Good morning and good afternoon, everyone. My name is Jinman Han. I'm the head of Samsung Memory Division of Samsung Electronics. Today I'm going to talk about memory market outlook and Samsung strategy of our products. This is the agenda. First part of my talk will be about era of COVID-19, how it's going to change the memory market in terms of DRAM demand and NAND flash. The second part of my talk is about strategies, what we have in our pipeline for DRAM and NAND flash products. I'm going to talk a little bit about the Samsung strategy about geopolitical uncertainties, how we're going to deal with all those geopolitical uncertainties. First, I'm going to talk about the era of COVID-19. Well, we had data here that shows that people who are having online shopping twice a month, and we asked them, this is based on Gartner data.
On the left part of this foil shows the online shopping trend. As you can see, comparison between pre-COVID-19 and COVID, you can see that there's a huge jump in the demand for education and online courses. That's partly because of the work from home and shop from home. Another thing that we can notice here is that people buy more electronic goods. As for myself, I just bought a lot of notebooks and TVs, and my daughter is now talking to her teacher across the Pacific, and it's really amazing to see how things have changed. Although this COVID-19 thing is ravaging the people on Earth, but I really appreciate the technological prowess, how technology allow us to prepare for this pandemic. It's really amazing to see how people are getting along with these technology benefits.
Also from the enterprise point of view, Gartner data shows that more than 86% of enterprise customers are willing to invest more on digital equipment and infrastructure. That shows that this COVID-19, in some way, positively introduced the paradigm shift about people's view on the technology, and people realized that technology has to be ready for any uncertainties that may happen in the future. This photo shows the DRAM demand. It shows until 2024. We believe that CAGR will be around mid-10s and 20%, this is very solid demand. Most of the key drivers of DRAM demand is obviously server and mobile. You can see that the mobile demand is pretty solid, as well as server. Although the portion of PC and consumer products are lower than mobile and server, but the growth is pretty solid.
Especially this year, we have seen very solid demand from PC side, especially very thin notebooks, such as Chromebooks and ultra notebooks, because of the work from home and study from home. We believe that that trend may not disappear. Even though the COVID-19 pandemic is gone, we still believe that this trend will continue to be here. Okay, I'm going to talk a little bit about the mobile market. The main theme of mobile market these days is about the adoption of 5G. On the left lower side, you can see that about 250 million 5G smartphone has been deployed. We believe that annual growth rate through 2025 will be about 35%, which is amazing because how this 5G penetration rate is accelerating.
Another good thing about mobile demand is that we are seeing mobile demand, not just from the industrialized countries, but we also see a huge demand rising in emerging markets, such as India and Middle East and Africa. The average content, average density of DRAM for mobile is right now a little bit above half of PC density. PC consumes about 8 gigabytes, but right now, mobile is about 4 gigabytes, which is pretty solid. We believe that in 2025, the average density of mobile DRAM will be close to 70% of those of PCs. Every phone will be, in terms of density of DRAM, it'll be very comparable to PCs. In the middle of this foil, it shows that there is a very solid demand from the emerging markets, especially from Middle East.
The high portion of handset is coming from Middle East and Africa region. Especially when you see the India, the demand uptick is pretty strong, actually. More and more people are starting to use smartphone as their main digital equipment, and we believe that trend will continue to be in the future. On the far right side, as I said, the smartphone gigabit per system is increasing with the annual rate of 13%. It'll be close to 8 gigabytes when we reach 2025. Another equation of DRAM demand is coming from servers. We all know that we are now watching a lot of streaming videos and movies and sports, and most of the consumption will be coming from entertainment.
At the same time, it's not just from the entertainment, it's also the data demand is coming from non-entertainment sectors, such as medical imaging and video surveillance, security and telemedicine. I believe that in the future, there may be more killer application for this data demand, especially from non-entertainment images. Okay. I'm going to talk about NAND demand, the storage demand this year has been phenomenal, because a lot of work from home activities create more and more contents and has to be stored somewhere. These days, SSD is replacing hard disk storage devices. This year, we have seen those trend pretty clearly. Most of the server SSD's demand is actually coming from cloud computing. When people buy more and more smartphones, for example, I remember the figure such as that, when people buy about 100 smartphones, you need to have a new server installed.
That interaction between mobile phones and server SSDs is very eminent from this year on. You can see that both SSD and mobile demand is pretty solid, and average CAGR is well above 30%. Combine that with the average density of DRAM, you can see that smartphone in the future will have a higher density storage. These days, I think it's about a little bit over 200 gigabytes, but you may see eventually 1 terabyte smartphones in the near future. Okay. This is my strategies. I'm going to talk about, first, about the DRAM products. As you can see, a lot of people are saying that DRAM is a commodity. That's true in some way. These days, DRAM industries are experiencing huge transition from old technology to new technology.
On this screen, you can see that there are four major products that Samsung is now producing. On the left top side, you can see the LPDDR5. LP stands for low power. DDR stands for double data rate. The 5 means it has a new interface. Its speed is, right now, up to 6.4 gigabits, which is very fast. You can find those mobile LPDDR5 product in the flagship devices, such as Android phones. On the right side, we have a DDR5. It has no LP. This is DDR5, is mainly for server applications, and it can, right now, support up to 4.8 gigabits.
This is the real workforce behind the server computation. These days, the main product is DDR4, but with the new CPU online in next one or two years, we're going to see more and more adoption of DDR5. On the lower left side, you can see there is a device called GDDR6. The G stands for graphics. As you know that we have, right now, new game consoles coming into the market, and those two consoles are right now adopting GDDR6. When people say that DRAM is commodity, it is true in some way, but it is not true in some sense, because DRAM industry has been working very hard to provide the perfect solution for any new applications coming to the market.
Like I said, GDDR6 is indispensable to make sure that next generation console is meeting all the people's requirements, including new ray tracing and all those new features. At the same time, it's really about giving semi-custom product for the new IT devices. For example, on the right lower side, you can see that we have a device called HBM, that stands for High Bandwidth Memory. The main application for this product is for high performance computing. You cannot find this device in your PC, but almost all the hyper cloud computing companies and high AI server are right now adopting HBM. DRAM is now transitioning into a new era of next generation interface, Samsung is right now ready with all those lineups.
To provide a new interface and new breakthrough to our customers, not only we have to work on defining new interfaces, we have to come up with new design techniques and architecture changes. For example, DVFS, Dynamic Voltage Frequency Scaling, that has been used by CPU companies, but right now we are adopting those technologies, including AVB and HKMG. It's a little bit about the process and design, but I'm not going to go into detail about that. The problem is, to meet the future demands, that is not sufficient, we believe. We have to come up with new way of continuing scaling. One of the main topic of today's talk is about EUV. EUV stands for extreme ultraviolet. Without the EUV technologies, meeting the demands from customer cannot be addressed just by coming up with architectural or design changes.
It has to eventually come from process innovations. One of the key technology solutions for process innovation, we believe is EUV. I'm going to talk a little bit more about EUVs. This is a little bit complex, but on the left side, that's the conventional ArF based machine. On the right side, it shows a little bit of mirrors, but that's the basic diagram for EUV. One main difference between ArF and EUV is its wavelength. ArF has 193 nanometers, EUV is about 13.5 nanometers. So ArF has a wavelength 14 times longer than EUV. When you have shorter wavelength, you can have more ability to finely define the pattern. You have to have a very sharp tool to have a very sharp lines. It's that analogy. There's a drawback of shorter wavelength light source.
What that is, it can be absorbed by lots and lots of mediums. Even the EUV can be absorbed easily by air or even mirrors. You have to make sure that you have to have a very strong power source to make sure that all those light source can reach down to the wafer level. That is the main technological challenges that EUV companies have to solve. Because the EUV can be absorbed by even air, the chamber has to be in vacuum state, and also it cannot go through the mask. It has to be processed by reflection techniques. The result, you can see that on the right side of EUV, you can see very finely detailed pattern. It has its own price tag.
Its footprint of EUV machine is more than three times larger, and the number of components it has is more than two times. It's also very, very expensive. You can believe me how expensive that is. For the last 10 years, we have been using this ArF pretty well. Right now we can produce our DRAM chips based on 1z-nm. 1z-nm is about 14 to 15 nanometers, just give or take. Below 14.6 nanometers, it's going to be very, very hard, if not impossible, to process that 14.6 nanometer design rule. For the next 10 years, we have to be ready for this EUV technologies. On the right side, there is an IRDS comment. IRDS is the semiconductor technical community, and it says that no new lithography technique will be needed for better resolution after EUV.
Looks like EUV is the ultimate machine for us to be able to process DRAM below single-digit nanometer for the next 10 years. I think it has been about 10 years when we started studying on this EUV technology. So far, we have been processing about 4 million wafers until now. The 4 million, that's a lot and lots of number of wafers, and imagine how much know-how we have gathered by processing all those 4 million wafers. Okay. If EUV is so powerful, then why don't we just buy an EUV machine from companies like ASML? Well, it's not that simple. EUV is all about ecosystem. As I said, EUVs can be easily absorbed by any medium. Even the pellicle technology that has been used for memory, any semiconductor can absorb EUV light source pretty easily.
You have to come up with a lot of different technologies combining together. The one thing is the mask technology. You have to have your own mask technologies, and also material and measuring technology. All those ecosystems, including mask and material and measuring technology, has to be combined to have a holistic solution for this EUV technology. You just cannot just buy a machine from a company and deploy it into the fab. I believe that more than 10 years ago, we started building EUV clean rooms. In addition to that, we have our internal organization structure that is optimized for this EUV ecosystem development. We have a designated EUV team. We called it as a Quantum TF. We also have a procurement strategy because Samsung is one of the IDM companies who has huge economies of scale.
