you've seen anything in the server end market, effects in the market with your partners after IBM open sourced POWER8 and had the OpenPOWER program and what they're doing there, if that's had any perturbations in your interactions with licensees and customers.
A couple points. They didn't open source Power. They announced that they're making it available to other partners. They have been licensing Power for at least 20-odd years, and still you've seen end user demand for Arm. IBM themselves license the Arm architecture. In terms of the OpenPOWER alliance, it's a limited number of partners, and what they have announced has been companies that are doing companionships to the POWER8 SoC. I think choice is good. Fundamentally, choice is good, and I think OpenPOWER is bringing another choice to the market. I think they will address a higher performance point than what the Arm partnership is targeting. Overall, I think it's a good thing for the industry to have more choice.
Okay.
Just looking at the slide 77 where you gave the ASPs, Tim. I was just doing the math, and it looks like your royalties could have been about roughly GBP 700 million if you had all Cortex-A processors with v8 in them, and that looks like a GBP 200 million difference in the royalties relative to what you had in 2013. Is that the right way to look at it, essentially you're saying if there is no growth and everything was v8 and your microcontroller share doubles, you would have had another GBP 200 million in royalties. Is that the right way to look at it?
Ian did the math. I'm not sure it's quite GBP 200 million. No. If 495 was $1 million of our royalties for last year, 6% times GBP 10 billion is GBP 200 million. That's an extra GBP 105 million.
Okay. Thank you.
100, not 200 in your example. Yeah. That was the way the math worked out. A hypothetical case, of course, just to illustrate the average royalty per chip issues.
Any other questions?
Hey, guys. Youssef Essaegh from Barclays. Tim, you said that you expected that the number of architecture licenses is going to grow going forward. Can you tell us what you think, which end market is more likely to drive that?
I'm not sure I said they would grow. I would say that when new architectures are introduced, and Arm introduces new architectures from time to time in quite a long cadence, then I think it's perfectly possible that there will be new architecture licensees, and it's perfectly possible that existing architecture licensees may take a license to a new architecture. That's really the point I was making. I think in version eight of the architecture, there were more architecture licensees than there had been for previous architectures, really because of the timing of the introduction of that architecture and the opportunity that it was offering to semiconductor companies to access some of the markets, particularly that Lakshmi's been talking about today. I don't necessarily say there'll be more, but I think it will continue to be a feature.
I think it's going to be highly linked to our success in some of these new markets. The way we look at developing our own CPU implementations is to create products which have applicability in many different end markets, where things that we know we can license to many different customers and get good return on the investment that we make on that engineering. We're covering the majority of applications with our product roadmap. The architecture license allows people to go into spaces which are economically a good idea if you're a semiconductor company, but as an IP company, not such a good idea. As Tim said, there was early adoption of version eight of the architecture because of its 64-bit capabilities as the opportunity in servers and networking is becoming a reality.
As that develops, it may well open up opportunities for other people to build chips which are hitting more specialized implementations, more specialized applications, which make sense if you're a chip company, but don't make sense if you're an IP company. Us building chips or processors that we only license for one customer doesn't allow us to get any scaling. It really is a function of how these markets develop and how the adoption of Arm technology goes in these markets.
In the next, say, three to five years, do you see potentially the opportunity for more cases like the one you just described?
I think there is quite good coverage right now of both these markets in terms of our standard products and customers building chips on the standard products, and customers targeting that with architecture licenses as well. We may see a small number of customers who decide to go that way as well, but I think it will be relatively small numbers. There's a question behind, actually.
Previously, you said that you would expect Intel to take about 5%-10% market share in the smartphone and tablet market because there's always room for a competitor. Intel and tablets this year seems to be likely to be heading above 10% market share, probably somewhere in the 10%-15% range. Some smartphone vendors like Asus and all have moved all their products across to Intel. I'm just wondering, do you and your partners believe that Intel's market share could be kept at these levels in tablets? What are the strategies to ensure that there isn't any further encroachment into the Arm territory?
