My name is Shawn Kim from Morgan Stanley. I guess we were here two years ago in this exact room, and it is really good to see you in person and everyone attending. We have a great lineup today. The format will be 40 minutes presentation, and then we will take 10 minutes for questions after that. Without further ado, I would like to introduce, please join me in providing a very warm welcome to Ben Suh.
Good afternoon. As Shawn just introduced, I am Ben Suh. I am Head of Investor Relations at Samsung Electronics. First of all, I would like to thank all of you, including those participating online, for our first-ever live streaming, for joining this year's investor forum. Actually, many of you probably know the reason we do the investor forum is that we try to address topics that are of high investor interest, but also we try to pick some topics that we feel that we would like investors to know better of, so to enhance your understanding of Samsung Electronics. I think this year is especially meaningful, because as Shawn mentioned, we are together for the first time in three years. It has been a long period of a lot of Webex and Zoom calls.
Those are very efficient, I think at the end of the day, it is quite meaningful to meet face to face. We look forward to some very good discussions today. It is great to be back in Singapore to be meeting with all of you. Now these days, as you are probably well aware, there is a lot of concern in the markets about the slowdown in demand due to interest rates, global inflation, recession concerns, and even lockdowns in China. While we have all these challenges, and it certainly is not the best of times, I would also like to point out that in the past when we had such big global macro challenges, Samsung has always emerged much stronger.
Today, we would like to share with you some of the areas we are focusing on for continued growth in Samsung Electronics in the field of memory, foundry, and our DX businesses. First of all, we have as a presenter, EVP Jinman Han, who is the Head of Global Sales and Marketing in our memory business, and he will address the changing dynamics of the future memory industry. He will share his views of the longer-term market outlooks, as well as key differentiations to address new opportunities. Mr. Han graduated from Seoul National University with a degree in electrical engineering. Over the years, he has been a key contributor in the company to areas such as both DRAM and NAND development, as well as flash solutions and SSD development. Currently, he oversees the global sales, marketing, and business development of the memory business.
Our next presenter will be EVP Sang-pil Shim, and he is the Head of Corporate Planning in the Samsung Foundry business. He will present our Samsung Foundry's technology leadership and customer strategy as a trusted supplier. He is going to cover the key pillars of Samsung Foundry and what makes it competitive. Dr. Shim earned his PhD in electrical engineering at Korea Advanced Institute of Science and Technology, and he has played key roles in the company, including in the development of advanced logic processes and also as head of Samsung Austin Semiconductor. Today, he is leading the sales and marketing activities of the Samsung Foundry business. Last but not least, we have EVP Se rin Han from the DX Division, and she will provide insights on the services business.
The services businesses, which is based on the huge Samsung ecosystem of products, is still evolving, and Ms. Han will help you understand its full potential. Ms. Han received her master's degree in integrated marketing communications at Northwestern University, and since joining Samsung Electronics, she has been devoted to establishing the Smart TV ecosystem and also in developing numerous services, including Samsung TV Plus and Samsung Gaming Hub. She is currently the Deputy Head of the Content Services and Platform Business. I am confident that our speakers will deliver interesting perspectives on our strategies for the continued growth of Samsung Electronics. Please give a warm welcome to our first speaker, EVP Jinman Han of the Samsung Memory Business.
Ladies and gentlemen, I am Jinman Han. It is good to be here in this precious and very competitive market. It is a real reprieve for me, and it is really good to see you face to face. There is nothing better than facing each other. Today, I am going to talk about the dynamics of the Samsung Memory Business. I am not commenting on any market current situation. It is so hard to predict. Any prediction is as well as good as your assumptions. These days, making assumptions is so tough. Everything changes. Every time it changes. I am going to talk a little more about the future, immediate future, and the long-term view, how our business is evolving, and how Samsung is going to address this evolving market. Okay. One of my colleagues told me this is his favorite slide. I don't know why, but I have three agendas.
One is memory at the center of digital life. I know that every company is arguing that their product is the center of digital life. I really believe that memory really is the center of digital life. The second agenda I am going to talk about is how the future of the Samsung Memory Business is evolving. Finally, I will talk about how we can address those evolutions by Samsung's point of view. Okay. Okay, first, memory is at the center of digital life. Digital life of both human and machine, I would say. In the old times, in the 1980s, it was really about the PC market. PC was the king of the digital hill, if I will. After that, we have a great growth coming from the mobile space. 1.3 billion sets per year, which is a huge market.
Comparing 50 million sets of server is really huge. Every three years, people tend to change or replace their phones, which is amazingly a good market for us. These days, average memory content for a smartphone is about five gigabytes. 1.3 billion times five gigabytes, which is amazing. Huge. From 2020, we are starting to see that the mobile market is taken over by server markets. Server from hyperscalers as well as OEMs like HP and Dell, and Lenovo. When first hyperscaler idea was emerging, we had an internal discussion whether it is good for our TAM growth or addressable market growth. The general consensus amongst ourselves is that, well, people might not buy a server as before, because if you have a very efficient big data center, then why would you want to buy a server?
We thought that market might not be growing as fast as before, but it turns out to be spectacularly wrong, right? We were starting to see various workloads and applications. Small startups are using hyperscalers to start their own businesses, right? I just had a noodle ordering from Grab, right? That kind of business model might not be possible without the introduction of hyperscalers. Any new technologies, sometimes we may feel are a threat, but it always turns out to be okay for us. Right now, we are starting to see new markets emerging from AR, VR, and IoT. Even though there are some struggles among the companies who are trying to make this AR/VR business big, eventually those days will come. It will give us huge opportunity again, I truly believe so.
From our point of view, revenue point of view, right now, DRAM revenue overall in the world is about $100 billion. According to our internal estimation, by 2035, it's going to be almost three times as big as today. Any computation, you need memory. There's nothing you can replace that technology. A long time ago, when PC was introduced, it's really about CPU, central processing unit, and your memory, and sometimes hard disk, right? That architecture has been here for a long, long time. When CPU companies, providers, hit the wall of improving their performance, they decide to use more cores as possible. More cores, instead of boosting their frequency. It's still the fundamentally same architecture. You just have more cores, right? Each core needs its own memory content to make computation easy. It's about six or seven gigabytes per core. That's the rooftop.
Recently, we are starting to see accelerators. Some people call it XPU, accelerator-type special function CPU, like computing SoC. One of the examples is GPU. We believe that this evolution will be on the right side, that you have a single huge repository of memory pool, and then every CPU or GPU or data processing unit or any GPU will access those big memory pool, sharing all the information without having to have its own memory content, which will make it very efficient for them because they don't have to buy huge memories for each application. They can have one single pool of memory so that any CPU or GPU can access them, which is kind of like, to me, a hyperscaler idea, right? They're having this one big memory pool, and then you don't really have to buy it.
You just have one big pool of memory, then every CPU, GPU can access it. That might limit the growth of memory content, right? From the intuition. I beg to differ. This will create new opportunity even for us and for our new customers that I'm going to talk about a little bit later. Currently, the bill of material for memory is about 30%-40% for server. If you buy a server, a substantial amount of the bill is from memory, right? That's why our customer, hyperscale customer, OEM server customers, are so keen about the price projection of memory. They have been showing concerns about the slow pace of node migration, they are worrying about the slow pace of price reduction over the years, right? Because their memory BOM accounts for about 30% or 40%.
We believe it might reach as high as 60% within 5 years. In the long future, we might say if you buy memory, you get everything else free, the future might be coming. I don't think this will eventually be like 89%. No, way. The server companies have great engineers. They're trying to optimize their memory usage so that they can limit the content of memories, right? In there, we will see a lot of great ideas, great architectural changes. We have a PC era and mobile and server. What's the next? That's the big question from my boss, my CEO. Where do you think the next big demand will come from? One of the most probable candidates are automotive. Like I said, on the DRAM 6 gigabyte, that is for a smartphone, then 152 gigabyte for a smartphone, right?
A few companies in the automotive industry are starting to use about 50 times more. Right now, they are using about 10 times more of DRAM and NAND flash content than a smartphone. Not every automotive company, but a very few of them. You know who I'm talking about, right? We have 1.3 billion smartphones every year, and we have roughly 100 million cars sold every year. Imagine what's going to happen when those cars will have 10 times more DRAM in the future. We're going to see certainly the almost same market as smartphone industry coming along, right? That's a huge market. It may take some time, and everybody has their own answers when that era will come. We believe that those days will come.
