Good morning, everybody, and thank you all very much for coming along this morning. Hopefully, we have an interesting morning for you with several presentations, and we'll follow that up with Q&A. I'm just trying to make the slide advance at the moment. Oh, spacebar. Okay. There we go. We are going to start off with Tudor this morning, who will give an overview and an introduction. If you want to sort of take away from today, and you don't take away the detail, then Tudor's presentation is what you need to take away. Graham will then turn to the business and talk about how we make our money. Then we'll have a break. In the session after the break, we're going to talk more about the technology with two presentations from Mike Müller, our Chief Technology Officer, who will talk about the microprocessor roadmap in particular.
Of course, as many of you have heard me say before, the microprocessor from Arm is worthless without the software that actually runs on it. James McNiven, who works closely with our software partners, will talk about what we're doing there. With no further ado, I'll hand over to Tudor to kick off.
Thank you all. I think what Mike just said was conscious is a content-free presentation, but I'll try to make it have some content. I know many of you are very, very familiar with Arm, of course. For the sake of those of you who are not quite so familiar with Arm, I'm going to give a quick overview before looking at what I see as our future for the next 10 years or so. As you know, I'm sure, Arm is the world's leading semiconductor IP company. We think of ourselves as an R&D outsourcing company for the semiconductor industry. It's a licensing and royalty-based business model. That is the basis of the IP business model. It's pretty clear that we're gaining share in the markets we're in, and we're in growing markets.
As a result of that, as a company, we're turning those revenues into both profits and cash quite effectively. Just a quick overview of the business model. There are several variants of this business model, but the fundamental version is what I'm describing here, where we license our technology to semiconductor companies. That has an upfront license fee. Then those companies build chips. That's the black thing there. The Arm core is integrated into that chip that comes from the semiconductor company. Those semiconductor companies sell those chips to OEM end customers who build the end products that we all like to buy. Every time one of those chips is shipped to a customer, the semiconductor company pays us a small royalty. Typically, those royalties are based on a percentage of the selling price of the chip. There are a multitude of semiconductor companies that we license.
In fact, we license over 250 different semiconductor companies. Each of those semiconductor companies can build more than one chip with the Arm core license. If you like, there's a sort of amplification effect going on here, where the one Arm licenses many semiconductor companies that build many, many chips. That's how Arm is able to get to market with these enormous numbers of chips that you see shipping each year. This is the world's fastest growing industry, but things happen remarkably slowly. It takes us two to three years, typically, to develop a microprocessor core. From the time of licensing that core into a semiconductor company till they get a product that they can ship in volume takes three to four years on average. Once that happens, those products have an amazingly long life. Any one processor can go on shipping for more than 20 years, we believe.
We've seen that with other microprocessor architectures. A proof point is that today, the ARM7TDMI that was first licensed in 1994 is still representing almost 50% of the shipments, over 40% of the shipments of Arm products today. That demonstrates just what an incredibly long life these things have once they get into volume production. You can see from this picture that Arm Holdings and the OEM customer are not actually a direct customer-supplier relationship. However, they are important to us, and we're important to them. We have, if you like, typically a sort of marketing relationship or a biz dev relationship between those influential end customers and Arm . That helps us explain to them why Arm technology is good, and they explain to us what they want in our future roadmap. The business has a very long lifetime.
The licensing business grows sort of linearly, typically between 60- 90 licenses a year. In fact, last year, 2010, it was 91 different processor licenses. In Q1 this year, you may remember, we announced 39 licenses. That's above average. On average, we think it's around about 90. Those are going to many different types of customers. What we see is, if you think about it, as we do that licensing, fundamentally, that enables royalties into the future. As I explained, those royalties can go on for a very, very long time. What we see is the companies buying Arm technology are, in fact, making very long-term strategic bets, if you like. The bigger Arm becomes, the more influential Arm becomes, the safer that bet. That's one of the reasons why I think the Arm model has become very strong in the last few years.
Arm Holdings is well known in the mobile sector, but we're seeing that the Arm technology is penetrating into many, many other sectors within those same semiconductor companies. On that graph on the bottom right-hand side, you can see the effect of royalty from licensees. Down the bottom there, 100 or so licensees that have licensed the Arm technology for more than 10 years, and they're still representing a portion of that shipping royalty. Yet on the top, in dark blue, the last 370 licenses licensed in the last five years or so, you can see they're contributing a negligible proportion of our royalty at the moment. Roll that graph forwards, and I think you can see how we expect it to extrapolate. That is why we're confident that the royalties are going to keep growing for Arm. Most of you are very familiar with this slide.
The key growth drivers we see firstly is increasing the Arm value per product. We're very fortunate that Arm is designed into these smart products that are using more and more Arm chips per product. In fact, those chips are becoming higher value within the product. We're also growing outside of our mobile base into non-mobile applications, as you will see in many of the presentations today. That dramatically increases the served market available to Arm. Finally, of course, it's not just about microprocessor cores. We have our graphics cores. We have our physical IP. The goal of this is to increase the Arm value chip per product that you and I buy. Excuse me, there is a seat just down the front here if you want. I said that we turn that revenue into cash and profits. Just quickly explain how this happens.
Many of you are familiar with this, of course. Taking the year 2010 numbers, we take our revenues. Most of our revenues, some 95% of our revenues, are dollar-based, of course. This is an international business where almost all revenue is dollar-based. Fundamentally, therefore, we're turning dollars into pounds. What we see is that the royalties are now more than 50% of that revenue. That has grown steadily over the last few years. It's now 53% in the last year. We expect that proportion will carry on growing slightly slowly. As a software-type company, if you like, we have a very high gross margin. We take our R&D expenses as they occur into the operating costs. Therefore, we end up with an operating margin around 40%. That has been growing slowly and steadily. We expect it to continue to grow as the proportion of royalties gets bigger.
Of that operating cost, there's a bit of a hedge against currency because approximately 50% of that operating cost is dollar-based. Not all our people are here in the U.K. We have a lot of people in the U.S. and in other countries that are fundamentally dollar-based. Therefore, there is some hedge to the pound/dollar issue. As I said, the operating margin and therefore profit should continue to grow as a greater percentage of the royalties flow through because, as you can imagine, the royalties are almost entirely profit. Therefore, we ended last year at GBP 290 million of cash in the bank. There is free cash flow in this company. That cash goes up. That's a quick overview of Arm. I'm sorry for those of you who are so familiar with it. I hope you're still awake.
I now want to talk about how I see us going forward for the next nine or 10 years. The first area is, of course, we see billions of interconnected screens. We've said for some time that mobile technology is getting reused in other markets. That's what we see here, that with the choice of different suppliers, OEMs are actually innovating with new types of products. Arm is now being used in applications processing, connectivity, and storage. This is all enabled by standard software. That standard software, such as Android and Linux and other operating systems, is enabling multiple form factors of screens. The same fundamental Arm technology is enabling a huge range of screen sizes and devices, from enormous HD televisions through to tiny things you hold in your pocket. The next market we look at is billions of real-time devices.
This is, if you like, moving on from connected screens more towards the sort of Internet of Things because the consumer is becoming increasingly connected. As you and I know, we expect to be connected. We expect to have data not just on our portable devices, but also on things like our TVs and other fixed-point devices. Mobile baseband, Wi-Fi, Bluetooth, GPS, all of these fundamental technologies that, again, are fundamental to mobile technology are being reused in many different markets. We see a growing demand for local storage because you can't always be connected. Again, Arm technology is used there. Arm processors are ideal for very efficient real-time communication and control. Many of the devices that we see that we think of as mobile devices and many of the fixed-point devices are starting to use this basic technology.
Therefore, that market for real-time devices by 2015, we expect to be in the region of 11 billion units. I come to microcontrollers. The microcontrollers are almost certainly the largest growth market for Arm in terms of volume because this really is about the Internet of Things, where everything gets smarter and everything gets connected and everything becomes electronic. We're talking about things that you really don't think of as electronics. M otor control, smart meters, security, airbags, toys, air conditioning, heating. These are not things you think of as electronics, but they all have microcontrollers in them. Arm Holdings is starting to penetrate those markets with our microcontroller products, which is why you've seen rapid growth in microcontrollers recently. Also, as you're well aware, there's a big green agenda going on around the world.
Governments are starting to mandate efficiency models for things like air conditioning, things like cars, of course, and even computer screens. All of these things need power-efficient controllers, which fundamentally are microcontrollers. Arm Holdings is very well placed for a lot of those markets. We see a great amount of innovation here. This is driving 32-bit technology, which suits Arm ideally, into markets that were previously driven by the 8-bit market. That market is the biggest single market that we can address, 19 billion units by 2015. Our share of that is growing. I expect we'll continue to grow. As this market evolves, we see huge opportunity. We've talked for a long time about how we expect to see growth outside of mobile. What we see now is that the non-mobile devices or markets account for about 75% of our licenses.
When we're licensing technology into a company, we have some idea of what they want to use it for. We estimate that about 75% of that application is actually today in the non-mobile space. Because of the lag effect of licensing to royalties, as you might expect, 60% of our royalties are still coming out of the mobile space. Given that I explained that the licenses are seeds, if you like, for future royalties, I think we can see how that is going to trend. The other trend here is that, as I mentioned, consumer electronics is using more and more Arm chips and higher-value devices. Many of the technologies that we're in today are using more and more Arm technology. We've talked about smartphones having five or so Arm processors. That same sort of model applies to things like TVs and many other connected devices.
As I mentioned, there are opportunities in other things outside the Arm processor cores. Not only higher-value processors, and we've talked in the past about how the latest Cortex family is attracting higher value to Arm, but also multiple processors per chip and other IPs, such as our physical IP and graphics. This is all enabled by this increasing opportunity in the market. The other trend we've seen recently is that our business is evolving. I said earlier that companies are making strategic bets, if you like, on Arm. Arm is now becoming regarded as a key partner in an increasing number of leading technology companies. I'm not just talking about semiconductor companies. I'm talking about end-market companies. We are increasingly being seen as key to the whole ecosystem. We have, if you like, a bit of a challenge.
It's a very nice challenge that we need to deal with a great number of influential companies. Each member of the management team in Arm has responsibility for developing relationships with various key ecosystem partners and companies. The other area where I think our business is evolving is that as we continue to invest in R&D, we need to do that in a way that matches closely to our customers. Of course, our customers are all based outside of the U.K., to a first approximation. They're based all around the world. We have multiple experienced engineering teams based in key strategic places around the world. This helps us to easily access our customers and give them the support they need. The other side of that, of course, is Arm has to be an innovative company going forwards.