Well, in terms of our strategy for maintaining high market share, it's about executing on our roadmap. It's about staying very close to our partners and understanding future trends of the market and making sure that the work that we do, coupled with the work that our partners do, continues to deliver leading-edge processors, highest performance, minimum power. I think the combination of our innovation and the innovation of everybody in our ecosystem, I think, is going to mean that we're going to have a very high market share going forwards for the Arm architecture. That involves a lot of hard work. It involves a lot of close collaboration with a lot of people in our ecosystem. When I look at what we're doing, what our partners are doing, I remain confident that we're going to maintain that very high market share.
Hi there. Alexander Duval from Goldman Sachs. I just wondered if you could talk a bit about how you see pricing dynamics of mobile device chips a bit more near term in 2014. TSMC, of course, has said it expects smartphone content to increase in 2014, mostly due to high-end smartphones. You've obviously given a long-term view today on ASPs being stable, but I wondered if you could talk a bit more about 2014. Many thanks.
Well, I guess so. We put up the ASPs on the chart. We expect that to remain the same for 2014. We don't see any major changes this year.
One quarter to the next, chip prices are going to move around. Fundamentally, as featured on your slide, new technology comes in at the high end. It trickles down. There is very aggressive price competition going on because the markets are large. What we have seen over the last year or so is actually ASPs staying fairly constant at these levels. From one quarter to the next, you might get a bit of fluctuation. Long term, we think that is a safe model. There is a question at the front.
Hi. It is Andrew Humphrey at Morgan Stanley. I just wanted to ask a bit more about networking. Clearly, your play there is to be an increasing share of that market over time in a market that is growing to low single digits percentage-wise. Clearly, there is a value proposition for your customers there, how are you working longer term, I guess, to mitigate the risk that you end up importing deflation into that industry?
If you look at that industry, as I talked about, that industry is actually used to a very diverse ecosystem in terms of supply base because of the kinds of specialized processing that I talked about. Semiconductor companies that are in that market today actually have a lot of value invested in some of the non-CPU core pieces like packet acceleration and all of those different things, crypto, a whole host of specialized processing applications. What we see happening is there is tremendous value for that. With the addition of Arm, us enabling them to outsource some of their core IP investments, they are actually able to leverage that into continuing to invest in other areas and innovate in other areas. Actually, I think that I do not see Arm coming into that market as changing that dynamic. Only for the better, actually.
There is a question at the back.
Thank you. Fati Naraghi from Newton. Just a question. We've seen a lot of litigation in the whole tech sector in the last few years, much more so than before, I think. Looking, you've had many more architectural licenses for this time than historically. As the industry moves towards 64-bit, what risk do you see of your licensees sort of litigating against one another? Do you have provisions in the terms of the license, et cetera, to restrict any sort of litigations amongst them?
Well, we don't enforce that our customers can't litigate against each other. I think that would be quite restrictive if we did. We respect other people's IP as much as we value our own. Obviously, I think our customers have got a right to go and defend themselves if they feel their IP is being inappropriately used by someone else. I think with the rise of architecture licensing, I don't necessarily expect that's going to have a big change on the amount of litigation that goes on in the industry amongst semiconductor companies. The semiconductor space is quite advanced and mature in the way that the cross-licensing goes on amongst the big companies on some of their fundamental IP. I think really my experience over the last few years is we haven't seen an increase in semiconductor companies litigating against each other.
Patent trolls has been a bigger growth, even that seems to be subsiding a little bit at the moment as well. I don't think architecture licensing changes that landscape very much. Yes. Phil, you had a question?
Phil Pearson from GLG. Just a clarification really, Tim. On the slide on the long-term growth opportunity, you describe license revenue growth reverting to 10% CAGR over time from a period 2014 to 2020. You refer to the backlog being guided to be roughly flat over the medium term. Do you need over that 2014 to 2020 period, is there a point at which you need the backlog to grow again, to continue to deliver that roughly 10% CAGR on licensing? If so, can you give us some sort of sense of at what point we would need to see some signs on that?