We never expected the smartphone business to be so big when Apple introduced the first smartphone. Some people say it's storm in a teapot, it turns out to be a huge market. The same thing might happen, maybe not, this is one of the candidates we believe that is going to give us a huge market for our memories and storages. Around from, I think 2018. Yes, 2018. On the left side, you can see that the data created by machine surpassed that of data created by humans, right? The machine is producing more data than humans. It's going to be kind of a singularity, right? When we draw the line, correlate that with our DRAM and NAND demand, it's perfectly matching. We believe that that data creation by machine will get accelerated.
You will see more and more demand for DRAM and NAND flash, right? When you look at the CAGR from now to the next six or seven years, it's going to be 17% for DRAM and 24% for NAND flash. That's the rule of thumb number when we say, "Okay, what's going to be next year's demand for DRAM?" About 20%, NAND about 30%. That's because based on this assumption. We believe that even though we see some fluctuations in the near future, that trend might be maintained a long, long time.
Yes.
Some people say it's just the beginning. I'm talking about the growth potential of DRAM, especially DRAM. I'm not talking about NAND flash today, but DRAM potential for growth is huge. Any competition anybody introduce with their new architecture needs DRAM. When I look at the architectural research ongoing around these days amongst the companies, is that they're always trying to figure out how to access memory in a more power-efficient, and how can they increase the bandwidth as much as possible. It's almost all about how to work with memories. They know that it is a very critical component.
Right.
Especially when you're talking about AI or machine learning, it's all about how to access memory efficiently. People come up with all different kinds of architecture just to make sure that they have the best memory management algorithm in their chip. It's going to be huge. Next, I'm going to talk about how it is going to evolve over the years. We know that memory is really the center of computing, both for human and machine life, but we want to see how it's going to evolve over the years. Memory business has not been so attractive to most of the investors, because it is so cyclical and market is so capricious, and during the shortage, we tend to invest a lot of money to build more fabs. During the oversupply, we tend to stop investing. Which makes all these typical cycles that make everybody frustrated.
One of my sales VP complained to me that he really liked oversupply situation because during the shortage, it's so tough for him. Every day, customers calling him, visit us, asking for more bits, which is so challenging and frustrating to him. During this oversupply situation, I think some people prefer to have a shortage cycle, but usually shortage is really painful time for suppliers like us. Anyway, that CapEx volatility and oversupply situation has been not so attractive to most of the investors. We know that. The second characteristic of memory business is that people believe that memory is a commodity, like copper or rice or oil. Right? What is a commodity? Well, it is easily producible and easily replaceable by other suppliers, and it has very high price elasticity. It doesn't require huge technological improvement. You just do what you have been doing, right?
The market is dependent on the supply and demand. People tend to believe that memory is like that, right? Easy to produce, anybody can make it, right? The third thing about memory business perception is it has very low profitability, and the operating cash flow is not even. It's not predictable at all, up and down. You always know that next downturn, next bust cycle will coming. Right? How can we make sure that our stock price can go up? Because you never know what's going to happen next year. That's the perception. Some of them, I agree, we still have that, but it's starting to change. That's the point I'm trying to make. Let's talk about the CapEx volatility and oversupply. In 1990s, we have more than 20, I think it's more than 20 DRAM suppliers.
Early 2000, number of players got shrunk and 2010 is also shrunk again. Today, we have three major suppliers, which accounts for more than 94% of total DRAM market. I didn't include any company who has less than 5% market share. I think it's fair to say that we have three major suppliers. This is, I think, very nuanced than it appears at the first glance. It has huge implications in the DRAM market. Samsung is now accounting for about 43% in terms of revenue, not bits. What's the effect of this so-called consolidation? Well, the CapEx- Divided by revenue is maintained around 30% for the last 10 years. Look at the 67% number. That was 2007. As an average, CapEx was 67% versus your revenue. It was amazing, very competitive. Right now, we are not doing that anymore.
We are very disciplined, we know how to generate cash flow. We don't want to over-invest anymore. I think that is the general consensus that we have learned over the years, which is good for our customers as well as to us. This graph shows oversupply, undersupply situations each year. Anything below the middle line is oversupply and anything above that middle line is undersupply. For the last 10 years, we realized that it is about 0.7% undersupplied. Well-balanced, you can say. The three major players of DRAM industry make our investment very, very reasonable and at the same time, market is pretty much well-balanced. Although with some up and downs sometime, but it's quickly recovered and we are doing that. That, I don't think is the typical characteristics of commodity market.
We might be able to see more stable supply-demand situation for the years to come. I truly believe that. Memory as a commodity. Like I said before, PC used to be the biggest market segment and all the memory suppliers are trying to give a one-size-fits-all solution to PC makers. The only difference is package type, some speed differences, that kind of thing. PC really was the king of the hill at the time. There is no real way for us to differentiate from each other. It's just all about supply ability and cost, price, a little bit different performances. Look at the market these days. Our DRAM addresses all these markets. Data centers, which require very high reliable, very consistent supply, no surprises at all. Also, it requires very thorough serviceability.
When something happens during the field, we have to address the problem right away by sending our technicians and engineers. In mobile, it's all about low power. Mobile, the cycle of DRAM product is about one year. Every year, we give them new products based on the latest and greatest technologies. It also requires very small form factor because your phone is getting thinner and thinner. They always keep asking, "How can you make your package height as thin as possible?" Gaming. Gaming is booming. Actually, gaming players, I think their demand is not as diminished as the rest of the industry. They're really good. I think they're trying to build more and more stuff next year because that component shortages is getting better every day. The gaming industries want us to provide them with very high bandwidth, very high-performance components.
Like a graphics DDR and that kind of thing. Automotive. They want to operate their car at a very high temperature. It's so hot, like 120 degrees, something like that. They want us to give them very reliable, resilient component. This is totally different from conventional old-time PC market. Every segment requires its own solution. There is a customer who even said that, "Okay, I don't really mind having single source supplier," as long as we give them good solution. That mindset has not been here before. They always want to have multi-source suppliers, but in this area, like gaming and data center, they sometimes want us to give them single source application, especially in high-performance computing. Even automotive, they want the best-in-class component, whether other suppliers have the same technology or not.
I don't think that is not the characteristic of a commodity anymore. Technology itself is getting difficult. For example, look at the left side. It's a number of steps in DRAM manufacturing. When we were producing 2X nanometer technology, assuming that as a one, we need 50% more steps to produce the die. Right? I don't think oil refinery industry is trying to increase the steps to produce the oil, right? We don't do that. I think they're using the same technologies, right? Us, on the right side, you can see that NAND is much worse in this sense. It needs more time for us to process the chip, right? More steps, the more process mean you have to have more wafer fab equipments in your fab. That means you need bigger space whenever we change, whenever we migrate to the next node. Right?
It's getting difficult as time goes on. Why are we doing this? Because we want to provide the benefit of cost reduction to our customers, right? Also at the same time, we want to increase the bit, number of bits, so that everybody can consume more bits without having to worry about BOM increase. Although, even though we are doing that, the BOM of server is getting bigger. We're trying our best, but it's very tough. The design will shrink, has its diminishing returns as time goes on. A while ago, I remember that we used sometimes have a 70% bit growth in DRAM, right? More than 100% bit growth for DRAM. It is not happening anymore. Also there is the revolution going on in packaging technologies, right?
I don't know who have some background in analog design, they don't want to use the most advanced technical node, technology node, right? Even though that doesn't make Mr. Kim happy. For analog or IO application, you want to have certain design node to make it very effective and power efficient, and reliable, right? CPU wants to have a state-of-the-art technologies, but some IO chips, you don't need that technologies. Memory has its own pace of node migration. How are you going to do it, right? People come to think of "Hey, why don't we mix and match all those small chips together so that we can put it inside one single giant package and connect all of them together using the optimal processes for each component," right? That is happening right now, and one of the jargon for that technology is chiplet.
We want to have a various number of chiplets, and we want to connect and stitch them together and put inside in a one single big giant chip and then make it to the market. That's a very good, very subtle implication for us. It might give us a new chance of developing business model, new business models, right? As a DRAM provider, we might be able to provide our DRAM chiplets to our customers, fully customed DRAM chiplets to our customers. Or we can ask our customers to give them their chiplets and we're going to stitch them with our memory, and we're going to develop package technology for them. It's kind of a DRAM foundry-like model with package included. It gives us a very nice chance of developing new business model, and it's going to be very interesting to see how that technology goes.
Low profitability, DRAM and cash flow is up and down, hard to predict, right? The reason our cash flow is not as great as other software or pure foundry players is because we have to maintain lots and lots of fabs. We also have to invest a lot of money, CapEx, to build new fabs, right? As time goes on, Samsung is, I think we are the one who really starting to see the benefit of depreciating fabs, right? We have a lot of fabs. A lot of them is also getting depreciated quickly, right? When your fab is fully depreciated, whatever you do with your fab is just profits, right? At the same time, we have to build more fabs, right? Maybe in U.S. or in Korea.
