Good morning. Good morning. I am Sang-Hyo Kim from Samsung Electronics IR. Thank you for attending Samsung Investor Forum in Singapore. Usually we have our investor forum once a year. Last year we had that in Hong Kong, but this year we choose Singapore. After that, actually, Singapore is getting more popular place for very important meetings such as coming political big meeting. Right? Anyway, yeah, I'm very happy to be here in historical scene. Today, as usual, we prepared three topics. First, Vice President Sanghyun Lee, Head of the Foundry Marketing Team, will walk you through our foundry business. With demand growing from areas such as AI, 5G, IoT, and automotive technology, the importance of advanced process technology is rising. Thus, this session will focus on Samsung Electronics' differentiated capability in continuously developing and utilizing next generation technology.
Next, Yongseok Choi of Samsung Display will talk about our automotive OLED business. OLED panels have become a mainstream feature in the small and midsize display market, especially for smartphone. We plan to expand the scope of our OLED business and from among numerous new applications, we see strong potential in the automotive business. Accordingly, we will provide an overview of our automotive OLED business, which will drive long-term growth based on advanced technology and mass production capability. Last but not least, the topic is about server memory. Explosive growth of data usage and processing has sent demand for high-capacity memory, especially from the data centers. At the same time, we also expect to see new demand from new areas, including AI. For this topic, Senior Vice President Sewon Chun, Head of the Memory Marketing Team, will offer insight into opportunities and our strategies for the upcoming data centric era.
I believe today's event will give you the future technology roadmap and growth potential of Samsung. Once again, thank you for coming and right, I would like to invite Sanghyun Lee. Please welcome him.
Thank you. Thank you, Mr. Kim. Good morning, everybody. First of all, I really appreciate your attendance here and your interest in Samsung Foundry technology and strategy. Today, I'm going to talk about three key items. First of all, I will briefly talk about the business update, and then I will talk about our process technology roadmap. We upgraded our roadmap based on last year's roadmap after talking with a lot of customer base. I'm going to talk about our platform solution. Platform solution includes all the IPs, libraries, infrastructure package, so that we can provide one-stop service to our customers. The three key items. Two weeks ago, actually, we had Samsung Foundry Forum in Santa Clara, U.S. to announce our first year anniversary. We became independent business unit in May 12th last year, and we built a lot of customer base interaction after that.
During my discussion, I will show you how we upgraded our infrastructure, our strategy based on those learnings during the one year. Our strength, as you might know well, by working with Memory and the R&D Center, Test and Package Center, we built very good technology node. Our technology has become world first in most of the key FinFET node. We have been building infrastructures on top of it so that we can provide best solution to our customers. After my discussion today, I hope you can find out why Samsung Foundry can become the world number one provider, the total solution provider to the fabless companies. Previously, before independence of Foundry, we had Memory and System LSI as two independent business units. Foundry was part of the System LSI, so we functioned as IDM company.
From May 12th last year, Foundry has become independent business unit. We are getting support from Research and Development Center, Test and Package Center together. Those two units are supporting Memory and Foundry all together. They provide the fundamental research and fundamental key solution for FinFET and the EUV technology and further. Now we are pure play foundry provider. Customers are building trust based on our pure play foundry business. Since independence, our President, E. S. Jeong, some people may know well about him. E. S. Jeong made trust as our philosophy of business operation. To get the trust from our customers, we provide technology, operation, service. For technology, we provide process technology, the best process technology to our customers, and infrastructure, which is very customer-friendly.
We also provide the one-stop shop from the design, the very design of the customers, to the fabrication of the wafer and the test, everything. For operation, we strive to provide flawless supply chain management. We deliver execution based on our commitment. In terms of service, we try to provide best service beyond the customer's expectation, based on our experience and knowledge. This slide shows our history as foundry player. In 2005, we started foundry business using the bulk CMOS. In 2011, we are the first High-K metal gate at 32 nanometer, and we started the mass production in S2 Fab. 2015, we became the first FinFET provider in the whole world, using the 14 nanometer. Our 14 nanometer, I will talk about 14 nanometer evolution later. We provided 14 LPE for the biggest AP fabless companies, including System LSI, Qualcomm, Apple.