I'm extremely confident that we have people well placed around the world to get the input required to keep driving the innovation that will take Arm forwards for years to come. Many of you have been at Analyst Day before. Some of you have been at many Analyst Days before. I'm just going to take you back a little bit over the last few years to show you what we told you and to correlate that with what actually happened. Back in 2002, some of you were here then, Bruce Bekloff, who was the IR guy at the time, stood up and said that we expected by 2006 Arm would be shipping 2 billion units. Many of you laughed. You ridiculed him. You called him 2 billion Bekloff. He was wrong. You were wrong. Actually, in 2006, 2.6 billion Arm chips were shipped. We undercored it.
In 2004, we predicted that by 2008 there would be a lot of growth in non-mobile as a proportion of the total shipments. We said actually that non-mobile would be 1.6 billion and mobile would be 1.7 billion. Almost parity, if you like. We were wrong because the market grew faster than we expected. The non-mobile, we actually slightly overcalled it, but we got pretty close on the 1.5 billion units. Mobile grew much quicker than we expected. That was largely as a result of this trend we're seeing to more products, more Arm products per mobile chip, mobile phone, sorry. Actually, mobile in that case was 2.5 billion. The ratio of 1/3 non-mobile to 2/3 in mobile carried on at that point. We were slightly wrong in our prediction there, but I think in a nice way.
In 2005, again, we were predicting that non-mobile would outstrip mobile by 2010. The same thing happened. Because mobile grew faster than we expected, we actually called the non-mobile dead right at 2.3 billion. The mobile, as you can see, dramatically outstripped our earlier predictions. That's been a great success for us. It's been a great bonus for us, if you like. We did foresee this trend, but we underestimated the incredible trend for how much Arm Technology is going into these high-end products. You could say our analysis is continuously conservative. In 2007, we didn't give you a projection. What we did was showed you how we expected those markets, the end markets, to grow and how we expected to grow into those markets. What I can tell you is that those were predictions for 2012, i.e. next year.
The numbers of the market size we predicted for 2012 were actually achieved last year. The total market is actually growing faster than we expected. I think all of these things make us feel quite comfortable that going forward, there's a great future for us. Let me talk a little bit about our 2015 opportunity here. Again, I'm not going to give you accurate predictions on what we expect Arm to be in, but I'm going to show you how we expect the market size to grow. I think we can make our own estimates of where we expect Arm to be in that.
If you take the high-value markets and the high-growth markets, you come up with the sort of green arrows there that the key markets for Arm, every market is important for Arm, but the key markets are smartphones, which are going to continue to have a very high number of Arm cores in them in a growing market with very high market share. Mobile computing, which today we have a negligible share in, but is going to grow into something quite respectable. Digital TV and set-top boxes, because that, again, can have quite a lot of Arm technology inside it in a growing market. Hard disk drives and solid state disk drives, the need for storage continues to go up and is guaranteed to continue to go up in future years. Then microcontrollers.
Our percentage market share in microcontrollers is still pretty small, but that market in total is almost guaranteed to grow very fast over the next few years because of the effects I told you about. We expect to see a growing market share in that market as well. I think you can see lots of opportunity ahead of us for Arm technology. Let me talk a little bit about Arm's vision, my vision for Arm, if you like, for 2020. Already, we are known as the architecture for the digital world. This is largely true. What we mean by that is that Arm's technology spans from the very low end, less than $0.50 chips, very slow speed, less than 50 MHT, right up to potentially pretty expensive chips at very high frequency. In reality, today, there is almost no shipment, negligible shipment of those extremely expensive high-speed Arm chips.
That's okay. What we see, obviously, is the bulk of our shipments in that sort of middle ground, in the hard disk drive, baseband, mobile computing sort of area. That's where the bulk of our shipments are today. This is what I expect is going to happen over the next few years. Of course, that market is going to continue to grow. I think we can see with great confidence that we expect to continue to grow in that market as that market grows. That will continue to be a very important market for us. I think we can see this microcontroller market is going to take off. It started to take off. It's going to continue to grow. In volume terms, that will probably become the biggest market over time. At the high end, there is progress. Excuse me. There's progress in mobile computing.
We might see Arm, we will see Arm getting into a certain level of PCs, I'm sure, in this sort of time frame. Servers also are going to happen around Arm. The volumes involved there, no matter how successful Arm are, are going to be tiny compared to the volumes at the other end in microcontrollers. As you understand very well, there's a trade-off between value and volume. The fact that we expect to see some volume in the server end by 2020, I think, demonstrates that Arm really will become the architecture for the digital world. That is what we really mean by the architecture for the digital world, spanning the complete marketplace with shipping product, right from the very cheapest thing you can imagine right up to pretty much the most expensive thing you can imagine.
I hope you feel that Arm has had a consistent strategy for the last 20 years. I'd like to think it will be a consistent strategy going forwards. I won't say for the next 20 years, but I will say going forwards. Our strategy has been quite consistently and is still to develop semiconductor IP that we license to leading companies and collect royalties on every part that's shipped. Our strategy is to continue to grow overall revenue faster than our costs. We can keep doing that in the medium term. We think that grows the company in a very stable way that enables the company to grow, but carrying on with the profitability that we have. Ultimately, therefore, I do believe we become, if we're not already, the architecture for the digital world. Thank you. That's all I have to say.
Fortunately, I have to go to Heathrow just after the break. I'm not available for any questions. My colleagues will be very happy to answer those questions. Thank you very much.
Thank you, Tudor. Good morning, everybody. Tudor spends a little bit of time there talking about the Arm Holdings business model. I'm going to go into a bit more detail on the evolution of our business model. Just before we start on that, I think it's worth noting, as Tudor said, the strategy, our strategy has been pretty consistent over the last 20 years. I think that's very true of the business model as well. Fundamentally, the business model that we have, the IP licensing and royalty business model that we have, is pretty much unchanged over that period of 20 years. It's been a very powerful approach. It's one that's served us well and indeed one that has served our customers well during that time. We expect that to continue. Clearly, we also flex it and adapt it.
I'm going to talk some more about that in the context of how the industry has changed and is changing. Today, Arm Holdings sits really right at the heart of the electronics industry. What we do, of course, as Tudor said, is centered around processors. We now have more than 25 different processors available for licensing, and they are optimized for a wide range of different types of products, from smartphones, TVs, to cars, to tiny smart sensors. Really, it's all about flexibility. It's about scalability, enabling reuse of the software investment that is going into the creation of these complex products all around a standard Arm processor architecture. Beyond the processors, of course, we also are able to deliver more benefits to companies creating these complex system-on-chips with our range of graphics, 3D graphics, IP, and physical IP.
I think if we look at the industry as a whole, the trend to outsourcing of all of this kind of technology is getting stronger. More on that later. As the industry dynamics change, we're seeing new types of companies, new types of disaggregated companies becoming Arm customers. If you go back 20 years, it was mostly integrated device manufacturers out there. Clearly, a lot's changed since then, the emergence of fabless semiconductors and the foundries. Increasingly, we started seeing companies like OEMs and in some cases, software vendors taking direct licenses to some of our technologies and service companies as well, also licensing our technology to provide chip design services for other companies. A wide range of different types of companies. We are always looking to evolve the business model to ensure that we can enable all of these types of companies to be successful with Arm technology.
Different types of companies have different needs. For example, major customers who are shipping in very high volume typically need cross-divisional access across their business. If you're a small startup company, maybe only got one product, you're looking for a low upfront fee to access the technology for that one product. This pyramid illustrates some of the different business models that we have to service the wider market. The way this works is that as you go up the pyramid, the business models, the value of the license rights that you get as a customer goes up. The amount of flexibility of what you can do with the license goes up. At the same time, the level of expertise and the level of investment that you need to make as the customer also goes up.
Typically, the level of interaction with Arm would also increase as you go up this pyramid. Just looking, if we start at the bottom of the pyramid, look at the different types of license. The yellow at the bottom there, academic research licenses. These types of licenses, it's really all about seeding within universities, research establishments. Of course, lower fees to access the technology, but no commercial exploitation rights. In this way, there are hundreds of seeds out there that potentially will turn into future licensing opportunities. Design Start is a program that we have so that designers can access the views of our IP, processor IP, for example, physical IP that they need in order to start their design. In order to fully take that through to the creation, the manufacture of the chip, they would need to come and take a fully paid-up license.
Again, it's a way of seeding activity in the market. The three boxes, the three sections immediately above that, are different types of what we call an implementation license for one of Arm's developed processor cores. They're bounded. The rights that you get are bounded in different ways. For example, multi-single use, fairly self-explanatory. You get the rights to do one use with the IP that you license or a fixed number of multiple uses. A term license, you're getting the rights to do an unlimited number of designs within a window of time. A perpetual license, of course, unlimited rights to do designs for specified IP in perpetuity. At the top of the pyramid, we have some more specialist licensing models that typically are only applicable to a smaller number of companies.
The subscription license is where a company will license a whole family of current and future IP that is developed during the term of the subscription agreement. A bit more on that in a minute. Right at the top of the pyramid, and really only suitable for a very limited number of companies, is the architecture license. I'll explain that in a bit more detail in a moment. What we see is that, going back to the example on the last slide, major companies, in order to service that requirement for use of the IP across many parts of the company in different applications, will often choose a term, perpetual, or perhaps a subscription license. For a smaller company, just doing that maybe single design, they will often choose a single-use license to start with, perhaps for one of our more mature IP processor cores.
You can see on the right there, the table on the right there, most of the licenses that we sign are either term, perpetual, or on a per-use basis. In fact, the vast majority of licenses, only 3% of licenses that we sign are either subscription or architecture. You can also see that the average time to royalty, this is to volume royalties, changes for different types of model in the pyramid. As Tudor already mentioned, typically, it's an order of up to four years for the first volume chips to be available once a company has licensed our technology. It does vary depending on the type of license. For example, an architecture license, and actually, I should just perhaps explain a little bit more about the architecture license here.
An architecture license is where a company takes a license to develop their own processor implementation that's compliant with the Arm instruction set architecture. It can run the same software, exactly the same software as runs on the Arm implemented cores. Of course, that means that the customer who's taking an architecture license has to develop that processor implementation first themselves, which adds, obviously, a considerable amount of time to the development cycle. That's why the architecture license, from the time that we do the license, has the longest time to royalties. Going to the subscription license, typically, companies that take a subscription license have a lot of broad experience of implementing designs based on Arm technology, which enables them to do that faster. They're at the lower end of the average time to royalties.