We are fully exploiting that advantages. One thing we realized by applying these EUV technologies for experimental purposes, we realized that it can easily reduce the number of defects. For example, it can reduce the defect ratio by more than 20%. On the right side, you can see that the two patterns. On the left side, that's the pattern with the ArF, with multi-patterning technology. In multi-patterning technology, you have to be very careful about overlaying all those layers. You can see that on the right side, EUV gives you very perfect pattern definition. Imagine how that can reduce the defects as well as the performance of transistors, even the cells. The result is a very fine pattern with the pattern precision and not comparable to ArF. Because EUV can reduce the number of steps, it can also reduce the cost.
It's a silver bullet for the next future 10-year process technologies. I'm going to talk about NAND flash. On the left side, that shows the concept of our approach for our sixth-generation NAND flash, which has 128-layer NAND flash. Right now, we are now transitioning from fifth-generation NAND flash, which has 92 layers, to the 128-layer based sixth-generation NAND flash. One unique thing about Samsung's technology is that we can etch 126 layers in a single time. We call that a single-stack technology. On the right side, you can see that two-stack technologies that we've been internally working on. 120 layers with single etching technology is very critical because eventually we have to stack more and more layers. You cannot have 256 layers in a single-stack technology. It's extremely difficult, at the same time very costly, prohibitive.
Before we are transitioning to two-stack technologies, you have to have as many layers as possible with your single-stack technology. If you can use that technology, leveraging that technology to have a multi-stack technology, then you can have more and more layers supported. We believe that Samsung is very well positioned to have two-stack technologies with many more layers. For example, since we have now a single stack 128 layers, that means that if you can just simply doubling, you can simply double the number of layers with two-stack technology. That can give us 256 layers right away. How many layers we're going to adopt for the next generation, that depends on our internal strategies, such as customer demands, our big growth target, and ramp up optimization. It's not really about how many layers you can stack.
It's really about what is the perfect optimized number of layer you can introduce to your market in a very timely manner. This one shows our best-in-class cost competitiveness, because 3D NAND flash is not created equal. For example, when you see that on the left side, you see that we have a reference line of Samsung's 92-layer TLC. There is another metric, which is called stack height. When you have a higher stack height, it takes more engineering resources and turnaround time to process those higher height stack. You can see that Samsung stack height is at least 15% lower than other suppliers. That's how we can achieve 128-layer 3D NAND flash with the single stack technology. On the right side, what's the result? Well, this data come from public data.
We can see that Samsung has the best-in-class cell engineering, and that gives us the best cost structure. The best cost structure means that you can have a lot of return on investment, so that you can put more engine resources to have a better solution for the next time. This is very positive cycle. I believe that Samsung has the most cost-effective solutions. Even though everybody's coming up with almost same layer of NAND flash or even higher number of layers, this is the reality. Samsung is pursuing very cost-effective solutions so that we can introduce to the market very cost-effective solutions. Okay, I'm going to talk about our strategies. As you all know that we have a lot of uncertainties regarding geopolitical issues. In the short term, we are now seeing unstable tariff risk.
Also, we have a little bit of difficulty having sales channel partners. In the long term, it might be possible that technology might go into the era of de-globalization. Eventually, we may see different trade sphere separate from each other. Not only just providing solution to our customers, we have to be worried about, we have to be ready for technology decoupling. We believe that Samsung is ready for that trend because, for example, this shows the worldwide TAM portion in America. Worldwide TAM portion is about 30%. China and rest of the Asia Pacific, it accounts for about 60%. In America, you can see that server portion is way bigger than other application. In comparison to that, China and Asia Pacific, most of the demand is coming from mobile applications.
Not just only we're going to see the decoupling, we might see decoupling of technologies, but the product portfolio that our customers demand from each region might be vastly different. In that era, what can we do about that? Well, Samsung, we have to come up with all different kinds of lineups ready. Not just about lineup, but also about the standard and technological directions. It has to be optimized and developed for each own region. In that regard, I think Samsung is more than ready for that situations. Although we hope that technological decoupling is not going to happen, we have to be ready for this future. We believe that Samsung is very ready for that. This is my summary. As I said, this COVID-19, although is a very painful experience, I myself want to meet my customers face-to-face.
I don't want to see my customer through the online small video conference tools, but that's the reality. At the same time, this gives us a new chance of global experimentation about how we can cope with any worldwide issues with this technology. That give us a huge insight about what you can do about in the future. At the same time, we believe that that will stoke more demands from memory, such as DRAM and NAND flash. At the same time, as a market leader, Samsung has to be actively working on the future ecosystem, especially for EUV ecosystem, so that we can provide solutions for the next 10 years. We are very willing to share this knowledge with our industry.
At the same time, we see that not only just DRAM, but NAND demand will be pretty solid for the next 10 years. At the same time, because we have a best-in-class cost structure and cell engineering, we are ready for any new demands coming from any new applications. Finally, as I said, Samsung has a huge coverage of worldwide customers, and we are already having a lot of complete lineups for our customers. Any geopolitical uncertainties, I think I can say that Samsung is ready more than anyone. Please have trust in Samsung, and we'll address all the issues and demands from you, and we'll do our best. Thank you.
We will now begin the Q&A session. You can type your question through the Q&A box, and please clarify your name and company when you type your question. The first question is raised by Dong-je Chang from Arete Investment. What is the cost scaling factor from 64-layer to 128-layer single stack, versus 64-layer single stack to 128-layer double stack?
That's a very good question. Cost scaling factor from 64-layer to 128-layer single stack, versus 64-layer single stack to 128 two-stack. We believe that at least lower 10%. I'm not sure about exact number, but the reason why you have to think about cost increase with the double stack is that you have to process two times. Right? At the same time, it's not just about the increasing number of stack, it's also between those two stacks. For example, 64 and 64, there is interface you have to deal with, it may also affect some yield drops, not so combining that, I cannot tell you exactly what exact number that is, but think about increased number of steps, as well as dealing with those intermediate boundaries between two stack. That will affect your yield. Eventually, then industry will figure it out.
I believe right now, that is truly affecting yields, so that it can increase the cost.
The second question is raised by Nicolas Gaudois from UBS. As you move to double stack NAND with [37], can you still realize a cost advantage versus other double stack, apart from the scale?
Of course. Because as I said, the height of our stack mold is way lower than other suppliers, and it gives us better TAT. Also, it's not really about height itself, it's also about the space, the pitch between cells. I believe that Samsung has the much higher tighter pitch between cells. Combining that with the lower stack height, that gives the better turnaround time. At the same time, that will give us much better yield. We believe that by even going into two-stack technology, Samsung can still maintain the cost advantages over other suppliers.
The next question is raised by SK Kim from Daiwa Capital Markets. Given current outlook on future demand, what's Samsung's investment strategy?
Our CapEx investment strategy is based on our internal demand modeling. Based on that, we plan our investment. One thing we have to think about is that, next year's bit growth will be mostly determined by this current year's investment. The investment has been done already, it's almost December. Our investment today will determine the bit growth of next year. We believe that DRAM bit growth will be mid-to-high teens to 20% next year. I believe that Samsung has invested according to that demand. For NAND Flash, although the bit demand growth is a little above 30%, we still believe that NAND demand can be very, very flexible depending on the average selling price. We believe that Samsung investment for NAND Flash will be assuming a little bit higher than what market thinks.
We believe that with the cost benefits that Samsung has, we believe that more demand will coming online next year, because this COVID-19 can be gone, I hope. Samsung ready for any new demand uptick next year. Based on that, we determine our CapEx scale. Bottom line is, we have our internal demand forecasting model, we are aligned with that demand forecasting. At the same time, we are strategically thinking about how bit growth will be like in the future based on the cost benefits that all the suppliers can make next year. For DRAM, we'll be based on the DRAM market demand, for NAND Flash, we will be more actively preparing for the demand increase next year.
The next question is from JJ Park from JPMorgan. Can you share your view on threats from China memory makers in both DRAM and NAND? What's the major hurdles for China memory to compete with mainstream players and potential threats?
I don't think it's really a threat. Any companies who have a willingness and capital can join the market. That's not really a bad thing. I myself started my career as a development engineer, I know how many hours we have spent time developing all those technologies. Memory and NAND Flash, as a whole, are not that easy to ramp up in a timely manner. You have to invest your time and money and some patience. Lots and lots of patience. It may take a little while before any newcomer into the market will hone that skill. We believe that any huge impact on the market will not be felt in the foreseeable future. At the same time, the suppliers, including Samsung, are working very hard to have a technological breakthrough, it turns out that it is getting difficult every year.
Any newcomer has to catch up with that technological advancement of other suppliers. I think it's not really a threat, but it's a very welcoming sign from my point of view. Yep. We'll continue to do our best, and I think the same will be true for any other suppliers. Yes. At the same time, these memory industry is kind of like working as a single community. Now, we are actually exchanging a lot of ideas in terms of development through the publication of papers and in forums. I think, the ecosystem of memory industry as a whole is going to be very, how would I say? It's going to be very full of ideas and breakthroughs, I hope, in the future. Yes. In that sense, I really welcome any newcomers. Thank you.
The next question is from SK Kim from Daiwa Capital Markets. There are concerns over Samsung's technology leadership, as one of our competitors recently announced the release of 176-layer V-NAND. In response to that, what is Samsung's stance and strategy regarding technology leadership?