We also provided 10 nanometer in the industry first. We made the evolution to S3 line for 10 nanometer and seven nanometer. We are building EUV, full EUV line in Hwaseong. We have many manufacturing sites worldwide. First, our S1 line is in Giheung. It makes 65 nanometer and below. We have S6 line six actually, which provides eight inch. I will talk about the details of the eight inch later. Our S3 line provides 10 nanometer and seven nanometer in Hwaseong. Actually Giheung, Hwaseong, they are very close, so you can just think they are very close altogether. S4 line provides CMOS image sensor, using 45 nanometer and below process node. We are building EUV line. We started constructing in February this year.
We have S2 line in Austin, Texas, which fabricates primarily 14 nanometer, and it expands to 65 nanometer to include all the legacy process nodes. We have two test and package centers, one in Suzhou, China, the other in Onyang, Korea. By having these manufacturing sites, we strive to provide one-stop service to our customers so that customers can enjoy the fast time to market. This year, we expect the foundry business, the overall worldwide foundry business to grow by about 5% or a little bit more than that. Total revenue should be around KRW 63 billion. As you can see in the bar chart, most of the revenue increase will come from the advanced nodes, from 20 nanometer, 14 nanometer, 10 nanometer, including the competitors' 16 nanometer and beyond as well.
When we had one 30 nanometer foundry, there were 22 foundry service providers in the whole world. Now at 10 nanometer and 7 nanometer, there are only three foundry companies. You can see that Samsung remains as one of the three foundry provider on the beyond node. We will keep on striving to become the best technology solution provider here. Last year, we reported KRW 9.8 billion of revenue, including the System LSI as well. Not just System LSI. System LSI has become a little bit less than half of our revenue, and we are actually expanding the customer base significantly. Compared to about two years ago, our customer base was doubled now. At the end of this year, we expect a huge growth in terms of customer base.
The reason why we could expand the customer base, you can see here From the advanced technology node to latest technology node, we are providing evolutionary process node on each node. We add RF technology for IoT companies, automotive, and we also add derivative process nodes for 10, 8, 7 nanometers, so the customers can choose the process node based on requirement. For performance, they will choose 7 nanometer and 8 nanometer. For cost base, they will choose legacy to 14 nanometer. That is our basic strategy. I am going to talk about our process technology node readiness. This slide shows our basic strategy of process node migration. First, starting from 14 LPE, here, the first FinFET in the whole world. We provided 10 LPE by using pitch scaling and the new process architecture.
We are going to provide 7 nanometer LPP using another pitch scaling and also EUV line. From 7 LPP to 3GAAE. GAA means Gate-All-Around. It is fundamental change of the MOSFET structure, and this will be another groundbreaking technology. We are going to provide this one as the sole provider, I believe. From 14 LPE to 10 LPE, 7 LPP, 3GAAE, there are always innovation of process technology, innovation of libraries. On top of that, we are providing evolutionary process node migration, as you can see here. From 14 LPE, we provided 14 LPP. E means earlier version, so E was used for most of the top-tier companies who doesn't need full set of libraries. 14 LPP, P means performance, and we provide 14 LPP for general customers. We provide full libraries, full IPs, and all solutions package as well.
For cost-sensitive customers, we provide 14 LPC and another upgrade, upgraded performance, another cost solution. 11 LPP is, up to now, our best cost solution on 14-nanometer node, and I think this is very competitive even compared to other foundry service providers. I think our PPA, performance power area, is one of the best. So right now, most of our customer base is 14 LPC, and it is migrating to 11 LPP now. 10 LPE, likewise, we provided 10 LPP, 8 LPP, 8 LPU, and the cadence on each node migration is one year. We provide a derivative node every one year. Starting from 5 LPP, there are some upgrade of our roadmap compared to last years. We provide 5 LPE, which is easy migration from 7 LPP, so that we can provide a very performance-enhanced version to top-tier companies.
4 LPE, 4 LPP will be mass production process node for global customers. The reason why we numbered 7, 5, 4 was based on our EUV technology and our working with the research and development center, we could provide very enhanced process technology. But you can see that in a single box here, the design rule is very compatible. From 7 LPP to 4 LPP, the basic design rule is very compatible. Or the 4 LPP design rule is a super set of 7 LPP, meaning that even though you migrate to 4 LPP, you can use the IPs built on 7 LPP. So customers can enjoy easy migration using our technology node migration here. This shows the roadmap on timeline from 2018 to 2021.