The three-year period for perpetual and term, that's kind of what I would describe as the norm for most companies. At the single-use end, the reason that's slightly lower time to royalty is because those companies typically are doing simpler designs. Although, as we said, only a few companies, as a percentage of the total, go down the subscription licensing path, we have seen an increase in the uptake of this model. We've seen some leading semiconductor companies, you can see on the right there, making long-term commitments to Arm technology through the subscription license. The way this works, it's a corporate-level agreement for a fixed number of years, giving broad access to a family of current and future technology during the period of the term. Companies using this typically will be designing across multiple application areas.
The availability of the IP under the subscription license means that the company does not need to come back to us to negotiate individual terms for each time they want to use a core. It makes it very flexible to allow technology to be rolled out across multiple applications and multiple divisions in a company. From Arm's point of view, of course, this is good. It encourages more use of the technology, more design wins, which feed through into more royalty. Looking at the architecture license, Arm designs processor implementations, the top left blue box on the diagram there, and licenses them to many, many semiconductor companies. Typically, those implementations are designed to be applicable to a number of different end markets, the broadest use possible. Sometimes, a company who is focused on a particular vertical market may want to, may choose to design their own Arm-compliant processor.
As I described earlier, this is the second line of the diagram now. The semiconductor company is now designing their own processor, making a large investment of people and expertise that are needed to do that. Going down an architecture license route is quite an expensive option for a company because, as well as the license fee to Arm, to have the rights to do this, there's also the cost of building a design team and all the expertise required to implement the chip. There is the usual royalty to Arm, of course, in addition to that. Our partners will have different motivations for taking an architecture license. Usually, it is around differentiation for a particular vertical market that they are targeting.
What we see generally is that most companies who have an architecture license will also continue to take licenses to other Arm-designed processor cores for other markets that they're active in. The basic premise, if you like, of the Arm IP outsourcing model is becoming stronger. What we're seeing is that as the complexity of the design of this IP increases, whether it's a CPU processor, whether it's graphics IP, whether it's physical IP down at finer process geometries, complexity is increasing. It's becoming more and more difficult to design efficiently for low power. Low power is in Arm's DNA. It's been what the company has been focused around from day one of Arm. It's more and more difficult. That means that it is cheaper for companies to outsource the development of this technology.
Outsourcing also enables companies to make use of mature and proven technology in the market, which helps to de-risk product developments compared with doing it yourself. It enables faster time to market. Time to market is so key in this industry. By using IP developed by Arm, companies can get new products, can innovate, and get to market faster than would otherwise be possible. We estimate that the direct saving that a typical major customer shipping devices in high volume would make over developing that same technology internally is around 60%- 70%. It is a significant saving. Of course, for a smaller company taking our technology, maybe just using that in one design in a lower volume market, the percentage savings that will be achieved will be much more than those numbers there.
The second major reason why outsourcing is so attractive in this market is the ecosystem that has been developed around the Arm processor. Today, we have more than 800 companies in the Arm-connected community, which is the ecosystem of partners working around the Arm architecture. These companies provide software optimized for Arm. They provide everything from the design automation tools required to do a complex chip design, the operating systems, the middleware, the software. They also provide services to help other companies to do their design. There is a huge pool of engineering expertise, thousands and thousands of engineers built up around the Arm architecture, to which companies have access if they choose to work with our technology. Really, no other company in this industry has the scale of Arm's connected community ecosystem.
The principle here is that by working together, we can innovate faster and enable new types of products into the market more quickly. For the leading semiconductor companies, all of that cost-saving opportunity and the ecosystem, it's a pretty compelling story really across all markets. Today, 18 of the top 20 semiconductor companies use Arm technology. What we're seeing is that the penetration within those companies is increasing. At the moment, only one company actually has more than 20 Arm processor licenses. Within these larger companies, Arm processors, in many cases, are still only deployed by some divisions and for some applications. There is still the opportunity to broaden the licensing base within these companies and, of course, to continue to license them new roadmap technology as we develop this.
Tudor mentioned our Mali graphics portfolio and our physical IP products, great licensing opportunity into those same companies for those areas of technology. That's the top 20. Looking across the industry as a whole, we can see that today Arm is used in just under 30% of all chips in the market. You can see top right there, you can see how that has grown over the last three years. To keep on that growth trajectory, we need to do two things. We need to continue licensing and working closely with those top 20 semiconductor companies. We also need to be licensing and working with the potential successful companies of tomorrow, some of which may well be smaller companies today.
For those smaller companies, without the capability to outsource high-value IP-like processors, it really wouldn't be possible for them to make the economics work of building a complex system on chip at any kind of manageable cost. The Arm business model scales very, very well to enable those kind of companies and help enable the creation of new products and applications. Around 80% of our licenses are signed with companies outside of the top 20. Around a quarter of our total licenses are signed with new customers. Quite a lot of that has been driven by companies taking licenses for our Cortex-M class of processors for applications like microcontrollers.
Often, you can see from the new customers chart on the bottom right there that a lot of those companies are actually established companies today, but who've never really had a reason before to license technology from Arm Holdings, but are doing so now for the first time. Just under a quarter of those are start-up companies as well. Ultimately, all of this licensing enables royalties to flow through later. What matters to enable that is the licenses and design wins that are going to lead to high volume in the application areas that we are covering here. We talked earlier about it taking two to four years for royalties to come through following the first license to a company.
What we've done recently is actually introduced this new format of slide, which we're going to keep up to date in the future, to try and give an indication of how we are doing on the penetration and sort of design wins in key semiconductor companies' products. The way this works is you can see there are a number of colored dots in the middle there. For each of the markets, for each of the application areas we're showing, the companies represented by the colored dots represent about 80% market share. You know if all of those companies are using Arm Holdings for most or all of their chips, then we're obviously in a very good position. A blue dot represents a company in that 80% who's mainly shipping Arm-based chips. A green dot is a company that has started shipping some Arm-based chips, but it's still ramping up.
Maybe they're also shipping some other chips based on other technology in parallel. A yellow dot represents a company that has publicly stated that they're going to use Arm Holdings technology, but the technology, their chips are still in development. They haven't started shipping yet, but they're expecting to do so. If you like, it's a design win, but it hasn't yet come through to generate any royalty for us. A red dot is either where Arm Holdings doesn't have a design win with that company, or in some cases, it might be that Arm Holdings does have a design win, but it hasn't been announced publicly yet. If I can just take one example to illustrate this. If you look in the middle of the chart here, the digital TV and set-top box line, the two blue dots there would be Samsung and MediaTek in this context.
You'll recall that last year, ST announced that they were going to start using Arm in set-top boxes. They are an example of a company whose dot on this chart would have changed from being a red to, in fact, one of the greens now for this application area. They have actually started shipping some Arm-based products in this market. The red dot in this example is actually Broadcom. Broadcom, of course, recently took a subscription license with Arm. We're hopeful that they will decide to use our technology in this space. If they do, and if they do and that's public, then that dot would change from a red to yellow and beyond. That's how it works. On the right there, you can see some references to the last couple of quarters and what we've announced.
We'll also be looking to keep up to update you with the companies that we have got new design wins during the period. Given that we're talking about a fairly small number of key companies here, that's not necessarily something that we should expect to happen every quarter. In fact, you can see here that in the first quarter, we did have two significant design wins with companies that are currently red dots, but they're not public. We can't say at this point who those are. We're also working all the time on re-equipping other companies who have market leadership positions in these markets with our latest roadmap technology. We re-equipped three of the blue companies and one of the green companies in Q1. In Q4 last year, we re-equipped five companies, as you can see, with new roadmap technology.
Our plan is we will keep this slide up to date as things develop. The dots won't change color until information becomes public. There's always been quite a lot of interest, really, in the market in the impacts of consolidation. What's the impact of consolidation in the semiconductor industry on the IP business? If we look at what's happened over the last 15 years or so, there's been clearly lots of M&A activity, but that has been balanced by the emergence of new companies and the disaggregation going on in the industry into more specialized companies. Actually, the relative number of top 5, top 10, top 20 hasn't really changed that much. If anything, the number of companies outside the top 20 has increased over that time. The licensing opportunity has, in fact, been affected in a positive way, really, rather than a negative way.
Consolidation events can often benefit Arm Holdings directly. There are some examples of some recent acquisition events on the right-hand side there. When the acquiring company uses Arm echnology already, when the acquiring company uses Arm Holdings and the acquirer uses Arm , then often part of the acquisition rationale will be a cost synergy theme around standardizing on Arm , which is good for the adoption of our technology, of course. Even if the acquiring company doesn't use Arm everywhere, bringing Arm Holdings technology in with the acquirer is also a good opportunity as we work with these companies. In general, consolidation has been beneficial for Arm Holdings. In this last section, I just want to take a look at some of the drivers of increased value that we're seeing from the licensing and royalty business model.
Firstly, as the complexity, sort of the technical complexity of creating these processor cores and physical IP and graphics cores increases, the development cost increases, the value of the ecosystem around the core also increases, which means that the value to the customer is increasing. That means that the technology commands a higher license fee and a level of higher royalty as well. Each generation of technology, we're looking to deliver more performance at a greater level of power efficiency. That's difficult. It has value. As a result, you can see on the right here that the core set for the Cortex-A class processors, typically for a perpetual-type license, is commanding about 2.5T the value of an Arm 7 license. It demonstrates an increased value. There will be a level of increased royalty associated with that technology as well. We're seeing strong demand for the Cortex-A class processors.
We have more than 35 licensees now of the Cortex-A9 and eight licensees of the Cortex-A15. Both of those cores are being designed into a wide range of products ranging from smartphones, mobile computing, digital TV, to servers. We're seeing very good traction for both of those products. Overall, for Cortex-A class processors, we've seen that licensing has ramped up very strongly, as you can see, 68 licenses signed in total for Cortex-A class processors. As of Q1 this year, only 4% of our total unit shipments come from those licenses, in fact, from 16 licenses within that 68. Clearly, a lot of volume opportunity to come through going forward. Looking on the royalty side, the average percentage that we see per chip is starting to increase. I believe there's three reasons why that's happening.
We're seeing the Cortex-A family of processors can command a higher royalty for the same reasons as I've just explained, driving the license fee value. We are also seeing that an increasing number of chips out there are containing multiple Arm-based processors, which typically results in a higher royalty rate for those chips. Finally, where we are adding Mali customers or licensing Mali graphics and physical IP, that's then generating an additional royalty into the mix on top of the others. Clearly, in some cases, we will have all three of those factors coming into play at the same time. That's what we want to see, with a significant increase in royalties resulting. To sum up, the business model that we have, the business model that fundamentally we've had for the last 20 years, it's been a very powerful business model. It continues to be a powerful business model.