When you look this from a little bit distant point of view, then the assets that we have might be a huge technological barrier for any new startups can challenge in the future, because it may take enormous amount of capital to replicate that kinds of depreciated fabs in your hand. Some people might say, "Hey, you know what? I've been hearing about these MRAM, FeRAM, or RRAM, or PRAM, Optane technologies. Those might take you guys over, catch up with you." You know what? Those memories, they are called emerging memories. I've been watching that emerging memory emerge, starting to emerge for the last 30 years. I don't think it's going to be emerged at all. As an engineer, I was in charge of developing PRAM myself. At the time, PRAM was perfect. It sounds perfect.
You can have a 10-year retention time. You don't have to refresh your cell every 30 milliseconds. No way. You can retain your data for 10 years, more than 10 years. The problem is, you may not be able to write and re-erase, or write and erase that more than 10 million times. Imagine how many read and write cycle is happening inside a server with your DRAM. DRAM wise, from DRAM's point of view, the endurance, we call the endurance, is almost limitless. Right? There's no way you can catch up with DRAM's inherent superiority of reliable operation, read and write operation. When we chart that on this nice octagonal comparison table, you see that DRAM has the most balanced characteristics. Perfect for your computing memory. I don't think you can count big on any emerging memory.
It may have its own niche market, but it's going to be very far-fetched to say that any emerging memory can replace DRAM industry for a long, long time. Yes. Look at the memory operating margin. This is not Samsung's number, it's the average operating margin. It is about more than 35%, right? Sometimes over 60%. Remember that the CapEx versus revenue ratio, about 30-something percent? The difference, gap is operating profit. For pure DRAM industry, for pure DRAM players, operating margin is pretty good, actually. Very good. It will be this count for a long, long time. The green line in the middle is actually Philadelphia Semiconductor Index, companies' average operating margin. Oh, by the way, there are no pure DRAM players in this market. The three major DRAM supplier are also having NAND industry, right? That might hurt the operating margin. That's the reason.
For DRAM itself, the operating margin is pretty good without it. For the last 10 years, there was no case when DRAM profitability is below SOX average profit margin. What is Samsung going to do for these kinds of DRAM markets? Are we going to chasing a ball or are we going to running to where ball is going? When people have this idea of having huge memory pool that can be accessed by any hosting CPU or GPU or DPU, we have to come up with concrete engineering solution, which is called CXL DRAM. Compute Express Link. It is first initiated by Intel, and Samsung is one of the board members for that consortium. That allow us to make our DRAM to talk to CPU or different kinds of accelerators with a new interface definition. Right?
Any host, SOST, can access our DRAM as long as they adhere to this standard called CXL. What's so different from conventional DIMM, right? Behind CXL interface, you can put DRAM. Our host is only seeing CXL interface. They don't really care about what's happening behind that CXL interface. That gives us new freedom. Instead of CPU companies define and memory companies get together and then define this spec, if you want to have this DDR5, you have to buy this CPU. Those era might be gone in the future. You are basically disaggregate your memory technology from the CPU and host. As long as you are adhering to the CXL interface standard, you can put any DRAM technology you want.
You can have a very cost-effective, but prone to errors, but you can put error correction capability behind this CXL, then it's okay. You can put very high speed DRAM attached to that CXL. It doesn't matter, as long as you comply with that definition. It's going to give us a huge product in terms of engineering. We are finally divorcing from the CPU. It's not really divorcing. We can have our own freedom of technology development, decoupled from the CPU technological progress. That will give us huge chance to support this pool of memory that can be accessed from all different kinds of hosts. Another area we are thinking about is the packaging technologies, 3D IC DRAM technologies. Using chiplets and three-dimensional stacking memory on top of the CPU, that gives us another new chance of a business model that we can provide to our customers.
The third thing we are talking about right now is, maybe with CXL technology, we can have our own as-a-service to our customers. Instead of just selling upfront and then pay back the money, get the revenue, done. No, we might be able to work with our customers so that as the usage changes, we can charge them as usage varies, like hyperscaler. It might be really tough because it has to be win-win scenario, win-win business model, so that make everybody happy. We're trying to crack the code using these CXL technologies. This CXL technology, this is our demo and a version of CXL technology. It looks like kind of a SSD. It has a bunch of DRAMs, but in front of that DRAM, you can see that the small chip, that's the CXL controller.
You can put a lot of interesting ideas inside the chip. Right? Depending on the feature of that chip, we can create new business model, as well as putting all different kinds of DRAM behind that chip. Because host is talking with that chip only, not talking directly to DRAM as before. We are one of the board members, only board member from DRAM industry. Samsung is the only board member of DRAM industry right now. Okay. The last thing is 3D DRAM technologies. On the left side, you can see the 2.5D technologies. 2.5D means you put your logic and DRAM at the same level, but your DRAM has a three-dimensional set technologies. If you put that three-dimensional DRAM on top of logic, you can start to call that three-dimensional. Putting DRAM on top of CPU is not just mechanical variation.
That means that your CPU can talk to our DRAM within a very short distance, which will give us huge bandwidth benefits with the low power consumption. If Samsung can provide that DRAM IP as well as how to packaging them, it might be giving us new business model. We can be a player as a DRAM foundry with full stop technical support. That's one of our vision in the future. This is as-a-service. Our goal is that within, I think it's five years, we want to have It's a very challenging goal, but we want to have as much as 10% of our revenue coming from this as-a-service. It's very tough, we want to crack the code, actually, really. This is the end of my presentation. Recapping, I talked about how memory really is the center of this physical era.
You need DRAM as an essential computational chip. Also at the same time, I talk about the characteristics of DRAM business has been changing, and is going to evolve more like a non-commodity, very specific, full custom, with the as-a-service and architectural change, all those kinds of things, which is not typical characteristics of a commodity market. Also I said that the operating margin of DRAM industry is very healthy compared to other Philadelphia Semiconductor Index. Finally, Samsung is working on 3 major directions. One is that we want to provide technology for a pool of memory with the CXL technology. In addition to that, we're trying to provide memory service model, including packaging solutions with our DRAM IPs on this 3D IC era.
In addition to that, finally, we're trying to build a new business model called as-a-service, with the benefit of this CXL and this packaging services. Okay? That's the end of my talk. I think the floor is open for your questions.
If you have any questions, you can raise your hand. Somebody will bring the microphone over to you.
Okay. Nick?
Thank you very much for the presentation. Nick from UBS. CXL DRAM. There's a lot of questions as to whether this is actually accretive to overall DRAM market, DRAM consumption, or because effectively making DRAM usage more efficient.
That could actually be dilutive. What would be your perspective on this, please? Thank you.
The major business will be taking place from 2025. That's what is expected to see, right? A long time ago, we were so fascinated by this embedded DRAM market. When DRAMs got cheaper, people lost their interest because embedding DRAM is much more expensive than buying a quality DRAM from the market. This CXL, some people ask, a CXL might be just the kind of passing technology that might lose its luster over the time, but we don't think so. A major pain point of our customers in DRAM industry, especially from server, is that our bit growth is not as high as before, because it's going to be very challenging for us to transition to next node because of huge CapEx and the diminishing return of design win shrink.
People want to use DRAM as effective as possible, right? If you have your own DRAM for your CPU, have your own DRAM for your GPU, own DRAM for DPU, you might end up having lots of DRAMs, right? People come to think about why don't we use this DRAM as a pool, and then everybody can access it. The CXL is part of that essential technology to implement that idea, that vision, right? I believe that that will unleash the full potential of next engineering revolution in the future from the architecture point of view. It's the real transition from von Neumann Architecture to the non-von Neumann Architecture, memory-centric.
Memory centric doesn't mean that you have to buy memory centric means that your whole architecture is well optimized so that you can have easy access without incurring huge costs, and you can share all those information at the same level because you don't have to worry about caching contents between caches. You don't have to worry about that. It has to be maintained inside. You don't have to have a duplicate copy of same data anymore, right? This is very essential. We believe that the market is real, all the major suppliers are developing prototypes for that technologies. We believe the real business will take off from 2025.
Yes. Shawn here from Morgan Stanley. Just a question on this era of de-globalization. We have a trend towards onshoring back from China. Maybe two questions. One is that in the short term, it may bring some excess type of a cost, so it may not be as great. Could you expand on that, the impact of what happens on cost if you reshore back to Korea? Then long term, is that going to be more sustainable, and therefore, is that something that's better for the long term?