We already prepared the 14 LPE to 14 LPU, 10 LPE to 8 LPP, and we are going to risk produce 11 LPP and 8 LPU now. On the second half of this year, we are going to risk produce 7 LPP, the first full EUV technology. Next year, we are going to risk produce using 5 LPE and 4 LPE, and then 4 LPP and 3GAA will follow. 7-nanometer readiness. The picture here shows our fab using NXE:3400 equipment, EUV lithography tool. On the right side, you can see our wafer per day capability. So right now, we are approaching to 1,000 wafer per day. Around 2020, we expect to provide 1,500 wafer per day. Most of the increase of the capability comes from source power.
We already demonstrated the 250-watt source power, and the other two items are high-sensitivity resist, photoresist, and also the availability improvement, which means that for EUV, we can almost fully utilize the EUV equipment all day. Theoretically, there is some period EUV should shut down a little bit because of the Sorry, I just forgot the terminology. But EUV equipment should have some break time, but based on our technology migration, technology innovation, we could increase the availability. This picture shows the mask defect, and Samsung is the only company who can use the mask defect inspection tool. Other foundry does not have this tool. It is because, since working with memory, we developed the EUV defect inspection tool from the R&D center, and we can fully utilize this one to inspect the defect.
The layout on the bottom line shows our migration from 5 nanometer to 4 nanometer. As you can see, there are some technology like MDB, 1 FIN, and others, which can help reducing the layout area. Those layout area reduction comes first of all from back-end-of-line, metal general change and also some of the library innovation. On 3 nanometer, first of all, let me introduce our 3 nanometer GAA technology here. This is the picture showing how our 3 nanometer is built on. We call it MBCFET, which means that multi-bridge FET. You can see that 3 nanometer is built using three stacked gates together. We are stacking three altogether, and we also provide variety of width, starting from either examples 15 nanometer, 30 nanometer, 45 nanometer.
15 nanometer can be used for high-density design and the low-power design, while the 45 nanometer can be used for high-performance design. As you can see, variety of layouts, this is for high performance, this is for high density. Our research center developed the device so that it almost approaches the ideal limit, fundamental limit of the CMOS performance. This is a kind of sub-threshold characteristic. Physicists and engineers usually believe 65 nanometer is the ideal limit of FinFET, and we approach this one using our 3 nanometer Gate-All-Around technology. On top of the FinFET and EUV technology, we provide differentiated technology on the legacy nodes, starting from 28 LPP, going to 18 FD-SOI. We have numerous volume productions using 28 LPP plus eFlash, and we are adding RF on top of it. They are used for IoT, securities, and MCUs.
There are many companies who require much lower power than 28 LPP, we provided 28 FD-SOI. We worked with STMicroelectronics to develop this technology, and we are now mass producing using 28 FD-SOI. We provide the RF capability, and by working with the tier 1 customers, we are going to mass produce eMRAM on top of 28 FD-SOI as well. They are utilized for automotive and MCU businesses. Next year, we are going to provide 18 FD-SOI, and on top of 18 FD-SOI, we will provide RF and eMRAM together. Basically, these differentiated technologies are very useful for consumer devices and automotive devices, which require very low power and high performance together. This shows our 8-inch process. Primarily, until 2016, we worked with the System LSI for those businesses, including eFlash, power, driver, CMOS image sensor.
From the second half of 2016, we expanded the customer base. We provided these technologies for global customers, and now we have customers using our fingerprint technology as well. This fingerprint technology is starting from 180 nanometer for LCD. Now we are for fingerprint on display FOD, which is utilized for OLED. eFlash is also advancing, starting from 65 nanometer, now we provide RF capability and automotive Grade 1 capability. They are used for many customers now. Likewise, power is very popular now. We provide both mobile power and the general purpose power, which has 90 nanometer plus high voltage, about 35 volt, and for general purpose, it goes about 70 volt. Our customer base is extended first in China, second in USA, and third in Eastern Asia now, and we are also expanding customer base in Europe as well.
For the future business, we provide variety of advanced package solutions. First of all, we are very close to mass producing using this FO-PLP. This is an example of FO-PLP technology, memory, logic, stacked using the FO-PLP, and this will be used for next year's mobile phone. We also provide I-Cube technology, which is Samsung proprietary technology for 2.5D, and we have the capability for four HBM stacking right now, and we are preparing 2.1D as well, which is based on RDL technology to provide the six HBM and eight HBM together. We have some discussion with customers who like to build 3D SiP. On top of logic, it could be SRAM, it could be some kind of DRAM or logic-to-logic stacking.