It gives us a wide range of licensing options to enable people to have access to Arm technology, really, whatever type of company they are and whatever they're trying to do. The licensing that we do increases our future royalty potential. The latest advanced technology that we are creating commands both a higher license fee and higher royalties. On that note, I believe we're now going to go into a short break until about 11:15 A.M. when we'll reconvene for the next session. Thank you very mu
I'm Mike Müller, the CTO of Arm . I've been with Arm since before the beginning. I'll make a few historic references as we go through. I was supposed to, it was okay if I lost the audience once or twice in my presentation, but hopefully, I won't do.
I thought I'd start by saying I am old enough to have started by learning to program on mainframe and minicomputers back in the 1970s. What I wanted to go through, something Warren talked about last year, is how the market sizes have grown as we've gone through different eras of computing, from mainframes and minis with relatively small volume shipped into the PC and kind of desktop internet era pushing a billion units. I'd like to thank some of our competitors for the wonderful innovation that that part of the market drove because they moved from the one form factor of mainframes, which is basically big boxes, to a glorious two form factors of a portable box and a small box on your desk.
Fortunately, moving on to the mobile internet, which is what has driven a lot of the growth of Arm Holdings in the last few years, there's been an explosion of the number of form factors that the market can address. As we look forward into the Internet of Things, there is an unspeakable number of different form factors that we'll see Arm electronics embedded in. Taking that kind of growth from a computer science view of what's happened to computing, I thought I'd look at what the drivers for that were. In the 1960s, 1970s, and 1980s, it was really about functionality. How could you get your systems to do more? That was the thing that drove that market. As you moved into the PC era, it really became functionality divided by dollars as cost entered the equation.
We then entered a mobile internet with the laptop, and really, the driver there was functionality divided by power and dollars because actually getting the thermals to work inside your product was the real challenge. What drove the real mobile internet revolution was energy on the bottom because it wasn't how hot it got. It was really how long could you run the battery for, and that's the difference between power and energy. It's power over time is energy. The new driver in that market was you had to have growing functionality, you had to reduce cost, but you also had to find a way of managing energy. That's what we grew up doing, and that's what our track record's been about.
If you look at that as the metric that has driven things in the past, what I claim is that that is what's necessary for all markets going forward. Whether you're looking at the obvious end with the Internet of Things and tiny, embedded devices or all the way through to servers, what's limiting the growth in servers is how much heat you can get out of buildings, how much it costs to actually put all the energy into those systems, and physically how you can pack them together. Whether you're doing small things or big things, energy has become the real driver for doing system design. That's something that we know all about. As Graham, I think, said earlier, it's a design philosophy.
It's not something you can just go to an engineering team and say, "Okay, can you now do low energy, low-power designs?" It's about a mindset and a philosophy. It takes a long time as we absorb engineers from outside of the company and get them to realize that the design trade-offs you make aren't the natural ones you expect if what you're driving is for performance. It takes a long time to re-educate people into looking at all the choices they make and how they actually drive engineering. It's not something you can easily transition in a company. It's also a design reality in that the Arm architecture is clean, simple, risk, optimized for making it simple, easy to build, low power. If you're trying to drive something like 8086 legacy and a complex Cisco architecture, you carry the burden of all that backwards compatibility.
However good your engineers are, however good your mindset is, you're carrying all of that baggage with you. There's a design reality about low power as well as a philosophy that goes with it. It's also about partnership in that it's about time to volume these days, not time to niche. I'll talk a little bit about servers in the future. It's a niche market. It's a few million chips. You can afford to do things in that market that you can't do if what you're really trying to do is deploy volume in hundreds of millions of units. You know speed bidding where you kind of make a few chips and go, "Oh, some of these are fast.
I'll sell them for a premium because there are some people out there who will pay extra for the extra 100 MH for their laptop." I have no idea why people do that, but they do. That's not something you can transition into a volume play. It's about manufacturing to volume, not manufacturing to early niches. When I look at this, I don't look at this as a historical view of how computer science has evolved from the 1960s through to today. I actually strip off the timeline because I actually think all of that is still relevant and all of that is applicable to the Arm markets. It's a question of expanding both into what some people see as the past, which is servers, as well as into the future of the Internet of Things.
I think all the speakers today will say we actually intend to have that complete spectrum of computing. That's one of my takeaways for today: it is everything from servers through to the Internet of Things.
Just so you understand how I look at the world of technology, and how that drives some of the things we do within Arm . Some of you are familiar with the good old-fashioned PC. Fortunately, that is being replaced entirely by Arm-powered products, both in the mobile space and all the way back to the desktop. If you look at the evolution of that around the home, which is another significant driver, there's lots of electronics you never see. The electronics you do see, of course, is your television. There's a lot of interest in moving that forward to 3D TV. I have to say, for me, it doesn't do an awful lot. What I'm waiting for to be a real revolution here is kind of the Star Trek transporter stroke holodeck, where you actually have really engaging 3D virtual reality.
When you have that, I think there are transformational things you can do in terms of what does person-to-person presence mean? How do you really do video conferencing? How do you have a gaming industry? That will be transformational. Until then, it's just a flat thing on your screen. The other end of the spectrum is this whole world, the Internet of Things, all of those embedded objects that you don't think about, the electronics that will be built into the concrete of your building with non-replaceable batteries that are not designed to be maintained. They are designed to tell you after an earthquake which bits of the building really did have structural damage and which didn't. Truly embedded into the environment around you. What brings all of this world together for me is ubiquitous environments.
That is networking at a micro scale, as I say, of all of the sensors in this room, bringing those together so that I seamlessly interact with it. To make that work, you need this kind of compute and storage cloud, which is what people think of with Amazon Web Services servers out there and high-speed Internet access. Once you have ubiquitous environments and compute and storage, you're building it together. Hopefully, for the investors in the room, some of you are managing a diverse portfolio because I actually believe all the money gets made up here in the services. While Arm Holdings is a great business, I think you need to balance your portfolio and say, who is using this? Who is extracting money from this? That service play is a significant part going forward.
I do another presentation I don't have time to do about washing machines and how there are people trying to do trials on how you buy washing by the load so that you drive sensible choices on product design for recycling and maintenance, rather than how do you build the cheapest box I can get into somebody's house, but when I throw it away, it's the consumer's problem, not the manufacturer's problem. Even things like laundry can become a service offering to drive total lifetime cost and recycling as low as possible. It's in the service. I want to start at the very other extreme of that, which is the process, the manufacturing, the transistors.
Within Arm , I've got a research group who take early designs of a process from a foundry and do various development chips within R&D over a period of a year or so as the process actually evolves and grows. It then transitions into a product business, and the product groups do initial trials of the IP design that they're going to sell and then do a final product release validating what they've developed and then prove that with actually taking out a processor. That process of product development is about another year on top of the year or two that's happened in R&D. That's kind of the life cycle of 3nm for Arm Holdings, which is now rolling into volume production. There was then a process shrink and a marketing war that said 32 is the wrong number.
It has to be 28, so there was a refresh for 28 nm, quite simple in R&D, not a lot of work to be done, could skip the alpha PDK stage, went straight to IP validation and a process of proof point. Some of the industries then pushing 28 as the current volume node. If you come back from that, R&D is working on the 20 nm stuff today. Last year, we taped out a 20 nm test chip with Cortex M0. That was last year. We're about a year away from the product cycle rolling through to mass deployment for 20 nm. If we look at that, that little chip that we taped out last year and blow it up and rotate it, it's a test chip with lots of tedious structures on it.
Part of it I'd like to compare to the very first Arm Holdings chip that we taped out in 1985. That chip in 1985 had 6,000 gates on it and was about 7 mm on the side. The test chip we made has a little processor subsystem. Within that subsystem is Cortex M0, which is our smallest product that we sell today. That's 8K gates. I sometimes look at it and go, 25 years of engineering to go from 6K gates, 8K gates, what have we been doing with our lives? The answer actually is quite a lot because those 8K gates do not look anything like the first 6K gates because we're solving a different problem this day. Today that Cortex M0 is designed for embedded control, not for building desktop computers, which is where we started with the original ARM 1.
We've actually reconfigured what the world looks like. It is, however, an awful lot smaller. If you draw it to scale, it's about the white dot in the yellow circle, and it's about 1 0,000 of the size. That's kind of Moore's law of manufacturing in action. The smart people in the room will be asking, why aren't there 10,000 processors on a chip if it's 10,000 the size? This is my one piece of math for you, but I'm told as investors you can do math. It's Amdahl's law. You could go, I've got 10,000 processors. I put 10,000 processors on the chip. It'll go 10,000 times as fast. What Amdahl said is, yeah, 32 processors, it all should work. You can only speed up the parallel part of a problem by going parallel. At the end of the day, you've got a serial part that's left.
If you take a problem, you split it up into a million pieces, at some point, you have to add all those million pieces back together. It doesn't matter whether you have a million pieces, 1,000 pieces, 10 pieces, or 2 pieces. There's a bit where you have to add it all back together, and the adding it all back together is serial. You can't parallelize that because what you're doing is aggregating everything. Your speed up is limited by the ratio of how much of your problem is parallel and how much of it is serial. The example here would be if 75% of your problem is parallel, you think that's an awful lot. 75% of my problem can be done in parallel.
The answer is you can only make that go four times faster because even if you have an infinite number of processors and you do 75% of the work in no time at all, you're left with a quarter of the work that was where you were serializing everything. The fastest it can go is four times, which is why some of you have bought Core Quads for your desktop machine and probably don't see much difference between that and your Core 2 Duo. If your problem doesn't have a lot of parallelism to it, you can't extract very much from it. 10,000 processors is great if you've got 10,000 things that you can do in parallel and they never have to talk to each other. If you have to add the problem together at some point, there's a real limit in what you can do.
Servers are one of the few places where you can look at it and say, maybe I actually have 10,000 things I can do at once without having to add it all back together again. In conventional products, that's not true, which is why you don't see hundreds of processors in conventional products. Go back to this original ARM 1, which we built for desktop computers. That's where Arm started. Let's compare it to a Cortex A9 SoC, which is the kind of thing that you will see in those products down there. When you look at that, yep, there's a processor, but there's also a lot of other components from Arm, whether it's a sophisticated 3D rendering engine or Mali 400, whether it's part of the system peripherals and memory interfaces or general I/O.
There's a significant amount of IP needed to actually build what is, you'll notice, about the same size as the original chip. I do another presentation that says all volume chips in time have been about 50 sq . That's the economic point and has always been the economic point in manufacturing. Take the processor core in that. If you take our product and build it out of the box, you get just under a gig. Fortunately, we can sell you some optimized physical IP, some transistors, some special gates that allow you to push that to about 1.3 GB. If you're allowed to use mixed VTs and push the processor a bit forward, you can use our POPs to build yourself a 2 gigahertz product. Starting to have what is serious computing performance.