Okay. Yes, it's really exciting for us to see new technologies coming online. As I said, we welcome newcomers and new technologies, because as a whole, we are a single community. The NAND flash technology is based on the cell types. It used to be floating gate. Samsung has been researching early on of the CTF-based technologies, we have almost mastered those charge trap technologies long, long time ago. Other suppliers, I believe, are working to transition from conventional floating gate-based cell technology to the cell CTF, so-called cell charge trap technologies, such as Samsung and some of other suppliers are using. In that regard, I believe that we are very confident about CTF maturity from our point of view. For some suppliers, that may mean new technologies.
They are, I think other suppliers are working very hard to catch up with what we have right now. As I said, the 128 layer with single stack, we can go immediately, for example, 256 double stack technologies. We have our own products in our pipeline, and determining the number of stacks to be introduced into the market will be determined by the maturity of the product, the yield of the product, and at the same time, the requirements of the customers. In a timely manner, we will introduce more and more layer-based technologies. For us, we believe that for next year, our 128 technology will be the most cost competitive, at the same time, very high-performance technology. Yeah. Okay.
This is the end of the Q&A session. Thank you for all the participants.
Well, thank you. Thank you for hearing my speech. Thank you.
Welcome to Samsung Electronics Investor Forum. Before we start the presentation, please be informed that we'll have a Q&A session after the presentation. You can type your question to the Q&A box at any time during the presentation. When you type your question, please clarify your name and company. This session will be presented by Mr. Shawn Han, the head of the Foundry Marketing Team.
Good morning. Good afternoon, everyone. My name is Shawn Han, and I'm the head of Foundry Marketing Team at Samsung Electronics. Due to the COVID-19, this investor forum this year will be held as virtual event. Thanks to the nature of a virtual event, there will be no limitation to the number of attendees we can have, which will give me the opportunity to talk about Samsung Foundry's growth strategy in front of a larger audience. During my presentation, I'm going to show you foundry market opportunities, business updates, and our strategy. Also, I explain how Samsung Foundry can support our customers and create success together in the industry. Today, semiconductors allow us to create everything we can imagine, and it is also taking a key role in the electronic devices. Moreover, the semiconductor industry, the responsibility for foundry is getting bigger.
As you can see on the chart, as the number of connected devices increases rapidly compared to the population increase, the demand for semiconductor is also expected to increase, causing explosive data generation. The foundry market has grown along with the innovations of the devices, and now even has been accelerated with the data transformation as more and more devices are connected together. We have found that foundry market opportunities are moving into a direction in two critical fronts. First, the customer's need on advanced node, especially under 7nm, drive the growth of the foundry market. Based on Gartner's market report, the foundry market is expected to grow at a CAGR of more than 9% from 2020 to 2025, and to reach a scale over $130 billion in 2025. Moreover, the number of foundries capable of supporting leading-edge process is getting smaller.
There are only two companies that can support 7nm and further advanced process technologies. I will discuss our business updates. Samsung Foundry's vision, in one word, is trust. Since we became an independent business unit three years ago, we consistently had the vision and strove to be the most trusted foundry in the world, with emphasis on service of technology, service of operation, and service for customers' success. In terms of service of technology, we develop innovative solutions like no other foundries can, and in terms of service of operation, we strive to provide a seamless supply chain experience for our customers. In overall service, we always put our customers' needs at the center of our focus. By successfully executing these tasks, we believe that we can become the foundry to lead the customer innovation in the total service.
I would like to point out some of the major achievements of Samsung Foundry as follows. With launching foundry business in 2005, we have started high-K metal gate volume production in 2011 for the first time, and FinFET technology in 2015. Samsung Foundry was the industry's first company on the 10nm volume production in 2016. Together with advanced technology leadership, Samsung unveiled 7nm EUV production and F3 line ramp up in 2018, and completed V1 line construction in 2019. As I mentioned in the previous page, Samsung started the foundry business approximately 15 years ago as an IDM business model, and became an independent business unit in 2017 as Samsung Foundry. Since then, we have worked very hard to provide leading-edge technology while building a business-friendly ecosystem. We announced Vision 2030 last year, and we are committed to investing more than KRW 130 trillion until 2030.
Our business will continue to expand with large investment we will diversify our applications and extend our customer base. Please find Samsung Foundry's global manufacturing sites for more information. Samsung Foundry has continually expanded our manufacturing capacity. Today, we have total seven global fabs, six for advanced 12-inch fabs, one for mature 8-inch fab, and test package facilities in three locations, including Korea, China, and the U.S. Our most recent fab is now being built in Pyeongtaek, Korea, which will focus on 5nm and below processes. I show you Samsung Foundry's process strategy in technology leadership. Samsung Foundry's process technology path divides into two parts. The mainstream path includes our advanced technology leadership with the first 10nm volume production, EUV adoption, and the post FinFET architecture adoption.
The other path covers our specialty technologies such as FDSOI and embedded MRAM, expanding differentiated offerings for 28, 18 nano FDSOI solutions for IoT, security, industrial, and automotive. With this concept, we will keep providing customers with value-added services in our technology offerings. We strive to maintain foundry technology leadership through continuous node scaling. As you know, process technology continues to innovate through new structures, new materials, leading to intrinsic scale down and transistor enhancement. In addition, process technology has been evolving from optimization and adoption of next-generation process modules. What is important from this is that node in the same branch share the IPs as they share design rules. For example, first let's take a look at the 14 nano family on the top.
The latest evolution is 11LPP that will provide new ultra-high density cell library by employing tighter metal pitch and layout optimization thanks to the maturity of the process. Same rule goes with the 10 nano family. As you look downwards, vertically from 10LPE down to 8LPA. We apply the same development principles starting from 14 down to 3. In the horizontal direction, we call it innovation. Vertical direction, we call it evolution. We modify the process technology to have better performance, better power, as well as the density shrink. For the 7 nano evolution, we have 5 LP providing UHD cells that is density optimized by improving the contact process. We also added 4LP and 4LPP that is based on the new process. The process nodes marked with the new flags are fully newly introduced. For example, 8LPA, 5LPP, and 4LPP.
With this concept, we will continue to provide our customers with value-added services in our technology portfolio. As addressed before, we have solidified a foot forward for growth by providing competitiveness of our advanced process technology with the start of shipments for 5 nano mobile products. In addition, we will continue our efforts to widen our lead in advanced technology process by completing the design of second generation 4 nano process technology. Let's take a look at on this 7LPP, which will give you more information on EUV lithography capabilities.
Conquering a mountain is no small feat. Only the best in the world can pave the way to the top. The blood, sweat, and tears eventually pay off. Like climbing Mount Everest, developing next generation lithography is extremely challenging. Samsung Foundry leads the industry with EUV lithography. As semiconductor patterns get smaller, image blurring becomes more severe in lithography. ARF multi-patterning has been used for many years, eventually EUV became necessary to manufacture leading-edge semiconductors. ARF multi-patterning requires several masks to define a single layer. Because of the increase in masks, cost increases while yield loss occurs. These side effects make EUV inevitable. Pattern fidelity of EUV is also incomparably better than multi-patterning. EUV utilizes reflective rather than transmissive optics. In order to maximize EUV reflectance, multi-layers of special materials are necessary.
Conventional tools are not capable of seeing embedded defects in between the layers. Samsung Foundry's EMDRS can capture these defects. EMDRS is a unique in-house tool, which enables defect-free EUV masks. Samsung Foundry has successfully performed R&D activities to overcome the challenges in EUV source, resist, and process. As a result, Samsung Foundry became the first to mass produce EUV-enabled chips with the smallest pitch amongst all 7 nanometer plus nodes. Samsung Foundry constructed a new dedicated EUV fab to support the growing demand of EUV and will continue to make significant investments. EUV will continuously be utilized for advanced applications such as 5G, AI, HPC, and autonomous driving. EUV was recognized as an insurmountable mountain to conquer. Samsung Foundry was able to make it to the top. Samsung Foundry will continuously develop technologies to support our customers' unparalleled success. Samsung Foundry.
Here's another very exciting new technology, which is Gate-All-Around transistor. 3 nano GAA node is the new technology node after 4 nanometer. This is the world's first GAA node, and its value comes from Gate-All-Around nanosheet transistor, with improved short channel effect and higher performance operation at low VDD. You can also check this advanced technology through a video clip.
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 make 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. However, 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 FET. MBCFET consists of multi-stack nanosheets. One advantage of MBCFETs is that additional area is not required to improve speed.
FinFETs need fins to be laterally added, while nanosheets can be vertically stacked. MBCFETs are compatible with FinFET designs. Designers can replace FinFETs with MBCFETs without changing the footprint. Performance can also be improved without area increase. MBCFETs 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 MBCFETs, Samsung Foundry will be the guiding light for the future of the semiconductor industry. Samsung Foundry.
Our current 8-inch value-added offerings with various portfolios start from embedded flash, power IC, display driver IC, CMOS image sensor, analog to power discrete. We provide diverse processes down to the design rule of 65nm and the low voltage of 1.2V. Technology is extensively optimized through many years of product manufacturing. Other than process technology, package platform becomes very important in meeting the increased demand for more functionality. The performance platform comes with package solutions based on heterogeneous integration package platform specifications. Also partners qualify their package solutions with Samsung wafers. There are various kinds of package solutions to choose from based on customer's need. For example, 2.5D RDL package is a low cost, high bandwidth solution. I-Cube is a very high bandwidth solution. X-Cube is a solution to enable short memory access latency and small form factor based on 3D TSV.
X/I-Cube is the integration of X-Cube and I-Cube to provide short memory access latency and high DRAM bandwidth. 2.5D RDL and X-Cube are already in volume production, I-Cube and X/I-Cube will be ready for mass production by the end of next year. Please take a look at the video, which has received a lot of attention from customers recently.