Since we have numerous experience on HBM and the CMOS image sensor to make the 3D stacking, we are going to provide this technology for AI, HPC, and other areas as well. Let me move on to Samsung platform solution. Especially, I'm going to talk about our solution for AI, HPC, and connectivity solutions. AI is categorized in three areas. For cloud computing, people uses HBM, 2.5D, 3D technology. For edge computing, like mobiles, automotives, people will like LPDDR and also GDDR as well. AIoT, simply IoT consumer devices like speakers, wearables, they will use LPDDR and low-cost solution. We are preparing for those three areas. How we prepare for those AI and HPC? Since we have experience working as IDM company, we know how to start from the very design, wafer manufacturing, test and packaging, and debugging all together.
We have those teams in Foundry business unit, we are talking with the customers. We are providing our total solution to our customers. Examples are memory and SerDes interface IP, libraries, analog solution, security solution. They are not just simple IPs. We provide total solution to our customers, that by using, for example, our analog IPs, they can optimize the top level architecture. We also provide design methodology and power signal integrity solution all together. This is not all. We build SAFE. SAFE is Samsung Advanced Foundry Ecosystem, Samsung's ecosystem. We have about 50 partners to participate in this SAFE ecosystem. The key members are shown here. We provide full line of design infra, we can provide a total solution, including memory and package as well. On the package, we just not provide the hardware package.
We provide design for testability and the power signal integrity as well. For 2.5D using HBM, the debugging is very serious issue now for many big customers. We work together with our partners and also memory company to provide our proprietary solution for debugging. By using our debugging solution, they can easily debug the issues. By having these solutions, we can provide short turnaround time for customers, they will enjoy fast time to market. This slide shows some examples of our SAFE ecosystem. We are building IP and the library very significantly. We have invested significant amount of money and resource starting from late 2016, we are building most of the key IPs working with these companies. We also have a strong partnership for EDA design methodology, we are expanding the design service partners as well.
This list shows the key examples of our design infrastructure and the packaging infrastructure. For memory interface, HBM2, 3, GDDR6 are the key IPs. We have been not the number one provider, but from now on, for HBM2 advanced version of 3.2 gigabit per second and the GDDR6 of 16 gigabit per second and above, we will become the first provider, and that's what we try to be. We are investing a significant resource for those IPs. Likewise, for SerDes, we provide up to 112 gigabit per second SerDes and a variety of die-to-die interface as well, working with Synopsys, Cadence, and Rambus. Okay. Our security IPs are one of the key IPs for HPC and AI, because the big data should be protected very carefully. Hardware security IP is very important because software security is always prone to hacking, right?
We have fundamental solution to provide those hacking using our hardware IPs. PUF, Physically Unclonable Function, is one example. On top of it, we provide three items, anti-hacking solution, chip-to-chip authentication, and the secure DRAM protection. This shows our example of security platform solution. We provide all security IPs and software all together to our customers so that they can just use as turnkey solution. DRAM, many security solutions should sacrifice the DRAM access bandwidth, but we have our proprietary technology so that the DRAM access bandwidth cannot be sacrificed. That is our solution we already provide to our customers. This is another key differentiating factor of Samsung Foundry, one-stop service. To provide a true one-stop service, we should be able to provide from the very circuit design, way for manufacturing, package turnkey solution, testing altogether.
Especially for the fabless companies like AI, they do not have enough engineering resource. They highly rely on this one-stop service. We already have those in our business, they just provide the idea and the top architecture, we do everything for those customers. For circuit design, we use our internal agent team. We also use our design service providers here. Likewise, in packages solution, we could provide package turnkey solution, meaning that we can provide HBM to enhance the package and agent service altogether so that customer can get seamless service from us. They don't have to worry about the HBM inventory. They don't have to worry about the debugging. Let me move on to connected solution. We see that the trend is moving from PC era to mobile era, now IoT and automotive era.
The number of connected devices will approach 125 billion in 2030. We see a tremendous business chance in this area. We are building a platform solution for mobile, IoT, and automotive. For mobile, especially for 5G, we already provide 14 LPC RF for top-tier mobile companies, fabless companies , we are also going to add 8 LPP RF and 7 LPP RF. For lower power solution and for lower cost solution, we provide FD-SOI RF, 28 FD-SOI RF, and 18 FD-SOI RF. You can see here that we are already the world first mass production company for FinFET RF, meaning that 14 LPC RF is already used. For automotive, we prepared to meet the AEC-Q100 Grade 1 qualification, 28 LPP to 10, 8 LPP, we finished the qual of all the process nodes, also IPs.