Of course, you can take something like the Motorola product, put it in a dock, HDMI connection to your display, and a wireless keyboard, and you have yourself a desktop computer. There was this little fun piece of benchmarking that said the existing Apple iPad 2 is about the same performance as a 1987 Cray supercomputer, which used to be the world's fastest computer. We've kind of gone full circle in many ways. The hardware, I was told I wasn't allowed to say is the easy part. What I have to say is the software is difficult too. James will be talking about how we build a whole software environment that goes on top of the hardware. What I wanted to say is it's all about designing the right core. If you take a processor and we have our physical IP POPs, you can make it go faster and faster.
It comes a point where you turn up the voltage, you start putting on fans, you turn up the voltage a bit more, you can push a processor just so far, but the power it takes goes exponential. At some point, you have to stop. The last few % are really, really expensive. We live in a world before you ever get to that point. If you want to go faster, you actually have to design yourself a core that's more complex. That complexity allows you to extract more performance. We have a roadmap of products that become more and more complex. I have to confess that the gains are getting harder and the power challenges are getting more significant. Therefore, more innovation is needed going forward. This isn't a question of just cranking the handle and doing another design that's bigger.
You actually have to work harder every time to get the gains. At the end of the day, you have to remember to design the right core because those bigger cores are less efficient for simple workloads. The little core won't run as fast as the big core. If you've only got a modest workload, then you need a modest core because the difference on the power scale on the left-hand side is factors of two, three, four in actual usage. That reflects in your real battery life. That's why we have a complex roadmap of existing products that scale from tiny things from embedded control through to 2 GHz processors at the top and beyond. Those of you who follow Arm Holdings closely ought to know that we don't make processor announcements at Analyst Day. This is our public roadmap.
It needs to scale from very small to very big. This isn't a game where one size fits all. If I come back to our Cortex M0 20 nm test chip and compare that with a 20 nm Cortex A15 2.5 GB processor, again, you can see physically, we still scale from very small to very big because we're trying to encompass a very, very diverse portfolio of end products. I want to talk about servers. If you take this data, I'll try and make sense of it for you. Here is Moore's law. The number of transistors, and this is for processors designed for the server market, the transistors go up and up in a broadly straight line following Moore's law. If you look at their performance and their frequency, in the last few years, it's flattened off.
One of the real reasons for that is you've hit the power burden. You could make things that go faster, but they'd burn more power, they'd get too hot. You can't sensibly build server systems that utilize that power, and therefore, it's rolled off. Cores aren't getting faster. They aren't really getting that much more performance in that server space, which is opening up a good opportunity for Arm Holdings where we come in below that. What has gone up is how many cores you have in a server. A server is in a box. How many processors do you have in that box? That's the number that started to go up. People are starting to scale out the number of CPUs they have. It's not about designing the world's fastest CPU anymore, which used to be the game. My server chip's faster than his server chip.
That's now changing to how many of these can I fit in a box and sensibly get the air conditioning to work. That's become an energy game or a low-power game, and that's where we play. The next thing to remember is the server opportunity is relatively small. There's like 50 million servers out there, maybe 10 million servers a year being deployed. If you're charging $1,000 a core, that's quite a good business to be in, but the numbers are relatively small. Power is up to 30% of the OpEx for a company to actually run big server farms. Acquisition cost is no longer necessarily the dominant function of what costs. 15% of the market is made up of what you might think of as the web hosting, Facebook hosting search. That's the market that I'm interested in initially because it's the one where there has to be scale.
It's where new startups have a new service, and as they get more customers, they need to replicate that service to more people. The way to do that is to have more servers. It's not to have a server that is a bit more powerful. It's much easier to say, I'll have a limited number of users on each chip and just replicate the number of servers I have and scale out the number rather than trying to push the performance. If you're doing scientific computing, you might go, I'd like a CPU that goes faster and computes faster. This is about being able to scale things, and physically, the easiest way to do it is to just put down more boxes.
That, again, is why there's an opportunity for Arm because as the world changes from I must have the world's fastest server to I must have low-energy servers that I can replicate and scale, you start to look in the design in a different way. It's no longer about big, expensive processor chips stuck onto boards. It's about how do I architect that system? How do I integrate the processor with the I/O, with the disk control, with the interconnect to actually make a system that I can start to replicate? You change the way you look at server architecture. You change what you're going to build. That is all then about building SoCs, not about building really large processors that go as fast as possible. The whole kind of system architecture changes and plays into SoC. It's about how do I get the highest performance per dollar?
How do I get the highest performance per watt? Joining those together is how you reduce the CapEx of running your big server farms. An example of that, we invested in a startup called Lazada a few years ago who are building those kinds of low-power systems. If you take that low-power system, I like to scale it up. Here's an environmentally friendly project where you go to the pole with a drilling rig and you drill yourself 70 boreholes 2.5 km deep. You lower down each one of these boreholes Arm Holdings-powered sensors that are all joined back together to a porter cabin, which is in the middle of nowhere in the middle of that picture. You wait for the ice to freeze.
When it's frozen, you've got a 1 km³ of Arm Holdings-powered sensors under the ice cap, which is a giant neutrino detector, all wired up to a single porter cabin. I claim it's a single computer. It's not some distributed network. That's a km³ . That's kind of the upper bound of my aspirations for servers, a km³ At the other end, we've done a project with the University of Michigan, which is a little 3D stack of a battery, a microprocessor, and a solar cell. That is about 3 mm on a side. That's 9 mm³ . That's how you start to build these little embedded sensors. That was last year. The current project is a 1 mm³ platform. The reason we've set a target of 1 mm³ is that's the largest that an eye surgeon will implant in your eye.
This is actually a little glaucoma sensor at the bottom that's measuring the eye pressure, a little chip that does the data logging with the processor, and a comms chip on top so that you can get your Tesco's RFID scanner out and scan it near your eye to retrieve the data. This chip wakes up every 20 minutes, logs what your eye pressure is over a day because apparently, if you've got bad glaucoma, the drug regime is really important. If you walk up the stairs coming to your doctor's surgery and they measure the pressure in your eye, it's no relation to what it's like during your normal working day. They need to do data logging. Cubic millimeter, it's as small as my aspirations are. I like Arm Holdings to scale cubic millimeter to a cubic kilometer.
I've said how in the past, functionality divided by energy by dollar is a driving metric. Going forward, I'd like to change energy to available energy. If you're going to do some of the Internet of Things applications, you have to scavenge the energy from the environment around you. If you're embedding it in the concrete, your battery life might be long enough. As long as there's humans out there, they'll be using electricity, there'll be a Wi-Fi network on, there'll be hum from the mains wiring, there'll be vibration in the air conditioning. All of those things can be used to scavenge energy to drive sensors. To get true deployment of that Internet of Things, you need to look at functionality divided by available energy and cost.
In the future, somebody might solve all the problems of the universe and get back to being able to just be engineers and worry about functionality. That will take breakthroughs in semiconductor technology that FinFETs and 3D transistors isn't. Maybe battery technology, charging, so you can charge a battery for a lifetime in a second. That would transform what products you could build. Until that happens, we're actually going to have to worry about real physics and real engineering. In summary, low power, low energy is what's positioned us for the future. It's a design philosophy and a requirement. It's all about building better systems on chip. It's not about building just better processors. That's an easy game. It's actually about building those systems. Those systems are about partnership because you're not going to scale from a cubic millimeter to a cubic kilometer by partnering with one or two people.
That's a really diverse problem that needs hundreds of different partners to address the complete range. Thank you very much.
Hello, everybody. I'm James McNiven. I look after the software ecosystem for Arm Holdings. The majority of my time actually is spent with my team working on the Cortex A ecosystem. That involves working with some of the big software companies that you may have heard of, like Microsoft, Adobe, and Google. It also encompasses more than that. My team is responsible for defining the roadmap for our own internal software engineering enablement, which we do. Hopefully, you'll see some of that today. The aim of today was to give you a little bit of insight into what we've been doing with ecosystems and where we're going there. Next, to look at some recent changes and some recent news in the software ecosystem. I'm going to finish off with a little bit of the Mali ecosystem as well. Now Tudor's not here.
I can say that he actually missed something out from his diagram this morning. In this 20-year or more lifespan for Arm Holdings products, what you really need to be very successful during that is, of course, I'm going to say, an ecosystem. This really encompasses a lot of different skills, a lot of different companies, different market segments, different regions. Some examples just from up there, everything from software tools through to operating systems to EDA tools, and even ones that aren't mentioned up there, like training. Something vitally important for all those software engineers out there is training services. With that, the consequences of the Arm Holdings business model that we talk about, Arm Holdings is only profitable if our customers are successful. For our customers to be successful, they need a healthy ecosystem to support them.
Something that we've certainly learned over our time with the ecosystem is that we do have to leave space for other companies. Those companies have to add value. They have to innovate. They have to make a profit. That even includes with Arm Holdings companies that actually compete in certain areas as well. A good example is software development tools. Arm Holdings has its own software development tools. My team are out there enabling other software development tool companies to make sure they support Arm Holdings really well. Those companies may be doing something different, something that the engineers particularly like in a different market that we haven't thought of. That's a good thing for our ecosystem as a whole. It is really for us, as I say, creating a nurturing ecosystem. We really have to understand what our customers need.
My team spends certainly a lot of time with end users, application developers, OEMs, and others, trying to work out what they need, what they're going to need in the future as well. We have to enable that to make sure that everybody's successful. If you look at the ecosystems that we build and evolve, that takes time. In fact, what I'm going to show you is there are actually several phases, distinct phases in the way we engage with an ecosystem. Phase one, we typically try and understand the market. We bring existing technology to bear, something that we already have developed. We might prototype something. We might invest in a startup, et cetera. That is what I class as kind of a phase one of our ecosystem.
The second phase, when we start to get some product and traction out there in the market, we start really working with market leaders. We start to then spend more time educating the rest of the market and certainly start really linking together the ecosystem. We start connecting all those different companies together so they can start innovating, making money, et cetera. What we aim to do then is move into the kind of third ongoing phase, which is really the life cycle of optimizing technology from Arm Holdings, engaging a broader ecosystem, co-marketing, training, et cetera. If you look at the right-hand side of those logos, I just wanted to say that if you go back over the last 10 or so years, you'll see that if you just look at those charts, the logos seem to have grown in number. That is kind of a nice thing.