Right now, we don't feel any impact of de-globalization, right? There's a concern, but it has not been realized yet. At the same time, if that really happens, but when everybody is on the same situation, when we are playing at the same level, it should be okay, right? Everybody will be affected by that de-globalization. Our customers might have to pay premium when that situation happens, but I don't think they disagree with that premium, right?
Actually, the question was more on if you have to no longer manufacture from China and have to come back to Korea, how does that impact the overall cost?
Having fab outside Korea or China is very expensive in terms of operation. In addition to building materials, labors, right? Everything is expensive. It will cause the price reduction not as fast as before, that's for sure. Right? For China fab, I believe we can sell all the bits coming from China for the Chinese market. We're not sure what's going to happen, if we can sell the disks outside China. One thing is for sure, that we can sell our disks to Chinese market.
Okay.
Hi. Thank you very much for this opportunity today. Just wanted to ask another question related to CXL. It sounds that we agree with what you said, that the memory-centric architecture, the emergence of that could help our industry a lot more-
Yeah
De-commoditize our industry.
At the same time, in order for that to happen, the CPU players also have to work with the memory players.
Exactly.
Once the memory-centric architecture becomes mainstream product, in a sense that the, I guess, the hegemony will switch from the CPU guys to memory, in a sense. The CPU has to let go of some sort of power to memory players. In that sense, are you seeing the CPU players becoming cooperative with development of this? Or are they still a bit more cautious in terms of developing this technology?
Well, actually, the initiative was from one big CPU company. The main reason behind that decision was that that's the requirements from their customers. We have both the same customers, right? It doesn't matter who's going to have a hegemony or not. It's really about how to address the customer requirements, and that's the voice from our customers. Give us the pool of memory technologies, CPU and memory companies has to work together, right? Just give us storage, right? We cannot go against what our customers want from us. In that sense, I truly believe we are on the same boat. That's the reason we have a board of directors from CPU and GPU and memories from all different areas of semiconductor industry. It's not really about hegemony, it's all about giving total solution to our common customers.
They are very active working with us, and we're also very active in working with them. You get to blink.
Hi, just a question on Memory-as-a-Service, which seems like quite an interesting business model.
It's still pretty much in its infancy. What are the key barriers that you see? Is it distribution? Is it competing with your own customers? Because it seems like a very big change.
Yes
of a business model from just selling memory into selling, I guess, a service like Azure.
That's right.
like AWS.
That's right. The easier implementation is coming from NAND flash.
Yes.
Yes. You can sell your storage not as a fixed density. You can ask your customers to give us usage level, then we can charge based on how much you use our storage. DRAM is really tough, actually, because DRAM is very hard. How many DRAM you use, there's no way, right? They continue to use DRAM 24 hours a day, right? There might be some possibility that we can give more value to our customers by signing with us. For example, signing with us a certain kind of a technological contract, like software companies are doing. If something happens, you have to address the problem within few hours of the problem. Or we can give them. There's a lot of different kinds of ideas flying around right now.
Like I said, there's no concrete conclusion yet, but for NAND flash is for certain is pretty straightforward. DRAM really have to think about the more value that we can give them through that CXL controllers.
Maybe just a follow-up on that. In terms of NAND flash-
Yeah
do you envisage that you would gain sufficient scale to be able to compete with the likes of AWS, Azure? Obviously, cloud is a scale business, right? At the end of the day. There are a lot of synergies by having CPU and then storage at the same time. Just really trying to understand and dig a little bit deeper into that.
As-a-service, it's not compete with our hyperscale customers. It's kind of a memory subsystem.
Right? They have their own services. They may have to use storage boxes. The storage box, they're consuming lots of money, their CapEx is so huge. We just want them to want to convert that CapEx of buying storage or DRAM into OpEx model. Right. Yeah.
Okay.
It's part of their purchase plan.
Okay.
Yes.
Yes, very clear.
Okay. Well, thank you. Thanks very much.
Good afternoon, everyone. I'm truly honored to be here to introduce you Samsung Foundry. I'll talk about the core values of Samsung Foundry, especially technological leadership and business strategy as a trusted supplier. As Ben mentioned, I took most of my career at development and some for manufacturing, and this photo was taken when I was in Austin. Two years ago, I moved to business side, sales and marketing. At the time, my colleagues told me that, "Welcome to dark side of world." I never knew he really meant it. Last two years, it was like a roller coaster. Very, very difficult time. I hope we are passing through really a bottom line of our economy. As a start, I'd like talk about the four core values of Foundry. They are innovative technology, and diverse offerings, and tailored service, and reliable capacity.
In terms of technology, scaling of leading edge continues. Stacking technology for heterogeneous integration will evolve even further as chip functionalities become diversified. As for the economy, high tech industries that contribute to the improvement of human life, such as 5G, AI, and HPC, will continue its growth with diversification and expansion of applications. Diversity of Foundry service is becoming more important. Today's Foundry customers are facing fierce competition. They must overcome the challenges through optimization and differentiation of their hardware and software. This leads to an increasing demand for a more tailored chip design service. Last but not least, the geopolitical situation has been very challenging. Localization of the semiconductor value chain has become a keyword for many governments. Naturally, reliable capacity provision has become more crucial. Samsung Foundry has long been preparing to realize these four values.
Let me introduce some of our effort and achievement. First, innovative technology. Samsung Foundry has been leading technology innovations for many generations. For example, Foundry first high-k metal gate was introduced for 28 nm, then Foundry first 3D FinFET transistor was introduced at 14 nm. As you may know, we introduced the first EUV technology with our 7 nm. This year, actually, proudly, we introduced Gate-All-Around technology, that has been started mass production from June of this year. As part of this advanced process, we offer many platform technologies from mobile to HPC, AI, and automotive. We are also adding value-added specialty technologies to our mature processes to realize products like RF, MCU, display driver, image sensors, and so on. This process is applied to many product families and is already widely used in our everyday life.
Our scaling continues and aims at extreme of technology. In the FinFET structure, we are offering 7- 4 nm with EUV patterning, and 3 to 1.4 nm are provided with Gate-All-Around MBCFET technology. The first introduction of a 7 nm of EUV process was an important milestone for Samsung Foundry. Since then, we have continued scaling by supporting smaller and smaller cells, moving 5 nm- 4 nm. For each node, we are always evolving technology in order to provide optimal fit for mobile and then HPC and automotive applications. We do this by developing and delivering more performance improvement for HPC and high reliability for automotive process. For example, the 4 nm, we have many derivative nodes in the parentheses. This is our official node name, which was announced at the Samsung Foundry Forum last month.
For example, SF4E means Samsung Foundry 4 nm earlier version, and then SF4. SF4P means more power boosted version, SF4X means HPC version with a high voltage capability and more reliability. SF4A means the automotive version with very consolidated reliability. Devices are becoming more and more scaled, so this makes it difficult to improve performance power area at 3 nm and beyond. We have established a new transistor structure, Gate-All-Around MBCFET, and this is a significant step forward. MBCFET is a highly innovative transistor structure, and it helps us have better performance power area compared to the existing FinFET structure. Samsung Foundry will continue to innovate in Gate-All-Around and will develop a technology from 3 nm- 2 nm and 1.4 nm. Now let's talk more about Gate-All-Around technology.
As you can see at left-hand side, Gate-All-Around has superior characteristics inherently because of a structural difference compared to FinFET structure. The SF3E, first version of a 3 nm Gate-All-Around process, started mass production for the first time this year, as I mentioned, and it provides a significant improvement in performance power area compared to first generation of a 5 nm. Gate-All-Around provides another advantage to designers, as you can see on the right-hand side. The FinFET can only provide the effective width proportional to the number of fins. Gate-All-Around can provide various effective widths by patterning the size of nanosheets, which provides flexibility and optimized performance for chip designers. Let's look at a short video, the introduction of Gate-All-Around MBCFET.
Whenever semiconductor technology faced challenges, there has been a guiding light to help navigate the maze. Evolution of transistors is a major component of technology development. Planar transistors have been used for generations with voltage scaling to save power. However, short channel effect ultimately limited usability. FinFETs were the solution and made further voltage scaling possible. Unfortunately, limitations are faced again. For ideal electrostatics, the channel needs to be fully surrounded by the gate. This is known as Gate-All-Around. Typical GAAs are nanowires. Integration complexities of nanowires outweigh the benefits. This led Samsung Foundry to create a unique version of GAA with all the advantages, but minimal complications. Samsung Foundry proudly introduces our Multi-Bridge Channel FETs. MBC FETs consist of multi-stacked nanosheets. One advantage of MBC FETs is that additional area is not required to improve speed.