The increased demand for more functionality continues to result in increase in size of chips. To overcome limitation of the chip size, chips are broken down into functional dies and connected at the die level to form a system. Vertical stacking of functional dies enables significant leaps in speed, power efficiency, low latency, and small form factor. This new 3D innovation is completed through matured TSV technology with many years of mass production experience in micro bump. X-Cube is the first generation of 3D IC wafer level integration solution offered by Samsung Foundry. The X in X-Cube stands for extended interconnection density and extended functionality. With X-Cube, system designers can enjoy greater flexibility to build custom solutions that best suit for their system requirements. Silicon demonstration of X-Cube has already been completed by stacking SRAM on a logic chip, freeing up space to pack more memory into a smaller footprint.
Furthermore, Samsung is providing silicon proven design methodology and flow for advanced nodes. X-Cube will continuously be utilized for advanced application such as mobile, AR, VR, wearable, and high-performance computing platforms. The evolution of 3D IC solution, X-Cube, will be the solution for beyond Moore's Law. Samsung Foundry will keep developing technologies to support our customers with unparalleled success.
Foundry business would not be a viable solution without partners. Samsung Foundry consistently works with their partners to build an ecosystem that can support all of the customers' needs. We transform customers' idea into silicon with EDA, IP, design service partners, cloud, process technology, and fab, as well as package solutions as a one-stop solution. To actively respond to the latest technology trend and lower the design barrier for developing competitive SoCs, we launched SAFE, that stands for Samsung Advanced Foundry Ecosystem, program in early 2018 and held an annual SAFE forum since 2019. In order to build grand ecosystem alliances with EDA, IP, DSP, cloud, and other partners, Samsung Foundry continues to maintain and solidify its partnerships. It's almost time to wrap up my presentation. To sum up my talk today, there are four key takeaways.
First, we will continue to maintain firm leadership in the advanced node technologies for the digital transformation, for the future needs. Second, Samsung will continue to expand differentiated offerings and solutions for various applications with SAFE. Again, SAFE. Third, Samsung Foundry will enhance our supply chain through continuous investment. Lastly, we will provide the highest quality of service to enable our customers faster time to market to ultimately grow together with our customers. Thank you very much for your attention.
We will now begin the Q&A session. The first question is raised by Dong Zheng from Arete Investments. On a scale of 1 to 10 for the degree of synergy, as 1 being no synergy, while 10 being the full synergy. What's the level of synergy on EUV between Foundry and DRAM? How is that ramp-up cost and cost from learning curve allocated?
Well, it's very hard to quantify the scale about that synergy between foundry and memory division. What I can tell you is, since EUV technology is the top of the line module process and still under improvement in many ways, we are still working together with the equipment company. Compared to the other companies, certainly we have more learnings from two different divisions. Some of the advantages of this synergy or mutual learning would be, we can share new learnings from memory as well as foundry, because both divisions are dealing with almost limit to the physical patternings. Usually, memory divisions are learning a lot more wafers, they can pick up more data from the real silicons, especially how to control the regular patternings. Foundry also can learn many different aspects for the random patternings. The basic knowledge can be shared between those two divisions.
There must be a synergy, but it's hard to quantify the level. We certainly have the advantages.
The second question is raised by SK Kim from Daiwa Capital Markets. How will Samsung close the gap with TSMC regarding five nanometer process?
Well, we started volume production of 5-nano process, the 4-nano process has been under development. In addition, through the launch of the development of second generation 5-nano and 4-nano processes with improved PPA. PPA stands for power, performance, and area. We are providing wider process portfolio and striving to maintain process leadership. As you know, furthermore, in order to respond to an increase in the future demand for cutting-edge products based on EUV, we strengthened the foundation for expansion by starting investment in production lines in Pyeongtaek Complex.
The next question is raised by JJ Park from JPMorgan. One of the issues for Samsung Foundry is the limited customer base relative to TSMC. What's your strategy to expand the customer base, and what's your advantages over your competitors from customers' perspective?
It's difficult to mention anything about customers. What I can tell you is, although we started our business based on the mobile products, where we have the strength, now we are moving into different applications, including network, data center, and other HPC applications as well. We are also working with many different companies in different segments, such as consumer products as well as automotive. We are in the middle of moving into other segments and applications.
We will take one last question from Daiwa, SK Kim. What is the general expectation for the second half 2020 and the outlook for early 2021? Thank you.
Well, there are many uncertainties because of this COVID-19 and geopolitical situations. What I can tell you is we are trying to overcome this situation through our product mix and developing new technologies, focusing on high performance computing applications as well as new mobile applications coming from this 5G era. We are also trying to diversifying the applications, as I mentioned, and expanding global customer base. Again, the outlook is not that clear. There will be a lot of uncertainties ahead of us, but we'll try our best to overcome the barriers ahead of us.
Okay. That's it for the Q&A session due to the time constraints. Apologize that we should now wrap up the Q&A session. Thank you for all attendees who has participated, for your time today.
Thank you very much for listening and your attention.
Welcome to Samsung Electronics Investors Forum. Before we start the session, please be informed that we will have Q&A session after the presentation. You can type your question to the Q&A box at any time during the presentation. Please clarify your name and company when you type your question. This presentation will be presented by Mr. Yong-In Park, the Head of Sensor Business Team. Please go ahead for your presentation, sir.
Hello, everyone. Welcome to Samsung CMOS image sensor, ISOCELL. I am Chan Yong-In Park, general manager of sensor business. It is a great pleasure to reveal ISOCELL at this conference. I think everyone uses smartphone camera. Through this conference, I expect you can understand about CMOS image sensor technology and its huge business opportunity. It would be great if this conference was held face-to-face. COVID-19 has changed many aspects of our lives. I want to ask you a question. I think you use many video conference. Do you like the video conference image quality compared face-to-face meeting? My dream is to provide the sensor like as a real human eye viewing in the short future. Let's talk about how image sensor can help. Today, I want to share with you our clear vision of ISOCELL. More pixel, more camera will bring us better life. CMOS image sensor.
CIS means CMOS image sensor. ISOCELL is our trademark. Pixel converts light signal to electrical signal. Pixel is the key technology for the image sensor. Pixel is composed of color filter, photodiode, and circuit. It is the most important to bring crosstalk between different color pixels. Each pixel should be isolated to prevent the other color signals bleeding in. It will make sense. The name ISOCELL is isolate and cell. ISOCELL stand for capture pure light and keep the image very well. For you understand, let's start with a short video clip, how exactly ISOCELL achieves and provide best image quality, especially in smaller and more pixel.
In general, when taking a photo with a smartphone, the light reflecting off of your subject passes through the lens and is captured by the image sensor. The captured image is then further processed by an ISP, or image signal processor, before it is saved into an image file. The image sensor's main role in the whole photo-taking process is to accurately convert the light, an analog signal, into digital. A CMOS image sensor is made up of a pixel array on the top layer and a logic circuit at the bottom. The pixel array layer converts the light received into an electric signal, which is then turned into a digital code at the logic layer. The pixel array is comprised of millions of pixels. The total number of pixels of an image sensor is also expressed as a resolution.
For example, an image sensor with a 64-megapixel resolution has 64 million pixels in the pixel array. A pixel has a micro lens at the top to collect the light, a color filter to sort the light by its intensity in red, green, and blue, and a photodiode to store the light. The metal line, or the logic circuit layer, is at the bottom to convert the stored light information to a digital signal and apply needed corrections. Smartphones today are getting slimmer and come with a full screen display, limiting available space for a camera module. To support higher resolution in a same or smaller size module, pixels in the image sensor need to get smaller and smaller, but without loss in quality. To absorb as much light as possible, pixel structure has evolved to a back side illumination, or BSI, from front side illumination, or FSI.
By placing the photodiode layer on top of the metal line rather than below, the BSI structure captures more light, which is essential in photography. By doing so, it increases crosstalk between the pixels, contaminating color information. In 2013, Samsung introduced ISOCELL, an industry's first technology that isolates the pixels from each other by adding a wall-like barrier around the pixels. The name ISOCELL is a compound word made from the words isolate and cell. By isolating each pixel with a physical barrier, ISOCELL increases a pixel's full well capacity or ability to hold more light and reduces the crosstalk of light particles from one pixel to another. With such benefits, ISOCELL enables the image sensors to deliver higher color fidelity than conventional BSI.
Before we dive into the details of our business, let's review the camera system by comparing to your eyes. A mobile camera is composed of lens, actuator, image sensor, and image signal processor. Like a human eye sensing system of lens, ciliary muscle, retina, and brain. Let's look at this human eye. Oh man, this is a very beautiful eye. You can see very good lens. It's only single lens. Very beautiful iris. If I look at it from the camera module point of view, we use about five to seven lenses. The iris is a shutter, is still a lot of gap to the near to human eye. Resolution wise, now, last year we announced 108-megapixel is world first, the most CIS. According to the scientist, the human eye resolution is about 576 megapixels.
ISOCELL's vision is to implement 600 megapixel sensor beyond the human eye resolution into mobile phone. More pixel is the first key strategy for ISOCELL. We are now pushing forward to implement DSLR camera into the smartphone through ISOCELL. Usually, DSLR camera interchange five lenses for better shot and special effect. Smartphones instead switch multi-cameras built in. High dynamic range video without imaging lag is also the key factor. More camera solution is the second key strategy in ISOCELL. Just two years ago, more than 32 megapixel smartphone camera was very rare. ISOCELL opened the new era of more pixel. ISOCELL started to provide 32 megapixel and 48 megapixel CIS in 2018 and changed the trend of a smartphone camera by providing 64 megapixel and 108 megapixel last year. More resolution is better. Let's look at the main camera adoption ratio.