In reality, though, every one of those logos represents a business that makes money, relies on the Arm Holdings ecosystem. That is incredibly important for us. Each one of those logos consists of 10s,100s , or 1,000s of engineers. It is that that I think really helps us in that innovation and differentiation that has come to grow the Arm Holdings opportunity. What I wanted to do really was just to take you through a little bit on the smartphone ecosystem. I think that's probably the ecosystem we're best known for. Obviously, this is something that we've been working on for years. If you look at the ecosystem as a whole, again, you can pretty much identify those three phases I spoke about. Phase one, we actually started by reusing some Arm technology. The ARM7 was actually not designed initially for mobile phones.
We worked with a thought leader there, learned a lot about the market, and it started off. Phase two, we really optimized our technology. We added some caches to our processors to make the software go, excuse me, a little bit faster. We also then started working with more market leaders and started growing that kind of base of Arm designs. Phase three really was when we kicked off into really optimizing our technology, getting those physically tagged caches so that the operating systems run faster, getting Java acceleration into our processors so the applications could go faster. That's when we also really started working very closely with operating system companies, middleware companies, and really connecting that ecosystem and making it into the broad ecosystem that it is today. That is a pretty hard ecosystem to replicate now.
Those relationships between those software companies, between those OEMs, and silicon partners have gone on for a very long time. That's certainly a key strength, we believe, for Arm. Extending that then into another market, hard disk drives. Going back 15 years, the market for hard disk drive controllers and other disk drive controllers was incredibly fragmented, representing an opportunity for Arm. We engaged then back with some thought leaders like Western Digital and Seagate. We reused technology, again, that kind of ARM7, which again has been pulled out and reused for another market. We invested in a startup and learned an incredible amount then of what was really needed to be successful. Going forward, we worked to optimize our technology. We improved the processor designs. We also brought something that we thought was of unique value to that ecosystem, which was this trace capability.
If you're a software developer debugging a hard disk, the last thing you really want to do with that head spinning is stop the processor because it tends to have some pretty catastrophic effects. We started introducing those key kind of technology ingredients that would help us grow in that space. The third phase was optimizing further. Cortex R4 was released, really helped start to accelerate our market share to the 85% it approximately is today. If you look at that as well, a fascinating point from that is that technology also now underpins other segments. Not only have we developed something for the hard disk drive market, but that now is used in engine management systems and in other segments to make sure you can debug your software and continue to optimize your product. Another example, MCU. Again, 10 years.
It shows you the time scales we're working on here. We saw an opportunity again because of two reasons. One, fragmentation, but also that growing complexity that we recognized in microcontrollers, that growing software stack really requires more complex languages for the developers to work with. That was the opportunity then for Arm Holdings. Again, we reused a bit of technology. I think it was the ARM7. We started working with some thought leaders in the market. We invested as well in a startup and, in fact, acquired Keil, the tools company, back then as well. We learned a lot. We created the Cortex M3 as our second phase. We started investing in those software developers and so on and so forth. Now we've introduced the Cortex M4. We think that's going to give us a very strong foundation to grow in MCU.
Hopefully, at this point, you're seeing the pattern in what I'm trying to tell you about ecosystems. They take time. There are distinct phases to it as well. You have to be in it for the long haul. For servers, we think there is, as Mike mentioned, a good opportunity for Arm Holdings. We're looking to apply the lessons that we've learned over several different ecosystems over many years to that ecosystem in servers. Supplier choice, supplier differentiation, and of course, the total cost of ownership, both in power efficiency and performance density, is something that is hopefully what we think will be a good differentiator for us there. If you look at where we are, two years ago, or two and a half years ago, we set up an exploratory team, a team of R&D people and marketing to go and work with some of those thought leaders.
That was actually based on a Tier-1 OEM request. A year later, we had prototyped a system. It was a small web server prototype working with one of our partners there, Marvell. We've been running a small part of the ARM.com website on that for the last about 18 months now to get a lot more background data and learn about the market. Last year, we set up a dedicated team to do that and have been really working on validating that TCO and optimizing some technology and started doing that ever since. Last year, we also announced the Cortex A15, which you'll see has some virtualization technology that is not needed certainly in every server segment or server usage, but certainly has some interest in the future there for those particular use cases. Going forwards, what do we have to do?
Going forwards, we certainly see as the market expands and the total addressable market expands, we're going to be working with some software companies that we'll need. Those include virtualization vendors. You'll also see us working with enterprise operating systems and enterprise runtimes like Java for enterprise if we're doing our jobs right, that is. Over time, you'll see us hopefully growing the addressable market from the 15% or 20% that Mike mentioned onwards and upwards. It will take time. We're certainly in it for the long haul. If you look at that kind of software, certainly for server, we're able to reuse a lot of the existing software stack that's out there. For running our web server inside Arm, we could use stuff that had actually been developed for that mobile ecosystem. That mobile ecosystem has effectively underpinned our server efforts so far.
If you look at the software development costs, they are rising actually faster than the hardware out there. That report up there is showing that pretty much a 45 nm was the crossover point for software costs versus hardware costs. That is driving more reuse of Arm technology within our silicon partners. You'll see that because of those costs, people are looking to reuse that software, reuse their hardware as well. We're seeing a trend accelerate. The home segment, for example, we're seeing that companies like ST, MediaTek, Samsung are all benefiting from the fact that a lot of that software work has already been done for the mobile market.
As we get new OEMs and silicon partners coming to us today, we can certainly have a fantastic discussion around building smart devices because that internet that they require, the browser stack, the networking stack, etc., has already been written and ported for Arm. We also think that the software development costs rising also really helps with the model of this disaggregated industry. Software partners for us are seeing more and more opportunity around Arm. I just wanted to tell you about a couple of examples here. This shows the complexity and sometimes the requirements to work in different regions and in different segments for us. A couple of companies I want to highlight are actually based out of China, a company called Shanda, a member of our connected community, very important to us nowadays. Traditionally, we'd have just worked with middleware and operating system companies.
This company, Shanda, has an online gaming presence in China. There are 10 million users paying monthly for their services. It is critical for us to help pull through those kind of OEM devices and get a good design win there. Another company, UC Web, has a product called the UC Browser. I'm not sure if many of you use it. It's certainly recently expanding out to international markets. They have statistics of 700 million downloads of their browser. They have 200 million active users. Every month now, they're seeing something of the order of 70 billion page impressions, which is incredible statistics and just shows you how complex that ecosystem is getting and how important it is for us to make sure we do worry about those different regions. We do worry about those different market uses of that software. Consumer devices slipstreaming the mobile ecosystem.
I'd like to use an analogy for how the mobile ecosystem is really helping the Arm software story generally. If you look at cyclists in particular, the advantage of slipstreaming is as a group, you go faster. You take it generally in turns at the front to do the hard work when you have to. What we're seeing is that with the volume in mobile and mobile computing, other market segments for us are able to slipstream that ecosystem. This is a great example as home. Home is getting the benefits of those browsers, those plugins like Adobe Flash, etc., ported to Arm, lowering the cost of entry. I think this is one of those key kind of helps to that trend of bringing more companies back to Arm for usage beyond just mobile. I think this really is a virtuous circle of development.
If you look at the size of the ecosystem again for mobile, you have billions of users. You have billions of devices. We're now seeing, I think now over, there's now over half a million applications now available for mobile devices out there. There are now hundreds of thousands of developers, which of course are creating applications that, again, loops back and pulls through even more. That's an incredible story, I think, for the Arm ecosystem. As I say, it's helping in other areas. This is one of those slides about saying, I'm not sure it's my chicken and egg slide. Is it license reuse that is driving software reuse, or is it, in fact, the software reuse requirement that's driving license reuse? I think it's a bit of both.
What you will see at the heart of a smart device is the compute system is somewhat similar. A Cortex A processor, a GPU such as Mali, the Amber Interconnect, so our bus framework that we have as a product as well. By having that common compute platform, you're able to reuse your software investments, reuse your knowledge. Our licensees can take that design and reuse it across multiple markets, spin different versions, adding their own differentiation into it. Our OEMs get the advantage of getting used to working with the Arm Holdings processor, understanding its architecture so they can work their products faster. Our software developers as well, by learning about Arm Holdings, really can use that same application on multiple markets on different platforms. Even though there are different platforms, such as operating systems on these devices, underpinning that, again, is the Arm Holdings architecture.
Once you have them trained up for that, they're able to acquire or, sorry, apply that knowledge to all these different areas. I don't think I need to tell you that this is helping drive that convergence between smartphone, mobile computing, and home. Let's talk a little bit about what's been happening in the operating system web and application world. We've seen an ongoing or continuation of the fact that there are still multiple operating systems available for Arm Holdings. That's a great thing for differentiating in the market space. If you look at the chart on the bottom left, the operating systems, if we went back about two years ago to the analyst day, the bottom logos there, the Android logo through to the web, are operating systems we certainly spoke about at the time. The ones actually on the layer above that are new since then.
You can see there's actually almost an acceleration or a continuation of that differentiation. Earlier in the year at CES, Microsoft announced Windows on Arm Holdings. We've had Google at last week's I/O conference that they held, which is their conference, their big developer conference, announcing that they are working on Chrome OS for the Arm architecture. We've had entrants such as RIM, who acquired QNX last year, using that operating system for their Playbook tablet. That's a fantastic story again of reuse. That QNX operating system was actually previously seen working in automotive systems on Arm. A great bit of reuse and critical mass in the Arm ecosystem there. Google also announced that the Android operating system in Q4, the next release of iScreen, as it's called, will be a combined release for tablet, for mobile phone, and for Google TV and will be open source at that point.
Moving into browsers and middleware, if you look at those phases I spoke about earlier, we've really moved from a phase of enablement and working with those thought leaders to really one of kind of almost phase three, I would say now, really optimizing that browser experience on Arm. Our engineers have been working out there with the general software community. We've improved JavaScript performance by five times in the last year. If you go back a year and take your Android device at that time and compare it to one this year, the language that executes these applications in your browser will be running five times faster now on the same processor. Going forwards, we are now absolutely optimizing our products to the web. The Cortex A15 runs at the same performance. It runs at many times the level of a Cortex A9.
We're taking this very seriously to make sure that we continue to develop products that are absolutely fit for purpose and optimized for purpose. The third area I wanted to talk about was apps. If you went back a couple of years ago, I think an app, that was really the first App Store. I think Apple announced their App Store two and a half years ago. Apps have really exploded onto the main scene since then. Most of the apps out there are built to be optimized for Arm. They're built using native development tools to get the most performance and most efficiency out of the mobile device, and so they run on Arm. We also do work on other technologies as well. We optimize these JITs, these just-in-time compilers for non-native applications as well. We're making sure that we're absolutely covering all bases from that point of view ould say this is another point of the Arm size of ecosystem. We are very fortunate to be in a place where all those investments are done by thousands of different developers, lots of different large companies such as Google, which means that our licensees get the benefit of being able to reuse those investments and quickly get to market because the software is already built for them. Let's talk a little bit about Android. One of the things I really wanted to highlight here is that this has been a real phenomenon for Arm Holdings in the last year, in fact, the last year and a half. Android is now really the kind of consolidation point for open source on Arm Holdings. Android is based on Linux, but really has united a lot of the differences that were out there in terms of shipping and consumer products.