I think it will go up and down. I think what's going to happen in the next few years is the gap that's been created between the backlog growing that much faster than license revenue. That's going to come into the P&L and drive license revenue. I think, yeah, in reality, although we talk about 10% growth out here, there's going to be a period of transition from the 30% CAGR we've seen in the last four years to that 10%. Part of that is the unwinding of the backlog. I think, looking out, in a sense, to the 2018 to 2020 period, if license revenue is growing at 10%, then out in time, once the backlog was normalized, the backlog would also grow at a similar rate to match.
That was my question.
Okay. This one.
Thank you. Jasmeet from Bernstein. I'm looking at slide 54, which is the concluding slide of the royalty opportunity in mobile. It suggests that between 2013 and 2018, a royalty growth opportunity between 15%-25% CAGRs. Then in one of the previous slides, you suggested 13% CAGR for unit growth of smartphones and tablets. Let's assume the revenue growth opportunity there is about 10%. It's suggesting between 5%-15% growth coming from royalty rate increases, which can be anything between royalty rate becoming 1.3 times-two times. My question is, what is the source of this uncertainty? Is that penetration of big.LITTLE, which is uncertain? Is that penetration of Mali? You sounded more comfortable and confident about v8. What's, let's say, why the range?
The short answer to your question is yes. As I said a little while earlier, the exact royalty rate per chip is all about the mix of technologies that are in there. How many cores, whether they are heterogeneous or homogeneous, does not really make any difference on that in the big picture. What the percentage of Mali attach is, how much of our physical IP is used, and that is why when we did that 15%-25% CAGR calculation, it was based on. We talked through the mathematics of that last year. It was based on growth rates in the different tiers, the kind of ASPs that we expect, how those ASPs might move around over time.
We modeled scenarios for different amounts of technology adoption across those different tiers, and that is why we ended up with that range, which may seem quite broad, but as we sit here now looking 5 years out in time, it is a bit hard to call that. There will be different factors affecting it. But again, that to us seems like a fairly sensible range to be looking at for the smartphone sector. Question.
Yes. David from UBS again. Just a couple of short-term questions. Tim, I guess you have now probably seen a fair few of the checks for royalties for Q2. I just wonder if you could give us a comment on your feeling around how things have trended Q2 versus Q1. Then just secondly, on seeing the acceleration in the second half to get to similar growth rates and royalties that you have seen in recent years as you had guided to before.
I guess we could not get through a whole analyst day without some questions about the end of next week or the week after. No, obviously, we are. If we had anything material to say about short-term trading or indeed full-year trading, we would be saying it. So we remain comfortable with consensus out there. As it happens, we have received a reasonable proportion of the royalties, but not the majority. We are not standing here making any update to current trading. So you can take from that we have got nothing new to say on that. In terms of the full-year, we remain comfortable with the full-year revenue consensus, which is in the sort of $ 12.95, just above range.
Thanks. It's Matt Ramsay from Canaccord again. Tim, in your slides, you called out what I think is interesting, another $20 billion TAM market, which is automotive, where the share is 7% for Arm today. Just to whomever on the panel, I'd just like to hear your thoughts about trends in that market, ASPs, royalty rates, et cetera.
I think automotive is really quite interesting as a growth market for adoption of silicon. We're seeing much more sophisticated entertainment systems into cars. That drives adoption of processors that look like application processors. With all the increase in driver-assist technology, you're going to see cameras in cars all over, trying to keep track of what is going on around you and trying to point out if some other driver is about to do something that you don't want. Just generally, there's going to be a lot more intelligence in the car trying to keep you safe. I think that's going to drive adoption of a lot more silicon in there, which is why that market is set to grow. Design-in cycles are long. Once you're designed in, you're in forever. It is a market that doesn't move very quickly.