This year, 35% of smartphone choose more than 32 megapixel, and expect 56% in 2024. I think more because more resolution is better. I want to make it more than 80% in 2024, providing performance and price-balanced ISOCELL product. Front camera has also similar trend. Please remember, more resolution is better. Let's look at the more camera. Although smartphone volume may saturate at around 1.4 billion unit, the total cameras in smartphone will continue to grow. Why? Dual camera, triple camera, and the quad camera adoption rate will be increased. Triple camera and quad cameras ratio is more and more. It means each smartphone use more cameras. The total resolution of all rear cameras in a smartphone is around 23 megapixel two years ago, and 59 megapixel this year. This is expected to double to 112 megapixel by 2024, but I expect more.
It will result in many wafers will be needed. That means more revenue opportunities in CIS market will drive there. ISOCELL has created the more pixel and more camera trend by providing tiny and excellent performance of ISOCELL technology and camera solution. Samsung started the CMOS image sensor 20 years ago. 2013 is the very historic year for the ISOCELL. Samsung CIS platform decided the ISOCELL as a reference platform for the Samsung CMOS image sensor. ISOCELL launched the world first 1.0 micron 16 megapixel CIS in 2015. There are many, many problems in that kind of technical barrier, but ISOCELL has removed the total technical limitation every year, and reaches 2.7 micron portfolio lineups this year. Let's talk about ISOCELL technology in more detail. This is a kind of training session.
For accurate color, it is crucial for each pixel to receive and contain a large amount of correct light information by maximizing full well capacity and minimize crosstalk each other. Photodiode is usually defined under the color filter with limited area because technical limitation. It means smaller pixels' photodiode area is more limited, and resulted, the full well capacity becomes smaller. Full well capacity means, as you see in the picture, the volume of the photodiode. If the full well capacity is small, let's look at what problem will happen. At the bright condition, when strong light signal came in, the limited full well capacity of the container will be overflowed. It means you lost some signals. Let's look at leakage. Some of the signals will leak to the neighboring pixels. We call it as crosstalk. Crosstalk is a critical parameter in image sensor performance.
If the leakage ratio is higher with the smaller pixel, you will get false color. It is very difficult to make a smaller pixel because of smaller means more crosstalk. ISOCELL uses a deep trench isolation, which physically separate between the pixels. You can implement photodiode in whole volume of this container and enlarge by taller when pixel becomes smaller. Full well capacity will maintain even though pixel is smaller. Captured signal is also perfectly stayed in their pixel by isolated wall. To implement this unique solution, ISOCELL reuse the well-proven Samsung's DRAM technology to implement this isolation very stable and cheaper. I think for the business, cheaper is very important to provide the worldwide mass volume. Samsung's DRAM capacitor is the foundation of ISOCELL's deep trench isolation technology.
Adopting DRAM technology allows ISOCELL business to reuse the legacy facility from DRAM, handle well-proven equipment, and shortening the set-up time. It results to increased profit and fast respond to the dynamic market requests. Let's look at the aspect ratio. Aspect ratio is between width and depth, is 55 in DRAM and 53 in ISOCELL. It's very similar. We are using very similar technology. For example, the highest building in Dubai has 163 floors on 140-meter foundation, and aspect ratio is about six. If we apply aspect ratio of 53 to the building, it becomes 1,460 floors. It's very tall. Not only that, if you make 100 megapixel ISOCELL should build 100 millions of buildings on a single chip. You can ask the first challenge, I think you can build it.
The building should not destroy or hit each other by strong wind, earthquake, vibration, or at any situation. The second challenge is even more difficult. There are very compacted buildings next to next. The sunlight from the top should reach to the first floor of all the buildings uniformly, very stable. You can imagine how much difficult we can make it. My engineers doing that kind of very fancy job. You may wonder why we go so far to develop such a smaller pixels. The problem is the sensor size is limited in smartphone due to camera module height. The module height is proportional to the sensor size. If you see the diagram at the middle The camera module in smartphone is composable five to six piece of lens, even eight piece of main lens.
At the bottom there, by field of view theory, the center lens, if you use larger sensor, needs longer focal lens. If you increase the sensor size, the camera become taller. For example, if you build a 0.8 micron 600 megapixel camera, the camera height will be 22 millimeter. Even current smartphone, theirs is only eight millimeter or less. It make smartphone big, bump, and ugly. To develop an ultra-high resolution CIS, it is mandatory to develop ultra-small pixels. You have to maintain same performance, not lose any performance, even smaller pixels. There is a trade-off and physical limitation. ISOCELL technology enables further pixel size reduction and contain more light signal. ISOCELL is highly competitive, especially in the more pixel market. If you make pixels smaller, then receivable pixel dimension is smaller. Very many problems in low light condition at dark.
At the low light condition, how we can take a brighter picture image? First, device innovation. How we collect the more light signal and not losing anything. I already explained ISOCELL technology can improve by capturing solid light signal and hold it. Second, pixel operation and algorithm innovation. ISOCELL introduced Tetra and Nona technology. Tetra means four, Nona means nine. At the dark condition, sensor receive the signal by two by two modular pixels, which is Tetra. Three by three is Nona pixel. At the bright condition, ISOCELL automatically convert the Tetra or Nona into higher resolution signal. Nona technology was implemented in our 108-megapixel of HM1 and HM2 product. ISOCELL offers highly detailed pictures with natural and vivid colors in all variety of lighting conditions. This year, 0.8 micron is the major. Next year will be a big year for 0.7 micron image sensor.
ISOCELL is already manufacturing 0.7 micron product portfolio family from 32-megapixel to 108-megapixel. ISOCELL technology is continuously improving to capture more light with smaller pixels to deliver better image quality each generation. As you've seen in Samsung's DRAM technology history, Samsung is good at making smaller and better. We reduce that advanced technology to ISOCELL. Let's move to more camera. Corephotonics joined ISOCELL last year from Israel. Corephotonics has many innovative camera solution IPs. The folded zoom telecamera was already adopted in smartphones. Corephotonics is adding valuable IPs every year and attracting customers with solutions that leverage ISOCELL. Due to COVID-19, smartphone market was decreased in first and second quarter this year. Let's review the main camera resolution trend. 20-megapixel and low resolution are rapidly decreased this year. However, here shown blue, more than 48-megapixel sensor showed steady growth, even crash period this year.
Third quarter-wise, year-over-year, more than 48-megapixel rear camera growth rate is about 125%, not 25%. Why? More resolution is better. Camera differentiation is more important because more resolution can show more detailed image and people want to get it. Samsung's ISOCELL created this market of more pixel and dominance. Major provider or only provider for the 108-megapixel ISOCELL has 100% market share. More pixel trend will be accelerated next year as well. ISOCELL has prepared this opportunity and we are very busy now. Most of our smartphone manufacturers are using ISOCELL. With a wide range of product offering from premium smartphone to mass volume smartphones, ISOCELL has solid customer relationship and a well-balanced customer portfolio. We work with them closely to provide the best ISOCELL and camera solution. As a business partner, we create and grow the market together.
Let's talk about some of our new technology. In addition to CMOS image sensors, we are working on an ISOCELL vision lineup that can capture things beyond the human eye. Time-of-flight sensors can capture 3-dimensional object information. Dynamic Vision Sensor can detect very fast movement, even in an extreme dark or bright environment. We are exploring and researching sensor technologies that can beyond visible light. If you see right side, the visible, like as rainbow, is very narrow compared to the whole solar spectrum. It means ISOCELL has still lots of opportunity to generate new camera and new area. For example, by detecting ultraviolet and infrared light, you can easily identify spoiled food or even see object in bad weather. If you see the right side of image, visual is some normal camera in the dark condition, but short-wave IR camera, you can see the like daytime.
ISOCELL technology is also compatible and competitive in other applications as well. For example, automotive application will use more image sensors than before. With more pixel and more camera, ISOCELL will surely capture and deliver roadside information in greater detail, improving accuracy. In addition, ISOCELL's applicable market can be extended to area like as AR/VR and drones. Last year, KPMG surveyed semiconductor growth opportunities. 72% of CEOs replied that sensor will propel the semiconductor industry in the future. Today, most of our sensor revenue comes from CMOS image sensor, and most of that is from the mobile communication industry. CIS usage in market like automotive, industrial, and healthcare are also expecting fast growing. By looking at various survey and expert interviews, the whole sensor market will exceed $143 billion in 2030. ISOCELL is very exciting and preparing these opportunities.
To develop extreme technology and adapt it to ISOCELL, I would like to make our world better. The dream will come true. We make sensors to help people and save lives. Thank you for your attention.
We will now begin the Q&A session. For participants who have any questions, you can type your question to the Q&A box, and please make sure to clarify your name and company when you type your questions. The first question is raised by Mr. Min Song Huang from Samsung Securities. How many wafers would you convert from memory to CIS in 2020 and 2021 respectively? Thank you.
Yeah. It's also my expectation. We already have more than 100,000 of wafers per month now, but we plan to receive one fab from the DRAM next year. Now, sorry, I can't tell you now the exact number, but we got a lots of order for more pixel CMOS image sensor, including 108 megapixels. Each wafer, the CIS numbers are limited, so I need many more wafers. We quickly transfer our one DRAM fab to CMOS image sensor next year, and to provide our valuable customer to satisfy to use it. You can imagine how many wafers for each DRAM fab.
Thank you. The next question is raised by Mr. Dong Zheng from Ariel Investments. I have a question on CMOS resolution increase. For the common users, it becomes very hard to differentiate two pictures beyond a certain resolution. What has your study found that resolution threshold to be? Beyond what pixel level does ISOCELL technology start having the biggest impact?