In fact, today, we're now seeing 400,000 Arm -based handsets being activated every day. So far, there have been, I think, 100 million Android handsets shipped out there in the market. A huge scale of growth in that market for us. The great news for us in our partnership is that Android is Arm optimized. From the low-level libraries up to the web browser, it's optimized and written to take advantage of the Arm Holdings architecture. The whole point of that is you can quickly get your new product to market if you use Arm Holdings. As an example, one of our engineers back in December took Gingerbread, which was the new release, I think, roughly at that time, and took two days to bring up that basic port on Arm Holdings.
One engineer in two days, which I think is a pretty good indicator of that reuse in action. What are we doing going forward? I always get asked the question about the software ecosystem. What are you going to do about the PC ecosystem? What do you have to do? For me, I think some of the key things there are peripherals and also some of those applications that still run on other operating systems for Arm . Peripherals actually are really moving now to being increasingly cloud-based. I don't know if you've tried using services like Dropbox or web-based printers. Basically, these are printers you can print to via email or via a web interface and storage that you can access via the cloud. Whether you've got your mobile phone or your tablet or your smart TV or your laptop, you can access your documents in one place.
This is really helping actually just open up more opportunity for smart devices, as well as these new cloud-based services to get a foothold in the market. In terms of Arm-based mobile computers presenting great opportunities, software as a service hosted in web browsers, so Oracle CRM products, SAP, et cetera, mobile-based computers are really allowing access to these services anywhere. What's Arm going to do going forward? The key thing we can do is to make sure we put the right tools in the hands of the developers. Those developers are going to be going off and doing all these wonderful things that we've never even thought about and could never hope to really understand actually at the beginning. What's critical for us is high-quality development tools so that those application software guys can go and create all the great stuff out there.
Arm has Development Studio 5 recently announced, actually, I think it was just at the end of last year, which is used for optimizing supporting software, including Android applications. We've invested in Linaro, which is a consortium of Arm partners who invest dollars, who invest engineers into working on optimizing the kind of low-level Linux software stack and also consolidating the Linux software work on Arm. Right now, that organization has grown from nothing to now 105 engineers in the last, I think, nine months. Pretty impressive progress there. Taking a look at Mali and the graphics ecosystem, I really think we've moved into phase two on Mali. We're now shipping in high-volume mainstream products. I think hopefully you'll see that the benchmarks are showing that it's incredibly competitive out there. What are we doing now?
Just like I said for the rest of the software ecosystem, we're going to provide the best-in-class tools, best support, best documentation to those developers out there. We certainly believe that Arm has that best address book in the industry. We can set up all those introductions between those PC gaming companies, PC software companies, and the mobile OEMs out there. How's Mali doing out there with Android? The good news for Mali is that it uses an open interface. It's OpenGL ES. What that means is it allows portability for applications to Mali or to other GPUs. It means there is no proprietary lock-in effectively like that. We've been testing a whole bunch of games very successfully. Hopefully, you can see some of those outside. There are occasionally issues that we come across, generally to do with proprietary texture formats.
The one thing I did want to highlight is that now Google and Apple are both openly recommending that the open standards used by the industry are now used. We have a process down now that, working with software companies who haven't done this yet, and there are still a few, but only a few now, we can convert that in about a day. It's been our record so far. Finally, I just wanted to highlight really what we mean by runs better on Mali. I don't know if you've seen this device outside, the Samsung Galaxy S2. I think there have been some absolutely fantastic reviews out there for it. Quotes like a smashing speed test ahead of launch, astonishingly fast, most powerful mobile handset we've yet tested. That right-hand screenshot is a standard benchmark out there for devices. You can run any device you want.
You can see that I'm certainly interpreting five stars as 5/5 is pretty good. If you haven't seen this device outside, please go and have a play with it, play some games, and see it for real. In summary, Arm Holdings has a very diverse, differentiated, and we believe healthy ecosystem. We've had very good success so far in mobile, disk drive, and microcontroller. We're applying that experience and our expertise now to server going forward. Software reuse, that software critical mass and that virtuous circle, I think, are absolutely enabling us in new markets and certainly driving the existing ones we're in. Going forward, I think Mali, as the graphics ecosystem, already includes the thought leaders. Now we've moved into that phase of working with market leaders and are now busy connecting the rest of that ecosystem up. Thank you very much.
I'll hand over back to Warren.
Thanks very much, James. I'm not going to take up much of the time. We need to get on to some Q&A. Just to sort of summarize where we've been, last night I was being asked a question by one of the sales side analysts who asked me, what is the one thing about Arm Holdings that makes it work and makes it really successful? I said that's an impossible question to answer really because it isn't one thing. It's a combination of things. What we've tried to show you this morning is indeed that combination of things. I think you'll see from Mike's presentation after the break, we really believe we have the best technology. Not just the best technology, but the best design philosophy. Remember, Mike talked about the low power is actually not just the best products and the best technology.
It's a design philosophy that is important spanning the whole range of digital products from the portable right through to server boxes. It isn't just the microprocessor. It's the system on chip design, the system design that goes with that. Before the break, Graham talked through our business model. We talked about the consistency of the Arm business model over many years, consistency combined with flexibility so that we can use our licensing as an activity to penetrate new markets and build our market share and therefore build our royalty for the future. James's presentation is fresh in your minds. I hope you've remembered there that it's all about time to market. Building ecosystems is a long-term game. We have been in this for 20 years. We have been building the Arm Holdings ecosystem for a long time. We've looked at different market sectors.
We've transferred lessons that we've learned in building the mobile ecosystem to transfer that into other ecosystems. We're replicating the process. The ecosystem is what really enables Arm to be what it is so that when we look at the growing opportunity that Tudor presented at the start of the morning about how we become truly the architecture for the digital world, all of those three best, best, best components go towards making that happen. How we can then scale from a tiny internet of things right through to, as Mike put it, kilometer-cubed computers. There is one best that's actually missing off this slide. I just wanted to thank Ian. You all know Ian Thornton. Thanks to Ian, the best VP of IR, for putting this whole session together. With that, I'll ask Graham to come up the front and Tim.
Tudor's had to disappear, but Tim's going to stand in for Tudor. He's a bit taller, but he's very, very versatile. Now we can do the Q&A. I think we've got the first one here. What we'll do is please ask your question. I'll dish it out to the most appropriate person to answer it.
Just give them a whole question if I may. Firstly, you talked about the server opportunity. Could you just talk a bit about what the remaining challenges that you have to overcome there? Is it mainly the software side or the hardware side? Where is the investment? Secondly, you talked about software, the benefit you have in using it across the ecosystem. What type of reuse would you typically get moving between, say, a TV or a mobile device? How does that compare to moving to different architectures?
OK, the server opportunity, I think Mike's going to answer that. While he's doing that, James will think about the TVs.
I think definitely the software side is a longer time to market than building the right hardware. It's building that software community and a confidence among aimed OEMs that this is something that they can adopt. It's relying on customer pull to pull through those solutions because at the moment, we're not the incumbent supplier. Therefore, there is perceived risk behind it. That takes time to demonstrate real benefits in the market before you get other people adopting it. More than the software, it's probably that cycle of design into product, actually being shipped to customers, seeing the benefit, it becoming a public thing, and then going back round. That's probably longer than the actual software, and the hardware is the easy part. It always is.
OK, and then the question was on how much can be reused, I think, from mobile into the digital TVs. Is that right? I think I'll probably, unfortunately, give you the answer as well. It depends exactly on what you're doing. Certainly, the lower levels can be reused. Actually, the highest levels are certainly reused, and we've seen that. The question is really how much of the middle stuff you're doing to differentiate yourself. I'm afraid that that's where I probably won't give you the best answer. Just from a browser point of view, we're seeing people just take the mobile browsers, reuse them, or reskin them, and the core engines are exactly the same.
I think the second question I'm looking at is, comparison? Sorry.
If I compare it to moving to a different architecture, what the challenges in that slide would be? If you move from another architecture to Arm or from an application on Arm to a different architecture, say Intel?
From an application running on top of a device like a mobile phone, it depends on the application. Some apps are written in languages that will just run on any processor. The majority of apps are written for Arm. You would have to rebuild your app, potentially rewrite any complex routines you've written. It could be anything from some apps to some simple apps, very little to highly or huge amounts of work, to be honest. It does depend on that app.
Yeah, the next one.
Morning, Digi Samar from RVS. Two quick questions and one clarification. On your 2020 cumulative shipment targets, do I understand correctly that you've actually upgraded your guidance? There was above 100 billion before. Is that about 150 billion? That was my first question. The more interesting points on the Intel announcement of 3D transistors and FinFETs. How do you see that relative to Arm Holdings and let's say your angle, which is low-power CPUs versus process technology? How reliant is the Arm Holdings manufacturing camp on getting EUV right? Will you need other materials to complete and catch up? The last question, I don't know if maybe for James, at the turn of the year, we had the two major computer OSes not running on Arm Holdings. One of them has converted to Arm Holdings.
I'm just wondering if your efforts, at what point do you think your efforts will pay off and you get all the PC or computers OSes running on Arm Holdings? Thank you.
OK, so just on the subject of 2020 volumes, and then Mike's going to talk about FinFETs before James answers the other one. We have historically not come out with a volume for additions very often. We highlighted the few times that we have done that. You're correct to observe that we changed this morning from greater than 100 billion to approximately 150 billion. It's 10 years out. There's a lot of variation that's going to happen between now and then. When we realistically look at the growth that we talk about on an individual segment-by-segment basis, we sum to a number that's much closer to 150 billion than to 100 billion. We're putting it there as an approximate target. Mike, FinFETs?
OK, FinFETs in what, two minutes? Are FinFETs, to misquote someone, magical and revolutionary? No. If you look at the ITRS roadmaps, are things like FinFETs necessary to get where we need to be to get down to 11 and 7 nm? Probably, along with about four other impossible things before breakfast. It's just another piece of technology. If you look at the major foundries, they've said they're going to do it at 14 nanometers. Intel said they're going to do it at 20 nm. As I said earlier, it's at 20 nm for server parts, shipping in hundreds of thousands of units. What are the issues to getting that deployed to volume? Will they actually get real volume on 20, or will it be at 14? Don't know. My guys pulled an interesting statistic.