FinFETs need fins to be laterally added, while nanosheets can be vertically stacked. MBC FETs are compatible with FinFET design. Designers can replace FinFET with MBC FET without changing the footprint. Performance can also be improved without area increase. MBC FETs are the most advanced technology that provides solutions from low power to high performance applications, including AI. Autonomous driving, 5G, and high-performance computing. With MBCFET, Samsung Foundry will be the guiding light for the future of the semiconductor industry. Samsung Foundry.
As a next step, we are preparing our 2 nm, SF2 and SF2P, as the enhanced version of 3 nm Gate-All-Around. We have a SF that boosts performance power area compared to SF3 process. We accomplished this through capacitance reduction and more performance boosting and additional load track cell. This SF2 process is in line for mass production from 2025. We are preparing SF2P process. SF2P further improves performance area power to SF2 by plugging in backside PDN. We will mass produce this in 2026. Based on 2 nm process, we are developing 1.4 nm process as the advanced version of MBCFET. By using 4 stack nanosheet and advanced backside PDN process and lateral scaling, area will be reduced by more than 30% compared to SF2 process.
This 1.4 nm process will see performance power area improvement at full scale compared to 2 nm process following the existing trend of node migration. Its mass production will be smoothly migrated through backside PDN-based phase SF2P process learning curve. We are preparing for the post 1.4 nm technologies. We are working on a new transistor structure that can sustain scaling through vertical stack structure and 2D material for channel engineering in a scaled device, as you can see here. Novel materials to improve metal line resistance is also now under development. This slide shows the specialty technologies Samsung Foundry offers. For specialty technologies on mature nodes, Samsung Foundry's focal point has moved from planar to FinFET process. By now, we have a great deal of FinFET process experience. That is, we have done mass production with the FinFET process more than nine years.
We believe that this record of mass production is the best proof of our capability and potential. Based on this experience and trend record, we are continuously improving the performance and cost of mature node and adding special technologies like CIS, RF, DDI, and eMRAM, shown here. Especially our 8 nm has a very unique value proposition. The other foundry company was not able to provide 10 and 8 nm except Samsung Foundry. As you may know, 8 nm is the most advanced non-EUV-based cost-effective node. We have a high competitive power of RF and eMRAM. 14 RF is now under mass production, and 8 nm is about to start mass production in 2023. 14 eMRAM is being developed for automotive and industrial application and can be easily migrated to 8 nm eMRAM.
Let me explain eMRAM and RF a little more detail on my next slide. First, eMRAM. Our eMRAM has unique advantages. First, Samsung Foundry started 28 eMRAM manufacturing for the first time from March 2019. It has been already three and a half years. The other is 28 eMRAM bitcell size was the smallest. From the customer point of view, there are several benefits using our 14 eMRAM process. For example, it is easy to later transfer to 8 nm eMRAM, and the other benefit is we can easily scale down the eMRAM bitcell because MTJ module can be implemented into the baseline process simply by plug and play. Regarding RF, recently, we achieved a meaningful result to our RF baseline. That is, we will be delivering the world's first RF product under 10 nm to a customer by next year.
This would be the second worldwide first since our 14 nm RF. As you may know, node scaling inherently degrade RF performance. Our 8 nm RF is a breakthrough technology with effort to overcome the paradox of node scaling for RF performance. 8 nm RF is a great example of how, with innovative methods, we can overcome the negative impacts of device scaling for RF devices. Our RF process is very competitive, and it will become the market leader. Now let's take a look at advanced packaging, so-called beyond Moore. We have been working hard developing heterogeneous integration technologies, and this is the history and the future of our progress. Millions of HBMs with 3D stacking have been shipped already by Samsung Memory Business, and we are expanding this proven technology to logic 3D IC for both mobile and HPC applications. This is called H-Cube.
We plan to start mass production of this H-Cube technology with a micro bump type in 2024, and bump-less type, called hybrid copper bonding, in 2026. As a horizontal packaging technology, Samsung is developing 2.5D I-Cube that integrates logic type and multiple HBM on interposer. I-CubeS, which packaged one logic die and two HBMs on a silicon interposer, began mass production in 2020. I-CubeS, which integrates four HBMs, has also been developed and ready for mass production. On top of that, we are developing I-CubeE, shown here, which has an embedded silicon bridge die in RDL interposer. In 2024, we are going to manufacture an even larger I-Cube product with eight HBMs and two logic dies. Now let's talk about the next value, diverse offerings. Foundry applications are expanding to various product lines such as HPC and automotive.
In order to meet these diverse needs, Samsung Foundry provides special technologies optimized for each platform. More specifically, this includes HPC platforms with advanced 2.5D and 3D packages, automotive platforms with advanced RF technologies, and IoT and 5G platforms with eMRAM, RF, and Chimney technologies. Through this, Samsung Foundry will actively support customers to create their business opportunities and unique value. Currently, Samsung Foundry is already growing in non-mobile fields such as HPC and automotive and 5G. We plan to expand share of non-mobile sales to more than 50%, aggressively targeting high performance and low power market by 2027. Let's take a closer look at the HPC and automotive market. HPC is expected to see higher growth than any other application, and will become a key driver for the foundry industry. In particular, CPUs, GPUs, and AI accelerators are expected to grow by roughly 20%.
Samsung has advanced process technologies used by industry-leading customers for each of these categories. Moving forward, we will use our experience on high-performance IPs and design methodology for big die to strengthen the ecosystem and infrastructure. In the automotive sector, we are using our expertise in mobile application process to extend our infrastructure from IVI to ADAS. Growth in this sector is focused on SoC, ADAS, MCU, infotainment, and xEV. In line with that growth, we have secured everything from processes to fab and design infrastructure for each auto grade. As you can see on the right-hand side, we have a very strong history of the automotive business with the first mass production of some of our products dating back to 2018. Our record of zero field failures over the last four years in the 14 nm automotive process demonstrate how prepared we are.
In order to maintain our competitive edge, we are also working on a process that utilize eMRAM from 14 nm. Many Tier 1 customers of automotive MCU area are working with us for this 14 of the eMRAM. Okay. For the third value, tailored service, I'd like talk about Samsung Foundry's differentiation, synergy, and one-stop-shop solution. Memory capacity, latency, and power efficiency are becoming very critical for HPC products. Samsung Foundry is collaborating with Samsung Memory Business to develop highly integrated system and packaging for the next generation of HPC products. For example, implementing direct 3D stacked DRAM on top of a logic die and moving buffer die from DRAM chip to logic chip can enable high capacity and low power consumption. To realize such an idea, we are designing a customer tester chip with memory logic stacking capability called SAINT 2.
First, we designed a customized logic die that includes the functions of DRAM buffer and validates the functionality of the HBM core die stacked on top as phase I. Then validate customized last level cache DRAM on top as phase II. We are expecting that the phase I product will consume 40% less power, and phase II product will consume 60% less power with two times faster latency compared to conventional 2.5D solutions. No other foundry can offer this kind of product as effectively. We believe these new technologies and solutions will open a new era of high performance computing. Samsung Foundry has been working relentlessly to create a strong and sustainable ecosystem.
Thanks to ecosystem called SAFE, Samsung Advanced Foundry Ecosystem, and close collaboration with our best partners from DSP, IP, EDA, OSAT, and PCB partners, we are now capable of offering total solutions for the advanced and special technologies. SAFE has been growing year by year with new partners joining our open ecosystem. We are currently engaged with 22 EDA partners, nine DSP partners, 10 OSAT partners, nine cloud partners, plus 56 IP partners who provide more than 4,000 IP titles. Let's talk about the last value, the reliable capacity. Capability to produce advanced technologies are being expanded rapidly, and we plan to continue this trend in the future. The expansion of the new Pyeongtaek Fab in 2023 is expected to increase advanced capabilities by 1.6 times. A new fab in Taylor is on schedule to launch in 2024.
By 2027, we are expected to expand our advanced technology capabilities by 3.3 times. In order to maintain a stable yield with rapid expansion, our new product development will take place at the Hwaseong S3 R&D line and safely transfer to Pyeongtaek and Taylor Fab. In terms of fab sites in South Korea, Samsung Foundry currently operates three sites near Seoul. Giheung and Hwaseong fabs are roughly five minutes away from each other, and Pyeongtaek is 30 minutes away from the two. It makes it easy to transport wafers and other important components between the fabs. Our fab in Giheung started mass production in 1983, mainly producing mature products. The production of new mature or 8 nm are underway at this site. Hwaseong began mass production in 2000, and that's where you can find the S3 line, the EUV R&D fab.