If you just look at the default mode of a camera capturing, it would be the same, because as I explained, ISOCELL also applied tetra or nona image default mode, is almost 12 megapixel, almost the same. Actually, my wife, after we took the picture, she zoom in her face, then you can totally different. If you zoom in in more detail, 12 megapixel and 108 megapixel, it's about nine times difference. You can take one picture, can cover most of area by using only one picture.
The next question is raised by Mr. Chris John from PineBridge. Your competitor's strategy seems to focus on big pixel size instead of higher pixel resolution. Do you think Samsung's strategy to focus on more pixels will allow you to increase market share versus competitors in the mid to long term? Thank you.
It's an excellent question. Actually, ISOCELL's more pixel already contain the big pixel. For example, our 0.8 micron, 108 megapixel nona. In nona mode is a 2.4 micron 12 megapixel, which is my competitor say is a big pixel. We already provide the big pixel performance and function. In addition, is a bonus. 108 megapixel is something we can differentiate. Also, people like it. If you are same price, one is to provide only 12 megapixel, the other one provide 12 megapixel and also 108 megapixel, which one do you want to buy? Even in the some market trend this year, the big pixel around the 12 megapixel is some deeply decrease, but more pixel, especially 108 megapixel, is a double, double. It's a huge increase.
I think if you see the very fine high-resolution image once, you will hate back to the blurred image. If you see, you will believe. Please try.
The next question is raised by J.J. Park from J.P. Morgan. You mentioned automotive is the new growth driver in CIS. Can you elaborate what's the major differences between mobile application and automotive, and rough idea on ASP difference and product specifications, et cetera?
It's a very similar, but totally different area of market. The first one is the temperature. Usually, the mobile phone is a commercial level of temperature, but automotive CIS request very extremely high temperature. The second thing is the mobile camera. Snapshot is more important for now adding the video, but automotive video is more important because always the automotive is moving. Also, dynamic range. Now in smartphone, around 80 dB over dynamic range is applied. But automotive area, like as if you pass the tunnel and go out, if dynamic range is lower, then you cannot see outside a very bright area because of limited dynamic range. Because the reason now, the automotive CIS request about 120 dB dynamic range, and they expect 140 dB more and more. That is a kind of major difference. Also, ASP is also different.
Usually, the smartphone use two or three years lifetime. As you can imagine The automotive area, it should supply five years, 10 years, longer time. Also, the smartphone use only module, but automotive area use a package. Some ASPs are automotive areas, a little bit higher than a smartphone.
The next question is from Simon Yu from Bank of America. When will 600 megapixel will be fabricated? Is it 2025, or is it going to be earlier than that? Also, could you elaborate more on the competition with Sony and Chinese and Taiwanese fabless companies? Thank you.
Yeah, 600 megapixel is our target. We need to develop much smaller pixel to implement it. 2025, maybe. Usually, my engineers, some pulling the schedule. Anyway, I would like to provide a 600 megapixel resolution in someday earlier than 2025.
The next question is from Dong-je Chang from Arete Investment. For auto application, will you still be using CMOS or CCD? How does one differentiate on global shutter capability? Thank you.
Yeah, CCD is a global shutter, mainly difference between CCD and CMOS image sensor. CMOS image sensor can add the logic and analog circuit same onto the pixel. There's a major advantage we can provide the single chip solution. Global shutter is a kind of a dream sensor for the image. Now we're also working on global shutter by using our secret, very advanced technology. It's also very synergy with the DRAM technology. Please expect our global shutter will be announced soon.
Thank you, everyone, for your time and attention. Apologize that we should now wrap up the Q&A session here due to the time constraint. Thank you, everyone, and thank you, Mr. Park.
Thank you for your attention and your very valuable questions. I hope we see you again. Please remember, more resolution is better. Thank you.
Welcome to Samsung Electronics Investors Forum. Before we start the session, please be informed that we'll have Q&A session after the presentation. You can type your question through the Q&A function at any time during the presentation. Please clarify your name and company when you type your questions. Today, we have Mr. Jongmin Lee, the Head of Mobile Strategic Planning Group. Please go ahead with your presentation, sir.
Okay. Hello, I'm Jongmin from the Mobile Communications Business. Today, I would like to speak to you about our foldable smartphone, which we first introduced last year under the theme, Fold New Era. As you all know, we successfully commercialized foldable smartphones. Today, I want to share with you how we have spared no effort to successfully deliver this innovative product to the market. They are having a lot of behind-the-scenes attempts and efforts to deliver an entirely new form factor into our customers' hands. I'm going to tell you a little about the way in which we unfolded this new era. As you are all aware, the smartphone market has been somewhat stagnant. Differentiation, especially in premium markets, has therefore become more important. However, we were struggling to find a new growth engine that would help us overcome this market situation.
While smartphones have evolved in many areas, including screen size, camera, and performance, the bar-type form factor has prevailed for over 10 years. Many people anticipated the release of smartphones with innovative form factors, but due to various technical hurdles, there had been no change from the bar type. In particular, foldable was the one form factor that many had long dreamed about. Even within the company itself, some did not believe that its commercialization would be possible. Above all things, we found it very difficult to convince others to create product that was unlike any other. However, after a long period of research and development, we finally introduced our first foldable phone to consumers last year, and many people said it was a real innovation. I will begin with showing you how our foldable product was born.
Let's listen to the story told by Patrick Chomet, who leads our product experience. As you can see, the people were fascinated with our foldable smartphone. Based on the positive response following the Fold release last year, we launched Z Flip and Z Fold2 this year under the Galaxy Z series, which is widely regarded for its highly sophisticated quality. We took a closer look at why people like our foldables. Our internal customer survey indicates that foldable customers selected the larger screen, productivity optimization, and distinct form factor as key buying factors in their purchase decision. Most Fold users were satisfied with the viewing experience presented by the immersive screen of the Fold. They also saw an increase in productivity related usage, such as searching for information or reading and writing emails, and they felt that the product is completely optimized for business.
In addition, the top three reasons for purchase mentioned by Z Flip users are all related to the distinct form factor, as you can see. We've focused on the voice of our users and have improved the competitiveness of our foldable phones through various approaches, including further development of features that users like and easing user concerns over foldables. Now, let me tell you about the innovative features we've been developing. UTG is a result of long-term cooperation with many partners striving to achieve the same goal. Second, there is inevitable gap between the body and the hinge in the foldable structure. We developed sweeper technology to prevent dust ingress into this gap. With the sweeper, we were able to minimize the gap and prevent dust ingress, which was one of the biggest concerns. Let's take a look at this video to see how this technology was created.
All right. As you can see in the video, the sweeper technology is a good example of innovation created by properly adapting elements from other industries. One of the best things about the Fold is its larger screen. We applied Infinity Display to the main screen, which allows us to enjoy a more immersive viewing experience. In addition to this, we also considered the fact that people's use of the cover display depends on their environment, and we applied a more expansive display to the cover screen. As a result, both cover screen usage time and customer satisfaction have increased. In fact, I'm using the Z Fold2 like that and utilize both screens depending on my environment. For example, the cover screen is used mainly while I'm on the move, like a subway. I'm using this device a lot.
In the office, I usually unfold the phone and use the main screen. In other words, I think it is awesome that you can use two different types of screens on one phone. Next, I'm going to talk about the optimized larger screen experience. To be honest, a tablet is a vast useful device, but it is difficult to carry everywhere. With Fold2, I have the advantage of carrying my phone and my tablet with one hand. I believe this is the perfect device for business people like us who read a lot of emails and documents. We optimized the screen layout to allow users to see a variety of information all at once on the wide main screen, as you can see below. Of course, if you just want to see a larger display, you can select the layout on the right.
Next, I will talk about the multitasking experience. When I used a bar-type smartphone, I didn't multitask that much. However, after using the Fold2, I realized the true convenience of multitasking. In particular, I think it is very cool that you can watch a baseball game and do your job at the same time. Now there is no chance I'll miss any home runs. In addition, you can easily attach photos by simply dragging and dropping. Let's see how this works. All right. You can feel how these small changes actually give us an incredibly convenient user experience. We must mention the hinge design when we are talking about foldables. User of Fold1 wanted to use the Fold at the desired angle. In order to provide the experience that user wanted, we tested various hinge structures.
As a result, we apply the cam mechanism so that the device can freeze them at multiple angles. It opened up many new experiences to users. Let's take a look. Now let me explain the user case that shows how the form snap of the hinge is actually used. User can now conveniently video call and watch video at desired angle without a separate stand. We call it Flex mode. Flex mode is also useful for taking pictures. Even without a tripod, you can take pictures and shoot videos hands-free by simply putting the phone on the table. In addition, we support AI technology-based auto framing, making it possible to shoot professional videos with the Fold alone. It also provides a new flexible camera experience. Whenever I traveled and someone took a picture of me, I often wondered how the photographer would look.
Sometimes I was disappointed when I saw the picture later, with Z Fold2, there is no need to worry. As you can see below, there is a rear camera and cover screen at the back of the Fold. Using this, the person being photographed can also check his or her pose in real-time. Not to mention that you can easily take high-quality selfies. As you see, the new form factor creates unexpected new experiences. More importantly, users are happy with these new experiences. Based on this feedback, we are trying to deliver a better experience. Of course, this experience is not created by us alone. We've been collaborating with top-tier companies like Google and Microsoft for a long time. As a result, we've been able to provide an optimized experience that takes advantage of foldable form factors.
In addition, we are strengthening our foldable ecosystem as more companies continue to join our ecosystem. Last but not least is brand collaboration. Foldable has become a fashion accessory given its unique design. We understood this and launched the Thom Browne Edition, which has gained explosive response from customers who are interested in the latest technology and fashion. Case in point, we saw fierce competition for lottery sales of Galaxy Z Fold2 Thom Browne Edition. Only one out of 100 people won a chance to buy it despite its premium price. As such, Thom Browne Edition has received a lot of attention and is contributing to the reinforcement of the premium brand image. So far, I have experienced 10 improvements in the Galaxy Z series.