They did a quick patent search for any patent involving the word FinFET, which is the technical term rather than 3D, Tri-Gate. The interesting results were, last, Intel 90-something patents, second, TSMC 140, first place with 370 or something, IBM. It's a technology that's been around for a long time, been biding its time for when the manufacturing issues get hard enough that you have to bite the bullet and do things like that. Will everyone be doing it in the future? Probably. Does it transform anything in the short term? No.
OK. On the PC side, certainly from a Windows point of view, there have been no announcements on timeline yet. In fact, I think it was about three weeks ago, Microsoft said, you know, come back at their developers' conference in September, and we'll have more information for you then. I'm sorry I can't comment on any other rumor or speculation.
OK, right. I think we've got one over here, and then we've got one towards the bank.
Hi, Anil Dharadhala from William Blair. You talked about some of your efforts beyond the top 20 Chipset companies. Can you comment about what are these smaller companies, I mean, trends that you're investing in, new trends with some of the smaller chipset companies? I've got a couple of follow-ups. From your vantage point of view, what do you think are some of the breakthrough areas in the semiconductor software and hardware level beyond some of the Intel announcements? Finally, the rise of Mali, do you believe the dynamics here inhibit you guys from succeeding in this space? Do you think the dynamics are pretty much similar to the rest of the industry, given the fact that you're competing with some of your customers here?
OK. The first question was about trends in which we're investing. Graham, you may want to add something to what I say on that. Mike, I think if you pick up the breakthrough question and James come back on Mali. The investments that we make, I think you probably saw a little repetition in some of James's slide. We're interested in penetrating markets. Arm is very strong in the mobile space today. You can see other applications in the spectrum where we have much, much smaller market share. Microcontrollers, we're in the sort of early teens % today. Servers, we're nowhere today. The investments that we make are all about taking Arm Holdings into those new application areas. If we see some interesting technology that needs stimulation to help us get into one of those new areas, that's when we'll make investments.
A classic example that we pointed out today in the server ecosystem was our investment in Calzata. We think that to really achieve low-power servers, yes, you've got to have low-power microprocessors. You also have to have low-power system design, design of the whole chip and the way the chip fits into the system. That's the clever piece of IP that Calzata has. That's the sort of thing that we invest in. I don't know if I've pretty much answered that, I think.
Pretty much. Yeah, I mean, if I guess the only thing that I would add, I think investment in innovation to enable low power across the whole spectrum, everything that we do, every market, that is important. That is a driver for what we do. Clearly, 3D graphics will be a great example of where we have invested and been increasing our investment over the last few years, an increasingly important area going forward. The one other area, sort of trend, that I would mention that is increasingly important is the area of security. Again, we have some great technology built into our latest Cortex A-class processors called TrustZone to help enable the deployment of secure services on the processor. Those are some of the areas that I would note.
Breakthrough, OK, if you look at mobile phone roadmap to 2020, 90% of the BOM of a mobile phone, I can, if you pay me enough, tell you what's going to happen. It's a fairly straight line. You can make reasonably good predictions. There are some fundamental challenges to be solved, but you can play it out. I'll be wrong on 10% of it, which will be the I/O device that doesn't exist today that will become must-have. I can't tell you which one that will be and whether it's around, but it'll be an I/O device. It'll be some form of interaction with the phone. People will crack the virtual keyboards, the gesture, mood recognition. It'll be some I/O device that'll be the breakthrough.
None of that will matter compared to the service offering that someone will dream up, like Facebook, that has little to do with the hardware but has been enabled by it. That's where the real transformations will take place.
On Mali, I think from a software point of view, certainly, I mean, I think something I mentioned where we've been through testing 300 of the top games and worked through that successfully. I think really going forward, the fact that we now have a market-leading device out there will really help because developers tend to gravitate to testing their applications on market-leading devices. I think we now have one of those out there.
From the underlying dynamics, given that the evolution of Mali's success, Mali is very similar to some of your previous families of processors, or does the growth and adoption of Mali possess any difference from other?
I mean, we think fundamentally, and somebody, I can't remember whose slide it was, we talked about 4 billion internet-connected screens. Behind each one of those internet-connected screens, we obviously expect to see at least one Arm microprocessor. Actually, we also expect to see a graphics processor. Do I expect to see a graphics processor in one of Mike Müller's neutrino detectors underneath the Antarctic? Probably not, I can't quite get my head around that. Clearly, we're going to have some sort of focus with Mali, and it'll be a sort of market-oriented, product-oriented focus. Therefore, the types of companies that we deal with to build the Mali ecosystem will be somewhat different from some of the other sectors. There's about a dozen sort of must-have companies with whom we need to work in Mali to get to 80% of the ecosystem that we need to develop.
We're gradually working through those dozen companies. Right off the top of my head at the moment, I can't remember all the names. We can talk over coffee, if you like, about that. We haven't got all 12 at the moment. I think we're on about seven or eight. I think the next one was Simon at the back. We've probably got time, looking at the clock, for about another two questions. Then we're moving next door to bite for lunch.
Thanks, it's Simon Segars from Goldman Sachs. Actually, I had a question about the business model in general. I was interested in your comment that you think roughly 60% or 70% of the cost gets saved by a customer by going with you. Of course, when market share was low, that was partly the beauty of what allowed you to capture as much share. For more than half your business, you've got probably 80%, if not 90%, of share. Customers may not even have a choice. They'll probably still go with you, even if they're only saving 30%. I guess my question is, why isn't that price that you're charging going up for a vast majority of your business?
Let me kick off with that. Maybe Tim might want to chime in. I think what we've tried to show you is that as the value in Arm Holdings technology increases, we expect to extract more of that value in terms of license fees and in terms of royalties. We touched with a graphic on the sort of incremental proportion that we get in license fee terms. I think back in February, at the results in February, we talked about incremental royalties. The answer is that it is a slow process. As Judah pointed out, over 40% of the volume that's shipping today is ARM7. Only 4% of the volume that's shipping today is Cortex-A processors.
When you look at the royalties that we earn off nearly 2 billion processors in a quarter, there's only 4% of that that is Cortex-A commanding a sort of higher premium than some of these effects are today invisible. Some of the other graphics also showed that we do expect this to flow through to much more of the volume in time. I think that's clear.
That's clear. The other thing that we obviously have to consider when we're thinking about the pricing of the business model is we have about a 28% market share, we reckon, of embedded processors. Yes, our penetration in mobile is about 90%. We're incredibly interested in upping that 28% in the other markets and gradually moving that towards the mobile level. The companies that are going to take us there are often in high volume, the companies that are generating our mobile royalties. Their appetite, and certainly their appetite for the speed of transfer from what they use today, often in-house, to Arm, would be greatly decelerated if we were perceived as an entity that exploited, if you like, our penetration. I think there's a balance between milking the position now and achieving pervasion at the pace that we think we're capable of achieving.
We actually think that the opportunity for pervasion at the moment outweighs what might be a relatively shorter-term opportunity of really turning the handle on mobile. It's that balance that we consider.
I think we had a question here at the front. We probably only got time for one more. I'm afraid it might have to be a bit of a lottery.
I had a question regarding the potential consolidation of the baseband business. Currently, there are many different competitors who are attacking that market, whether it's Qualcomm, Broadcom, Marvell, NVIDIA, MediaTek. If those, as that market sort of sorts out its winners and losers, what is the risk to your licensing business that those people licensing those highest-end cores would ultimately fall off? Is the increased subscription license contribution enough, or the new source of license revenue from the architecture enough to offset that? Have you run that scenario in case that does happen?
Yeah. I think from a, so looking at that scenario, typically, companies who are in the modem business will typically be using a Cortex-R profile core for that particular application. Those cores, of course, the same cores, as well as being used in modem, they're also used in automotive, in hard disk drive, a number of the embedded control applications that we've talked about today. I think from a licensing perspective, actually, however that might play out in the industry, it's not going to make an awful lot of difference. It's unlikely to move the dial. What's perhaps more interesting is the long-term royalty potential. It doesn't really matter which companies are shipping the Arm-based modems in volume. They'll still be obviously getting the royalties through. I'm sure I should finish.
I understand. Is there a risk that there's going to be a competitive tactic on the license side if there's a consolidation or it gives up the licensing? What's the best thing that you can do for it not to offset?
I think any effect, like a VAT, would be very small in the grand scheme of things, looking at the number of licenses we assign overall per quarter if that scenario came to pass. It's interesting, over the last, you know, we've been asked a variation on that question at more or less every analyst day for the last 10 years. Essentially, is our market saturated? During that time, consistently, whilst there's been very good arguments for why the market should consolidate, our licensing has consistently grown. We've continued to add between 60 and 90 licenses per annum, and the number of semiconductor companies who've been actively licensing Arm Holdings technology has grown from about 120- 250-odd companies. I agree, it's very hard when you look at an individual little piece of the market.
You can see some activity happening there, but the net effect on Arm Holdings' licensing business just doesn't seem to have a detrimental effect. I think we have one last question. Whoever gets the microphone first.
That was, by the way, the best way that question's been asked in the last 10 years.
I'm Dr. Arjun Nani from Nomura. You spoke about servers as well as mobile computing. At the moment, the dominant form factor for creation of data still seems to be a desk-based computing system. How do you see that evolving? Do you have plans to enter into those kinds of systems? Do you see devices like the ATRIX, where you have a mobile system which gets plugged into a monitor and converts into a desk-based system? Basically, a system dominantly used for data creation rather than consumption.
The answer is that we can't tell you what the dominant form factor is going to look like. As Mike said in answer to the question about the mobile space, it's going to be some I/O peripheral, and we can't tell you what it's going to be. Similarly, with content creation, the form factor of choice at the moment is something like a desktop machine because it has been for the last 25 years. Whether or not that will continue is something that we actually can't tell you. What we can tell you is that from an electronic system point of view, a content creation machine looks very like the sort of machines that Arm Holdings technology is going into today, and the form factor will be what it will be. It's a job for somebody else to come up with that form factor.
Hopefully, that form factor will include some other intelligence, so there'll be opportunity for even more Arm microprocessors in the dominant form factor of the future. From a point of view of our business, it really doesn't matter.
Thank you.
Thanks. I think we had, because we're running a little bit over the schedule here, we better finish the formal Q&A at that point. All of us will be next door there, as will be Ian. The other Ian and Drew, who's standing up at the back, will also be out there. We'll be happy to talk any further questions in a more informal environment. Thank you very much for coming.