Advanced products are developed in S3 and then transferred to Pyeongtaek or to the U.S. S3 line has a long history of experience making it highly effective, efficient at transferring products safely. With this experience, the S5 line in Pyeongtaek has succeeded in transferring 5 nm process with the shortest time. Pyeongtaek is planning to continue to expand its new fab, including phase II of S5 line, and which is scheduled to start mass production in 2023. In the United States, we have been operating the Austin fab for more than 26 years, since 1996. We are now building the third fab in Taylor, which is located about half an hour from Austin and which is scheduled to start mass production in 2024. We successfully transferred our 14 nm process with a stable D0 level from our S1 line in Korea to our S2 line in Austin.
Therefore, we have the capability to execute stable production at the Taylor fab through this experience and the knowhow of S2. Let's switch gear to the business side. Step by step, we have been adapting technology for each application sector and made a constant investment in infrastructure to support such technologies. We continue to make great strides in building our capacity through massive investment and are constructing new fabs in both Korea and the U.S. every year. We have secured the IPs required to support our technologies, and we will continue to work with full dedication. At the end of the day, the culmination of all this effort will ensure that we will be thoroughly prepared within the foundry industry. On the customer side, our global customers has more than doubled since 2019, and we are focusing our business worldwide to reach more than five times that number by 2027.
Recently, we secured many first-tier customers in the HPC, network, and automotive segments. As a result, we now lead several sectors in the world's 5G and automotive market. We have the top market share in the global 5G RF sector, and are a strong second in the 5G modem. We are also competing for the biggest market share in the ADAS sector by cooperating with top automotive customers. We will continue to spread our sales coverage across all applications by expanding our global client in the U.S., Asia, and Europe. Samsung Foundry is well prepared for each application, actively making necessary investment in a timely manner to support the success for our customers. By 2027, we expect to grow three times. By expanding from the mobile centric to HPC and automotive areas, we are building a healthy and diverse portfolio.
Samsung Foundry pursues realizing of customers' every innovation with various partners who build a foundry ecosystem. Going forward, Samsung Foundry will become a most reliable company focusing on the four values: innovative technology and diverse offering, tailored service, and reliable capacity. Thank you.
We'll take 10 minutes for questions. As the audience starts to think about next question, maybe I can start with my first question. There is this clash of chips that will happen sometimes in the next two to five years. In your opinion, how do you see this three-way competition evolving between yourself, TSMC, and Intel, over the midterm?
When I meet many customers these days, they think the current geopolitical risk is really serious. Many customers want to have a second source. From business point of view, Samsung Foundry are lots of opportunity to have many customers who want to have a second source. Industry-wide, many EDA vendors are preparing for more easier and efficient migration from foundry to foundry. Samsung Foundry was a little behind to compare to TSMC for 4 nm and 5 nm. I believe the 3 nm will be really important game changer. Samsung, Intel, and TSMC are all working on Gate-All-Around. Samsung is the first company, introduced earlier the Gate-All-Around this year, and the main production for mobile from 2024. Intel also announced today start Gate-All-Around 2024, and TSMC from 2025.
As I mentioned, Gate-All-Around is a really efficient technology for more performance and power. From 2024, 2025 timeframe, we are really focusing on this 3 nm maturity and power performance area competitiveness. That will be really important battlegrounds.
If you have any questions, you can raise your hand.
Thanks, and thanks for presentation. Staying with 3 nm Gate-All-Around, I mean, I can't have noticed, but you don't really have a HPC flavor yet for 3 nm. Yet you seem to expect that HPC customers will grow significantly in importance, how do you get there? If mobile mass production is 2024, when do you actually expect mass production for HPC for 3 nm? Secondly, one of the feedbacks I'm hearing from your customers, is while they're impressed with your rollout of GAA now, they at times find you maybe a little bit slow in bringing library IP to them, to help porting designs to a process. In other words, you may at times either like the library IP offering that TSMC has or perhaps you don't really service it as quickly. How could you improve on that front, please? Thank you.
The first question, recently, many customers actually, they want to have a more earlier HPC product launching compared to previous times. Previously, usually they are starting launching 2 years later than mobile or 3 years later then. Because of higher competition, recently many tier 1 HPC customers, they ask us to provide GAA even at the same time of mobile. That's our homework. Now we are working how we can provide the earlier version of HPC, more reliability and high voltage operation. Currently, many our HPC customers expected the launching of a product is in 2025 timeframe, 2025, 2026. Even though I didn't include SF3X yet, but we are working with our customer. What is their request and how we can prepare? My expectation is, some customers start in 2025 or 2026 timeframe.
Second question, yeah, that has been our weakness, because TSMC has a more wider ecosystem and because they have more customers, like 500 customers and many products. The ecosystem is self-boosting, right? Samsung Foundry has less customers, but, as I mentioned, right now our customers increased twice in few years. We are really focusing on this IP and library ecosystem. Recently, we are working with many library partners, not only the American company, but other Asia company also. Really we are focusing on this library and IP, this ecosystem. Maybe we can also provide some incentive because we cannot expect a TSMC model, right? Voluntarily, they cannot start this investment. We are working in the various aspect to motivate the partners. Yeah. For 3 nm and, I believe we will be in the better position than the previous node.
Maybe I will ask another question. This is in relation to your future CapEx. Moore's Law is very expensive. The capital intensity is rising. You are going to use probably High-NA eventually. How can you justify return on that invested capital in the future, and will there be enough customers to pay that premium?
Yeah. That is a really important point. Similar, not only foundry, but memory solutions. The CapEx is increasing rapidly. We believe that the long-term partnership between foundry and our customers are becoming more and more important. One of the way is, we are now discussing how we can have this long-term partners commitment between foundry and customer. Some customers willing to pay some investment and pre-deposit to bind our capacity and their demand together for long years. That is one year, one way. The other one is foundry is a different business model compared to memory. For example, 14 nm is almost 10 years on the business, we expect more than 10 years, 15 years still, we have a very good visibility to our business. After depreciation, right now, 14 nm is the cash cow, actually.
We have to think about long-term business, not just five years, but 10 years and 15 years. That is a different way of thinking. Another way is this semiconductor is becoming more and more important strategically for many governments. Also, we expect to have some incentive from many government, current countries. For example, our Taylor Fab, we are working closely with United States to have some support of this CHIPS Act.
I think in interest of time, I think we can end it here. We'll take a 10-minute break. Thank you very much for this presentation.
Okay. Thank you.
Thank you very much for preparing this nice set, Stuart, especially for me. I hope service business will and devoted for contribute to elevate our set business as well. It's very honored to be here to presenting DX new business here. As I introduced, I'm Serin Han, leading service business in visual display division. Today, I will be speaking on behalf of the broader service business within DX division. If you have any questions, I will be happy to take at the end. Let's begin with the device that Samsung is well-known for. On average, Samsung launches over 400 million customer product every year. More than 50 million home appliance, 40 million in TV and monitors, more than 300 million mobile devices. Even as we speak, there are 13 Samsung devices released somewhere every second.
If you count to 8 seconds, we have 100 devices somewhere in the world. In addition, more than 1.5 billion device are connecting, which brings a huge potential to go further. I would say this is one of the largest device ecosystem in the world. What makes Samsung's device ecosystem so unique is it has been a part of consumers' daily lives for a long time. Most platform ecosystem make tremendous efforts to gather users and build up relationship with them. We, Samsung, have direct access to various consumers due to its nature as a consumer electronics and connectivity technology with services. As a manufacturer, longer lifespan of home appliance is perceived as a weakness because we want to sell more device more frequently. Refrigerator has a lifespan about 10 years, TV for seven years, even mobile is up to three years.
If you wear ecosystem lens, it could be turned into rather strength or opportunity. You can easily imagine that those surface devices are occupying users' family and personal spaces. Before going to shift my discussion from device to business directly, let me spend more time to share what consumers want today. In the past, key buying factors of consumers were a set of feature and specifications. Marketing and sales message were mostly on what's new or spec comparison with competitors. Consumers are asking, "What can I do with it? What value can I expect? How does it reflect to my lifestyle?" The value proposition of a product is being evolving to focus on expected outcome out of a holistic offering composed with feature and services together. This tendency is not only for single choice of a product.