I'm also curious about how the next foldable products will be like. I'm looking forward to seeing more users have new experiences in the near future. We will continue to present a variety of products to solidify the Z series as a mainstream of the premium segment. The next foldable could have slimmer and lighter design, or it can be a product with differentiated technology, or it could have a new form factor. The foldable is not an expansion of the existing smartphone. It has created a totally new category. As a leader, we are constantly innovating. Thank you.
Thank you. We will now start the Q&A session. Please feel free to ask your question through the Q&A function. The first question is raised by SK Kim from Daiwa Capital Markets. How many foldable smartphones do you target to sell next year? When do you think is the time for foldable smartphones to be the mainstream in the smartphone market? Thank you.
Thank you for asking us. Actually, I'm sorry, I can't say the exact number of foldable smartphone sales. Actually, this year, our sales performance is much better than last year. I am absolutely sure next year will be much better. It will be popularized in the market. Foldable smartphone is in the mainstream of our smartphone, especially premium smartphone. Please look forward to seeing our more innovative and the new phones with new innovative features. Thank you. Yeah.
Next question is also the following-up question from SK Kim from Daiwa Capital Markets. You said you are using the Galaxy Fold 2 by yourself. What's the biggest advantage of this foldable smartphone that you're using?
Okay. Yeah. Actually, I explained many things in the presentation. Actually, I'm a big fan of the Fold 2. Actually, I have used Fold 1 and Fold 2 and even Z Flip. I'm a big test user. The most favorite thing is watching video, like Netflix or YouTube. In the subway, I just watch like this. It's enough to see my YouTube video. Let me show you something. Okay. I can see just like this, or this is enough to see the video. Okay. Oh, actually, there's no video. Yeah. When I unfold my foldable phone, actually, I can see the video in the big screen. I can share to see the video with my colleague. The other one is, as a businessman, I have many cases to see the email or some document with my phone.
In the past, I used tablet a lot. Sometimes, I didn't bring my tablet, but these days, I just bring Z Fold 2 itself. In summary, I think this is quite business use and entertainment use, both. When I bring my wireless keyboard, that is a perfect tool for me. If I have a monitor in somewhere, I can connect it using a DeX. It is almost like a PC experience. Yeah. That's why my family, like my wife and my children, also wanted to use this kinds of the quality device. Thank you. Yeah.
There's no more questions from the attendees, so we will wrap up the presentation for the mobile segment here. Thank you everyone for your attendance and attention.
Thank you. Yep.
Welcome to the last session of Samsung Electronics Investors Forum. Before we start the session, please be informed that we will have Q&A session after the presentation. You can type your question to the Q&A box at any time during the presentation. Please make sure to clarify your name and company when you type your questions. This presentation will be presented by Mr. Incho Hwa, the Head of AI Solution Team. Please go ahead with your presentation, sir.
Thank you for the introduction. Good afternoon, everyone. I am Incho Hwa, the leader of AI Solution Team at Samsung Research AI Center of Samsung Electronics. Today, I'm going to introduce On-Device AI, which is one of our AI research directions. I would also like to show you its potential and our plan to innovate user experience through Samsung products. First of all, I would like to start with the definition and concept of AI to help you understand AI more easily. Then as a way to realize and operate AI, I will explain both cloud and On-Device AI. What is AI? It's an intelligence that is demonstrated by man-made machines, just as AI's claim indicate. To put it in another way, it implies machines replicate the human mind. AI's ability spans a wide range of human's ability.
For example, the same as human perception, AI can analyze information from environment and process a variety of senses, including vision, voice, smell, touch, flavor, and et cetera. Also, AI can replicate language ability, which facilitates the communication among humans and accumulates knowledge. In addition, AI can also make a plan, which includes ways to complete the target goals. If all these AI abilities are realized, machine can understand the world as we humans do, thus AI can help us better. The graph on the screen shows the three pages that AI has developed through. It was 1950s that the concept of AI was first introduced. Due to the lack of the computing power, first AI winter had come. From 1980s, a large number of studies were done to create AI, which has similar expertise to human expert.
At the time, by describing expert analogy as logics or rules, machines can solve the specific domain problems based on that. However, it was quite difficult to make and accumulate the knowledge well. The second AI winter came again. Years after that, with the rapid increase of computing power and globally connected internet, a new era of AI has come. Now, machines can automatically learn logics from big data. In short, now you could think AI is the same thing as deep neural network, which mimics the human brain structure. With the deep neural network, what other things AI can do? In 2016, we witnessed a surprising AI event, AlphaGo. As many of you already know, Google DeepMind was matched against the world's top Go player, Lee Sedol, and scored unexpected victory over Lee by four to one.
What's so special about this game is that AI finally has the ability to pick the best move among 381 factorial options and beat an expert at Go that's been around for thousands of years. After the match, many people started to realize that the era of AI has come closer. Another example is the adaptive cruise control. AI can detect a safe distance, speed, or lanes from vehicles ahead and automatically assist driving. This technology is widely available these days, and people in the automobile industry even say that an entire new world will come in just a few years, where fully autonomous driving is everywhere. There are various voice assistants coming out as well. Voice assistant understand the user's command and perform tasks quickly. This consists of a delicate combination of AI technology, speech recognition, language understanding, and speech synthesis.
All these are applied with deep neural network. These AI technologies are playing a major role in innovating user experience on products as well. Some products have always been designed to give users the best vivid multimedia experience. Applying deep neural network on this area, AI can innovate the multimedia quality of products such as image, video, and sound quality. AI can act like a human, so it can support a natural interaction, such as a conversation and on authentication. AI can learn and evolve based on the user's usage pattern of the product. Product can operate automatically without any user intervention. Therefore, as a product manufacturer, AI-based user experience innovation is essential for Samsung Electronics to provide the best products in the world. There are two ways to implement AI, cloud and on-device AI.
Cloud AI has its operation centering around the cloud, while on-device AI operates on each device. Not all functionality of the cloud AI can be transferred to device due to resource constraints of device such as computing power, battery, storage, and so on. Certain parts of the cloud AI, for example, personalization of the users or device-specific functions, can do better on-device AI. Within the nature of the on-device operation, it could be better to enhance the privacy, to provide a reliable and fast operation, and easier to process conditions of the surroundings. Next, let me give you examples of how each of these user experiences can be realized specifically. Oh, sorry. Sorry about this one. First is the privacy enhancement.
As shown in the picture, when AI-based voice assistant is operating, if it is based on the cloud AI, as usual, the user's voice should be sent to the cloud to be processed. In this case, to utilize the voice data with the user's approval, definitely, voice assistant providers are required to protect the user's personal data, including voice, through advanced technology, such as anonymization, and so on. On the other hand, if there is a voice assistant based on On-Device AI, because it processes the user's voice on device, sensitive user data, including voice, does not need to be transferred to cloud. If a user says, "Read the most recent message," in this case, if we are assuming the cloud AI, the recent message should be transferred to cloud, since it has to synthesize the response voice based on the message.
On-Device AI does not need to send the message to the cloud, since it can synthesize the response on device. In short, even though cloud AI provides the best protection on the personal data at the moment, On-Device AI does not need to send the sensitive data to the cloud, such as voice message and so on, at all. That is why On-Device AI can provide enhanced privacy with users. The next user experience is reliability. On-Device AI works only on device, On-Device AI could work steadily without effect from the outside of the product. For example, let us assume that you are using some features based on the cloud AI, network connection is unstable, such as in elevator, in flight, or even in case of the server maintenance. You are absolutely unpleasant with those situations. Right?
In contrast, On-Device AI can operate reliably no matter when and where you are. The reliability of On-Device AI gives an advantage, especially in basic functions of a product, which should be stable at every time, such as voice interface, video sound quality improvement, and so on. On-Device AI is also quite good to understand the context of the user as well as the context of the product quickly and precisely, because literally, AI is on device. As shown in the picture, when a user started a camera app and tried to create a super slow motion, On-Device AI can recognize all user's command. It can suggest the easier way to access those functions. For example, in this case, On-Device AI can tell you to say, "Record the super slow motion video to Bixby," instead of taking every step.
Of course, it is possible for cloud AI to provide such a feature, since it requires continuous tracking of the user's status and device context, it is unrealistic for cloud AI to provide the services consistently. Of course, On-Device AI is quite faster than cloud AI because it does not need the network connection. By learning the efficient AI algorithm, which has the similar performance to the cloud AI, and acceleration technology such as NPU, On-Device AI could execute faster than cloud AI. I touch these technologies in the next few slides. From now on, I would like to tell you about technologies around the On-Device AI, what technologies we already have and how well prepared Samsung is. Compared to cloud AI, limited resources are inevitable to On-Device AI.
As the space moves from the cloud to device, AI has to be operated with relatively limited computing, memory, storage, and power, as shown in the figure. The current AI, which is represented by deep neural network, has mainly focused on having human-like performance in a cloud system, where infinite computing power is provided. Just recently, it became possible to operate in the resource-limited environment, thanks to improvement in computing environment and development of deep learning technology, along with the chip and software. Currently, there are two major research directions of On-Device AI. First one is about how to make efficient structure of a deep neural network or represent it efficiently, the other is how to accelerate it with hardware or software to operate deep neural network. From the next chapter, we will look at what these are specifically and what technologies are being prepared in Samsung.