Although I can see where more and more technology can be used, it will take a frustratingly long length of time because of all the testing and safety requirements that have to go on in automotive.
Youssef Nassar from Barclays again. You've shown us on slide 33 your roadmap on the hardware side, and then 37 on the achievements you've done on the software side for servers. I was wondering if you look at these two timelines together, what do you think for in the order that you told us that you want to be able to penetrate the different type of server markets, what do you think is still left to be done? Basically, what comes up at the right of the arrow because it stops in 2014.
Oh, on this timeline diagram?
Yeah, absolutely.
Okay. I talked about the fact that we have a base Linux platform. We have Ubuntu that's demonstrated, Red Hat that's demonstrated. We start to see more and more software partners, but also starting to see more applications being ported over to Arm. That's going to take time. In terms of trying to think, the middleware pieces coming together. The workloads that we're going after that I showed, I think we have a base software platform ready to go to enable that. I think someone asked, "Hey, you're saying HPC telco out in time." As you start to expand the number of workloads that are addressable by the Arm market, more and more of those applications will get ported. To answer your question, we have a base software platform ready to go today for the applications and workloads that we're targeting today.
As we start to see more and more of the market getting interested in Arm, those software applications have to get ported. Does that answer your question?
Yeah, it does. What I was trying to think about is, clearly there is different type of servers. There's also different type of companies that have tons of different things to run on these servers. What I was trying to assess is basically, say for instance for anything storage, you may have now 100% of the top 20 apps that anyone would need. If you want to move to big data, maybe Hadoop becomes necessary. You have that as well. What's still left to be brought?
One thing to perhaps make maybe a little explicit is that there are different abstraction layers underneath, right? A lot of the applications that are being targeted by Arm sit on top of middleware like languages like PHP, Python, and all of those different kinds of languages. Those languages today run on Arm already. A lot of the applications could run directly on whatever infrastructure there is today. It's a question of perhaps tuning those a little more and how it rolls out from there. From a base software platform perspective, we have the OSs, we have the middleware, we have the boot infrastructure, we have all of the pieces that you need to make that happen. It's just extending that. I can't say that I think in mobile there's X many apps, and in infrastructure, the answer is it depends.
I think we really focused on the infrastructure that allows then people to build their applications on top of that. That's really what we will see beyond the work that we've done here. People starting to roll out these servers, build their applications on top, and over time, that will go from being very broad-based, which is where we've started, to filling in more of the kind of niches in filling out the entire space.
Okay.
We better make this the last question, and then we'll go to sandwiches next door where we can all chat some more.
Thank you. Hopefully it won't be too long then. It's Andrew Dunn, RBC. Just on networking. You've had 5% market share last year, and you're targeting 25%-35% by 2018. Call it 30%. How should we think about the trajectory of getting from 5 to 30? Are there any particular catalysts or technological hurdles or whatnot we should be thinking about for the next few years to get you to that 30%?
I think generally when Arm technology starts being adopted in new markets, it goes quite slowly and then starts to take off. I think with all the design wins that we have, we are poised for that to take off. This is a market that's not driven by consumer trends. Nobody buys a new router because it's green. It's on quite a slower cycle compared to a lot of the markets that we're typically in. I think probably the uptake towards that 25%-30% range is more in years four and five, and I think you'll see it ramp up at that point.
Okay. Your existing licensees and networking are continuing to develop on MIPS and power, at least in the near term. Is part of your share gain substitution or is it new markets to them that they're entering?
Some cases, yes. In some cases, what you said. It's a mixture of things. There's a lot of prototyping and activity going on right now.
It's about equipment upgrades. It's about as network architecture changes, it creates discontinuities where people are looking to service that with the Arm architecture instead of the incumbent architectures, just because, as we spelled out, that's where the investment's going. It is a mixture of all.
Thanks.
Okay. Well, thank you all for coming. Hopefully, we've given you an overview of where we see growth, and there's an opportunity to chat some more as we move next door and have a sandwich and a cup of tea.