Consumers' expectation level on user experience is getting higher and more complex. According to a study, U.S. households have 22 devices at home on average, which has been doubled compared to in 2019. We have all faced these challenges, making our homes as smart and simple as we want. Making seamless connectivity across the devices requires continuing investment and technology democratization. Up to now, I have gone through what consumers want at the section of so-called experience. Let me move on to service section, where experience is completed and generates more business opportunity at the end. Each service plays a different role by purpose at its development stage, from experience enhancement to monetization. There are many services to enhance the optimal device experience within Samsung. Samsung has collaborated with various strategic partners for open innovation and enriched experience.
For example, Netflix, Spotify, Disney+ for optimized content experience on our new foldable screen, Prime Video for HDR10+ or 8K resolution, Xbox for perfect gaming experience. I couldn't imagine Galaxy Watch without Samsung Health. Monitoring health condition, key activity metrics, recommending fitness and nutrition program, and helping sleep management. Samsung Wallet is replacing physical wallet at this now. Also, it will be a critical component of a future lifestyle in virtual and decentralized world in near future. Discovery and distribution services are typical ecosystem services you can easily imagine what looks like. Those services offer more choices to consumers and open up new business opportunity for partners and developers. Ecosystem provide distribution and promotion vehicle to partners, and partners provide commercial value to Samsung in return. A commercial term like revenue share is a common business model behind.
This kind of a business model is a hygiene factor for platform business for Samsung to operate it and invest for further and better-performing platform. Samsung Smart Hub with the home screen and app store are example of TV business. On mobile side, Galaxy Store provide a variety of apps, themes, and fonts to maximize a Galaxy specialized experience, and key vertical service provided by industry-leading partners are prominently placed for users to discover and enjoy best-in-class TV and mobile experience. Some services are designed to perfectly match the device. I call it device-native experience. By combining hardware characteristics and its competencies with software service, Samsung can deliver truly native experience to consumers. These services are operated with a self-sustained business model, either advertising or subscription.
Some of them are significant enough in monetization now, some are in good pace to the destination, and some are still exploring new opportunities. Let me introduce Samsung TV Plus, a free ad-supported streaming TV. Literally, TV comes with TV. Samsung TV Plus now reached out 24 countries and 465 million devices globally across the TV and mobile. Since last fall in 2021, it has been 100% growth in customer viewing. This year, we project nearly 3 billion hours stream by highly engaging consumers across the globe. This is the TV Plus.
[Presentation]
Art Store is another good example, The Frame is one of the most famous Samsung's lifestyle screens. The idea of this unique product has been initiated to extend the screen value of TV, which remains as an ugly black screen for 18 hours a day. Art Store service concept was designed as a native experience to The Frame. When we have come up with the idea is that you stream music every day. Why not with artwork? That's the starting point. Now, Samsung Art Store is a digital art subscription service with the largest market and audience in the world. Over 2,400 artworks are available through partnership with famous museum, galleries, and artists, as well as new artists who don't have enough channel to be exposed to the people in the world.
Since this is our service, so we can provide revenue share to our partners according to their consumption. Another case is even more interesting. A device called the QuickCook, so named to mean quick cooker, was launched in Korea last summer. This is a very stylish all-in-one unit with one tap platform from shopping to cooking. Hardware itself is equipped with four different cooking functions, air fry, grill, toast, and microwave at a time for efficient preparation. Users can have the product for free with subscribing meal plan provided by our partners in HRM, so it's home replacement meal and meal kit industry. QuickCook is a cooking platform with SmartThings and provide integrated experience. From shopping at Samsung's online food mall, scanning barcode, selecting cooking mode of device automatically, and to having recommendation for next use. Family Hub is another good example demonstrating Samsung's continuing effort to innovation.
It is a smart fridge with various smart features combined with, again, hardware and services. Collectively, it has uplifted fridge value from a refrigerator in the kitchen to kitchen and home management hub. I hope for us to deliver another successful story for the market within near future. I have mentioned customers are getting more conscious on seamless experience across their all home devices. I think services to connect all the dots is the one that Samsung can do better than any others. This is a fundamental part for Samsung device to be transformed into actual ecosystem, and that's why Samsung DX Division is redefining, striving Samsung SmartThings strategy with more priority. Smart home journey needs to be silently orchestrated from finding the right device, setting it up, and to doing more with it.
Calm technology is centered on SmartThings to make this happen silently. Frictionless experience without hassle, intelligent experience to manage home, and extended experience with other services. Let's see a short video clip on what Samsung SmartThings. I hope you enjoyed the video because it's a little bit kind of this time, so it can wake up your brain. As you may remember, the title of my presentation today is, Device Ecosystem Gears to a New Growth Engine. I have explained how services contribute to make the experience that consumers want and viable ecosystem so far. Now it's time to talk about new growth engine based on device ecosystem under DX umbrella. If I make my point simple, there are two components to drive post-sales revenue stream. First one is Samsung's global distribution platform.
We have a highly engaged audience in more than 200 territories in the world with direct accessibility. The power and reach of our platform continue to attract many partners. They are choosing Samsung as a major distribution channel of their content and services. My previous slides on Samsung services have briefly illustrated how services on Samsung provide mutual benefit to both users and partners. How they can be operated and monetized for both partners and Samsung. Second one is unique user data and insights generated by various devices and services. Let me explain further in coming slide. As you can see, building blocks from the slide, Samsung is able to have invaluable user data and insights through services offered on the top of the device ecosystem.
Given proper user consent and data security protocols, we can collect data to understand them better and deeper, knowing what device they are using, when, where, and how. It drives valuable user insight eventually. The Economist has called data as the new oil of this century, and other industry experts called it the new water. Whatever you call it, its importance and unlimited value to business has not been changed, even is being increased. Search data from Portal opened up data business with material impact to digital economy, followed by social data coming from social network services. Evolution of data business is still ongoing. Why not by Samsung? Even though I limit data's value to manufacturer's standpoint, we utilize the data for hardware business immediately. Data enables a more personalized experience for user lock-in and provide actionable insight for new and innovative products.
As an ecosystem owner perspective, we certainly leverage the data for new business opportunity, like advertising. Let me go deeper into actual business case of a visual display where I come from. No matter how big your device ecosystem is, it is less valuable if neither being aligned with the market trend, nor being prepared for capturing them. Back in 2016, half of U.S. household were already watching a connected TV. Connected TV includes smart TV and streaming media device, like a stick and box. It has been continuously increased. By this year, more than 70% of U.S. home own a smart TV, which is excluding any other device such as streaming media devices. The reason I have quote U.S. data is too frequently is, U.S. is the single market with the most advanced platform business.
Samsung's business also as U.S. market is the number one, followed by European market and other emerging country. Indeed, 90% of Samsung TV sold out globally are all smart TVs. What do users watch with smart TV? OTT streaming services. It may be fair to say the rise of OTT services make people to choose more connected TV. As you can see from the graph, global OTT services such as SVOD is a subscription-based video on demand service, and AVOD, ad supported video on demand service, to grow. Based on other survey by J.D. Power last year, the average consumer in States had 4.5 paid streaming services. In order to win this competitive market, OTT companies continue to invest multi-billion KRW to generate new content. Legacy TV advertising business is being shifted to connected TV advertising more and more.
The market trend has been validated at actual TV data that I have. The share of viewing time of linear versus streaming, linear means legacy live TV, streaming means OTT watch, has flipped from 70%-30% in 2019 to 30%-70% in 2021. Live TV regime within TV has remained for several decades. However, it has been taken by OTT recently Samsung, as a 15 consecutive years of number one market share, never stop to pursue innovation in software area. Barely, we also have substantially paid the tuition fee for various trial and errors. We had introduced the world first TV app store in 2010. In the bottom, the left side, you might not be aware of, not surprising. We launched officially the Smart TV in 2011. There were market doubt and criticism on Smart TV itself and its usability for a long time.
Samsung was not an exception in the mood and criticism. In 2015, we daringly transition existing platform operating system to Tizen for better platform usability and expandability to other CE devices. It was painful, but it paid off. We introduced the whole new Smart Hub experience with code name Eden in 2015. The reason of code name is Eden was rethink TV, going back to the zero point. It provides a single access point for live OTT and external sources by treating all media content equally, which was not normal because the 2016, still live TV experience was dominant to all the TV watching time, and no one think about OTT and external sources should be treated with live TV experience. Single remote control supports the single experience.
When you introduce the single remote control, there was a fascinated feedback from the market with single experience and single access point. The result, well accepted by customers and our partners, well appraised by media and market, luckily. Those momentums let Samsung embrace the market evolution into the business. With this, service business went into full swing from then. In 2018, we renewed Samsung TV Plus that I have introduced a few minutes before, and began to monetize it meaningfully. We also introduced the Art Store at that time. In 2022, this year, we have introduced the first screen home experience as ready for grabbing all media consumption. At the same time, we are taking another wave beyond the media experience from now on. There is a small slide and looks very similar in bottom line.