A basic study to increase the efficiency of the neural network was published by Google in 2017. The study showed that a network named MobileNet, which detects the visual object in the scene and has performed three times less calculation, can show similar level of the performance to existing deep structured algorithm, GoogLeNet and VGG-16. In short, there are many ways we can find to dramatically increase the efficiency of the networks while maintaining the best performance. Since then, many network structures and methods have been suggested as alternatives. Our researchers, with this inspiration, have tried to make a cloud-based voice recognition operate On-Device AI as well. In 2019, our colleague, Chanuk, and his team conducted a study that showed 13 times smaller neural network algorithm can have a similar performance to cloud AI. The size of neural network from 500 to 38 megabytes while maintaining 91% accuracy.
The smaller the size is, the less the amount of the computation as well as memory is. That means it is possible for AI to be operated On-Device. Recently, our researchers have come up with new ways of expressing neurons, which is the element of the neural network. Neurons are originally represented by 32 bits or eight bits, but they devise the way to represent the neurons less than one bit by applying the error correction algorithm in communication system. With this algorithm, the neural network could be compressed by 40 times. This achievement will be announced at the NeurIPS 2020 conference in the next week. In the next chapter, we are going to talk about how to accelerate AI On-Device, which is made up of this highly efficient neural network.
Many of you may have heard about recent NPU that perform neural network energy-efficient way and fast. As shown in the graph, we assume that the CPU can perform one visual task, such as detecting object in the image. GPU can perform three times faster NPU about 20 times faster. Therefore, as the number of the AI operated in this product increases, NPU is essential to accelerate efficiently. Regarding the NPU, recently, many NPUs have been introduced, those are mainly focused on visual tasks. We already also have the similar type of the NPU. In addition to that, Samsung researchers also conducted a NPU, which is specialized to the voice processing. As shown in the figure, voice recognition NPU can operate with about four times less power than CPU. Now we have a base NPU, which can accelerate both voice and visual tasks.
There are many cases where several On-Device AI algorithms can work together at the same time, rather than just one AI algorithm works. Figure shows the example of the augmented reality on mobile phone. In this case, application should know the depth of the image as well as the segmented object to mix virtual object into the real scene at the same time. Depth estimation and scene segmentation could be implemented On-Device algorithms, as shown in the figure. In this case, by using the Linux Foundation open source software called NNStreamer, which is led by Samsung Research, these AI algorithms can be efficiently coded and automatically scheduled to maximize the parallelism. Dynamic combination of the AI algorithms could be efficiently supported On-Device also.
Many of the studies I showed you earlier are conducted not only at Samsung Research in Korea, but also at global AI centers around the world. Silicon Valley, New York in America, Toronto, Montreal in Canada, Cambridge, Moscow in Europe, and numerous researchers in 15 overseas R&D centers. I can say Samsung is so confident to make the advanced technologies around On-Device AI with these researchers. Convergence of products became a trend these days, AI features are not limited to a single product. We needed to build the On-Device AI platform, which helps AI features to be easily applied to all Samsung products. Figure shows the platform structure that we are building. We have AI algorithms that enables AI features like voice, vision, reasoning, system software and hardware like NPU that accelerates AI algorithms.
Also we have dev tools which support developers to extend the new AI algorithm, as a result, new applications with the AI algorithms. Samsung's On-Device AI services and applications can be served at the right time at various devices. Every year, we sell over 300 million products every year. With our hardware, software capabilities for many product, we will keep innovating user experience through our On-Device AI platform. Let me tell you how this On-Device AI technologies has been realized so far, and how it will be provided in the future. Let's watch the video first. Based on On-Device AI, we created a new camera experience named Single Take. As you can see in the video, instead of taking a snapshot at a specific moment, Single Take magically pulls out the best moment while you hold the mobile phone and automatically applies the optimal filters or effects.
Without any expert knowledge about taking photo, users can get the best pictures with a Single Take. Next example is 8K television. As you probably know, 8K content is not so popular yet. However, even when you watch the popular content which is SD or HD quality, you could feel the same 8K experience with On-Device AI, called the 8K AI Upscaling. As is shown in the figure, we created more realistic content using On-Device upscaler than the signal processing technology used before. Based on this technology, Samsung has opened the 8K television market and continues to lead the market despite the lack of the original 8K content. In home appliances, we are trying to change smart appliances to AI appliances. Previously, when you are using the washing machine, you have to manually set the washing mode, which may be the best option for your clothes.
Since machines operate based on setting, clothes sometimes get damaged, and you feel upset all the way. Now, as you can see in the figure, On-Device AI, called the [audio distortion] , remembers the users patterns and consider the weather or season, laundry machine could recommend the best options to users. In addition to the existing products, there will be much more products AI-enabled in the future. First one is AI glasses, which could be the next mobile device. The figure illustrate how it would be like to take pictures or to search information through AI glasses. To support these features in real-time, On-Device AI should detect the important situation and event, or to detect the target object is essential for these use cases. By On-Device AI, a wearable mobile device can monitor the environment continuously and then gives total new experience to users.
In the future, definitely, robots will have the same intelligence level as humans, and this intelligence will become a helper for humans. It will be around you no matter when or where you are. On-Device AI will carry out a perfect role as your caregiver by understanding and providing information that you need. As we have seen so far, Samsung is preparing On-Device AI optimized for each product. With On-Device AI, we are going to keep innovating the basic features of existing products, mobile, TV, digital appliance, and so on. We are going to enable the new type of devices, such as AI glasses, robots, with On-Device AI. I want to say that this On-Device AI is not just for a single product. In other words, we will evolve On-Device AI, which can make all AI-enabled products collaborate with each other to give you the best experience.
Thank you for listening. That's all I have.
We will now begin the Q&A session. For those who have any questions, you can type your questions into the Q&A box. Please clarify your name and company when you type your questions. The first question is raised by JJ Park from J.P. Morgan. On-Device AI seems to have many advantages over centralized AI based on cloud. If On-Device AI becomes popular, how to resolve comparability issues among various devices made by different companies, unless they adopt Samsung solution?
You are wondering of the auditability among the companies, right? If we apply the On-Device AI to each device. Definitely, company-wise, we can test a lot of situations with the various devices. Still, when we are providing the connected experience to you with the multiple devices, we always do some testing with the various environment, with many combinations of the products. In that case, I think definitely we can provide the best experience based on Samsung's product, but I don't know how we can extend it to the other device compatibility. Maybe we might make some standard to be compatible with each other, if it is necessary for the users. I think it's not concrete yet, but I think we need to consider the case.
Thank you. The next question is raised by SK Kim from Daiwa Capital Markets. Is the new hardware absolutely necessary for the On-Device AI? If so, the price of such products seems to be increased.
Thank you for your question. It's a very good question. Currently, NPU is not mandatory for all On-Device AI features. It means that there are On-Device AI algorithms which can be operated in CPU, GPU through model compression or optimization. However, if we have the specialized hardware like NPU, we can serve the extremely low power or real-time capability to differentiate the product. If there are the cases to have the multiple AI features concurrently, I think if there is an NPU, it's good to have, definitely.
Thank you. The next question is from Simon Oh at Bank of America. I have two questions. First of all, what is the incremental cost to add AI in the low-end phones and low-end home appliances? Please go ahead for the first question.
Yeah. First one is that you are asking about the incremental cost for the low-end devices. Currently, we are testing to spread our On-Device AI technology to the low end, especially. In some cases, CPU over the quad core is a low-end device spec. It's quite enough to run in some cases. High-loaded AI algorithm cases, definitely memory or CPU, even NPU, is necessary for the user cases. It depends on the product differentiation. If the low-end-only devices want to apply the On-Device AI, we need to profile some functionality of the On-Device AI, which is only working on the CPU. We will provide a low-end answer at the high-end profile of the On-Device AI for each device.
The second question from Simon Oh was, what is the AI penetration ratio among Samsung phones and consumer electronics?
It's quite a difficult question for me, actually, because I'm from the research side. On-Device AI. Mainly, we are working on the flagship model first. As I already told you, if we needed to expand that kind of functionality to the low end, we will embed more powerful CPU or powerful NPU to use the features. It depends on the product decision. Yeah, that's the situation right now.
Thank you. There was a follow-up question from SK Kim from Daiwa Capital Markets. Would Samsung continue the cloud AI in the future?
Definitely, Samsung will continue cloud AI. It means that currently, several Samsung representative AI-based services, such as Bixby and recommendation services for TVs and home appliances, are provided through cloud AI. When on-device AI starts to be deployed on our products, some parts of the AI functionality, like voice recognition of Bixby, could be migrated to on-device because it has a benefit compared to cloud AI. Collaborative recommendation service will still be in the cloud to make better recommendation based on global trend. In other words, on-device AI and cloud AI, each has its own specialized area, and it will be used to collaborate with each other to bring the whole experience together.
We are now waiting for the next question. If there is no question raised right now, actually, we got a follow-up question from Simon Wu from Bank of America. Which OS will be promoted more actively, say, Bixby versus Alexa? Do you have any outlook for the Tizen?
Actually, the Tizen as well as the Android, we are considering both cases, because we have a variety of products, like Android-based mobile as well as the Tizen-based TV and digital appliances. On-device AI, which we are preparing, definitely it will be deployed on both sides, not only one. You are asking about the Bixby and Alexa. It is also not covered by me. Sorry about that.
For those who have any questions, please feel free to type your questions in the Q&A box. As there is no question from anybody, actually, we got one last question from Rex Liu. Sorry about the confusion. The last question is from Rex Liu. Will Samsung open the NPU API to outside developers?
Still we are investigating. It means that still inside the first, maybe if there is opportunity to be open, I think it might be open to the public, but I'm not sure when it will be.
Okay, as we don't have any more question, we will wrap up our presentation here. Thank you everyone for your attendance and your attention.
Thank you