However, this journey clearly reflects how Samsung TV has captured the market trend and transformed the opportunity to service business in the past decade. With the largest TV scale in the market and key services and solid business platform to address the opportunity, Visual Display's service business has grown up 86% on average per year in the past six years. Our business can be more about the ecosystem play, as I explained before. This was platform business. The platform business is one part of our service business, and the other area is advertising business. Advertising is based on Samsung as a platform, supported by unique audience base Samsung has, differentiated data generated by the service end users, and as a technology capability to target audience and serve to advertisers on the top of the Tizen platform.
As we have grown by taking dynamic market transition on time, the service business need to be agile for further growth without rest. Samsung has announced the Tizen platform licensing program to non-Samsung branded TVs last year at Samsung Developer Conference. This year, we confirmed onboarding partners at the same conference. This is an initiative to scale out our Tizen platform as a backbone of independent service business. This year, we have launched Gaming Hub. Gaming Hub is all-in-one cloud gaming services. You don't need a console to enjoy cloud gaming on your TV. Proposal of this service is not only to expand the TV user base to younger demographics. Those people, the interest in TV is not that high as we want, we would like to expand our audience base to those gamers as well.
At the same time, we would like to prepare next growth drivers of service business with this gaming service. Like any other services at the beginning, we are focusing on right experiences to target, right audience, and now. The value of device ecosystem and recurring revenue is rather simple and basic. Great products and services give consumers more reasons to buy, stick with us. As a result, more scale and data you will have. Larger scale and audience base will help expand your partnership, generate further revenue with them, and even identify new service and business opportunity of our own. Platform monetization will enhance the data-driven business structure. After all, recurring revenue will be reinvested for experience innovation and financial soundness of business. There is no reason for Samsung not to get into the great value chain. I have mentioned about the DX device ecosystem.
Let me going back to the first starting point. If Samsung's device turns into viable and fully functionable ecosystem, there are much more things to do. DX division has all blocks and dots in action. That's why we believe Samsung device ecosystem will become one of the largest audience platform to drive a new growth engine. I have no doubt, with solid market leadership of each business unit and continued efforts to create new experience across business unit, and corporate initiative within DX division will unlock even greater business opportunity in near future, I'm sure. Am I so fast? I would like to end my presentation with other key agenda in Samsung, which is not the least. We can never emphasize too much on the importance of privacy and security. Our new business relies on user data of device and services connected to each other.
In order to maintain the momentum to grow further safely, we need to take good care of them by following the principle, where we put our customer first. Our privacy principle, unwavering transparency, choices, and protection. Protection is also the foundation of everything we do in security. All are protected by Samsung Knox security platform, which is certified for use by the top government agencies around the world. Thank you very much for listening my presentation, and now it's time for your question.
I can start the question first. It's two parts, my question. You mentioned about the TV service business growth potential. Can you just quantify for us the actual results so far? Because it's something that's new. Second part would be that aside from the TV business, there's a lot of interest about the Samsung mobile services. What do you expect the returns and long-term strategy from that part of the business?
Very good question. Thank you. I cannot share the exact numbers with you. However, the service business across the DX division are increasingly becoming material. Also it is considered as one of the most profitable business within company. With TV business, the contribution to the overall financial performance is to increase given the nature of business. I have explained, the service is monetized over the lifecycle of the product, versus TV sales affect the financial performance of a single year. This is a huge potential to grow still. In terms of mobile business and other business within DX, so there is a little bit different stage of service business by business unit. Mobile division is more focused on enhancing and optimizing user experience, as I mentioned earlier.
They are more focused in and how they can resolve consumers' pain point, how they can create differentiated product with the services. However, it is the first phase of services to be developed more. Since I explained the Samsung TV business spend the 12 years to invest continuously. Which was adventure to resolve customers' pain point thoroughly, and then we can find the business opportunity after that. What difference between the past and present is the company understand the service business in nature. When we resolve the customer's pain point, it is not one time, one-off solution. They are more focused on the platform structured approach. Which makes a huge difference in future when everything is ready. The mobile could carry even greater potential because they have greater addressable market.
DA business could be a little bit different because DA doesn't have any experience in other things, but they are connected to other devices within DX device ecosystem thoroughly. It could not be generated individual service business by their own product. However, once we combine those device ecosystem together, the DA division will be benefited definitely. I hope this answers to your question. Please.
Maybe a slightly different way to ask the question. If I look at MX and when I focus on the mobile portion, excluding networks, about 20% of revenues, give or take this year, is non-smartphone, non-tablet, effectively wearables and services as well. Approximately how much of that would be services? Should we assume that gross margins there are comparable to the benchmarking industry, which is Apple, and Apple obviously discloses that. Thank you.
I'm sorry, I cannot be representing those questions and answering the questions. All the devices is going to be smart. That is the kind of destination we are looking for. I believe MX is also looking for the same direction. Please contact our IR team to get the detailed answers on them. Sorry about that. Please.
Thank you so much for this. Just a couple of questions here. In terms of how you actually incentivize as a headquarter, do you incentivize your head of business units to actually contribute for the DX? Given that there might be some conflicts of interest that the BU heads may want to sell as many devices, while obviously headquarter wants to have seamless experience for the users. Are you going to allocate your profit, DX profit with BUs, so that you can actually incentivize them to be more integrated within the systems?
I think the profit of DX is including all the profit from the business unit, that's not a matter of financial benefit. Before, we have support from the very independent approach and siloed approach by GBM, because we are too much focused on their own device to sell in the market. We have learned very painful lessons. Those approaches didn't work at all. As I mentioned, the people's expectation is getting higher than we think. We have to provide some connected experience across the device, which will increase actual value of individual product as well. Now there are good thing and good thing is within DX, all the leadership team, including head of GBM, they are all in consensus to make this experience as one for the further business profitability as well.
For service business, how we can return our business value to GBM so we can contribute to this profitability crucially.
Sorry, just another question related to that. In terms of seamless user experiences, I have lots of Samsung applications in my phones. Not necessarily that everything is very seamless. That's because obviously a lot of the application devices that you have. There might be some good benchmark for you. How are you encouraging, again, the BU heads, to integrate into your master user experience platforms that you may have?
Yeah, sure. It's the kind of real sound of market. I know there is some pain point still exist. During the last couple of years, we have the older listed of the problem, why we cannot have a seamless experience. As I mentioned, it was because the siloed approach. I mentioned about SmartThings as not only for the IoT platform, it's more the fundamental infra platform for across all the services. We have several, the meetings and task force and the working group to think about how we can make those infrastructure to be working. That's our approach at this moment. I think within couple of years, we can really make the seamless experience. The other thing is when we discuss about the connectivity, each person in each department is looking for their own job.
They would like to increase, enhance the connectivity quality within that wall. It doesn't work. That's why we are again too much focused on the DX level, corporate level initiative with SmartThings.
Hi, thanks for the presentation. Just one question on how do you see the competition coming from the Chinese, I guess the appliance producers? I think they're kind of adopting a similar strategy platform, et cetera, and their quality is also improving. What's Samsung's edge now and how you build on it going ahead? Thank you.
Yeah. We are suffered from the competition. There are some huge competition in the western market, like U.S. and European countries, where this service business has started earlier. We also have very aggressive chasing from the Chinese manufacturers. We are aware of that. The Chinese manufacturers, although they have a good ecosystem within their devices in China, it doesn't necessarily mean they can go to the global level of execution. Samsung has a global footprint and global platform and operation system, and we have a strong presence in advanced market, so that we can make our services to be distinguished from the competitors. However, the other difference is data usability. For other companies like Samsung, in other words, except China, we have a strict regulation on the data itself. We have to protect our users.
Within domestic market of China, there is much more flexibility in utilizing data to make their business. That's the market difference. However, we have to take those barriers to overcome in the near future. Thank you.
That wraps up all of our presentations for today. Once again, thank you very much for joining our investors conference, Investors Forum. Also, I'd like to thank Morgan Stanley for hosting the event as well. Through these afternoon sessions, I hope you're able to see the basis for new growth opportunities in our Memory Business, the strength of our four value pillars in our Foundry Business, and also the services growth opportunities in our DX Division. I look forward to separately meeting with many of you to have further discussions, and also look forward to future events where we can continue to share our growth strategies. As always, I appreciate your interest and support of Samsung Electronics. With that, I would like to now call a close to the 2022 Samsung Electronics Investors Forum, and I wish you all have a great rest of the week.
Thank you very much.