At the fourth quarter of this year, we are going to provide the solution on 7 LPP. For IoT, low power and connectivity nonvolatile memory is the key for IoT. Our solution is 28 LPP RF eFlash is one solution, the 28 FDS plus RF plus eMRAM is another solution, and we will provide 18 FDS as well. We are working with library companies and our IP partners to build IoT platform solution on top of these process nodes. Let me summarize on our connectivity solution. For mobile and connectivity, we provide silicon-proven RF process technology, we also provide RF millimeter wave support up to 110 gigahertz. We provide special technology, package technology, antenna array technology, so that customers can use the package as the antenna solution.
For automotive, we already provide AEC-Q200 grade 1 from 28 to 8 nanometer, we will provide 7 nanometer solution by fourth quarter of this year. IPs are also ready to meet the ISO 26262 certification. We can provide automotive service package as well. For IoT, we provide FD-SOI technology with RF and MM capability, our proprietary security IP and total solution on top of the IPs. Our IoT reference platform can be used for the easy design, we have strong IP team who can customize our IPs if they need different solution. I talked about 3 key items of Samsung Foundry. One thing is our business update, second thing is our technology roadmap update. The key update on technology roadmap, just summarizing it again. Last year, we announced 4 LPP as the Gate-All-Around.
We changed the name to 3LPP, we pulled in the schedule by 2 quarters because our R&D center found out the solution earlier than we originally planned. That is the key difference. Four nanometer moved on to FinFET process family because we further found out process migration from the FinFET technology to 4 nanometer. I'm very proud that we can provide 7 to 4 nanometer as FinFET family. We can provide 3GAA, we're the 1st solution, Gate-All-Around solution. I'm very proud of it. I just talked about the platform solution for AI HPC and the connectivity as an exemplar. We can provide those kind of solutions for automotive, for mobile, for consumer as well. I think that 1 change of trend is from just a pure foundry provider, the foundry company should provide turnkey solution. The key is package solution, agent solution, IP solution.
We are, I think, ready to provide on 14 nanometer and 10 nanometer. We are building those solutions on 7 nanometer. I hope you found out Samsung Foundry could become a strongest foundry provider. Thank you very much for your attention. Please have a good day. After my discussion, there will be LED and the memory discussion, please enjoy our discussion. Thank you very much. Please give me any questions, but, today please confine our discussion on our process technology and the platform solution, okay?
Yeah. Hi, thanks for presentation. First question is on Gate-All-Around. Could you elaborate a little bit on the key process challenges in bringing Gate-All-Around away from FinFET, and the timeline for risk production at this stage? Second solution, just rebounding on what you just said, you added it to packaging potentially being a key differentiating factor, effectively only of a foundry which has this internally in terms of advanced technology, apart from wafer level packaging and 3D would be actually Intel. How do you look at packaging and packaging solutions as differentiating factor to win business, especially in AI, vis-a-vis your foundry peers? Thank you.
Okay. Sorry, can you repeat your very first question?
Very first question is, regarding Gate-All-Around,
Yeah
what are the key process challenges in forming the Gate-All-Around versus normal FinFET structure? With that in mind, when are you targeting risk production?
Okay. The reason why we developed Gate-All-Around, first of all, the purpose is that from 14 nanometer, 10 nanometer, 7 nanometer, even though we could get performance enhancement, size reduction, the fundamental limit was we could not scale supply voltage. Supply voltage was always 0.75 volt. That was physical limit, fundamental limit, not just for Samsung. It is the fundamental limit of FinFET technology itself. We pursued a very different solution. Gate-All-Around, the difference of Gate-All-Around is FinFET, the channel is surrounded by gate in two dimensions. Left side and right side. Gate-All-Around is very different. Gate-All-Around surrounds the channel in four dimensions. That could enhance the channel mobility, I think, the current capability. As you saw in the picture, we build MBCFET using three stacked Gate-All-Around devices. We could enhance the current capability of those devices.
I'm not a guy on process node, I think the Gate-All-Around is not easy to develop. There will be a lot of defect issue using Gate-All-Around, reliability issue as well. We already demonstrated some silicon performance on papers, we also demonstrated on our press release as well. I'm very comfortable that Samsung Foundry will be the only provider of Gate-All-Around technology. In terms of mass production, it highly depends on customers' adoption of Gate-All-Around. We started talking with tier 1 customers, the timeline of mass production is not decided yet. For the package, your question was, our package solution for AI, HPC, those areas. Okay.
Our package solution for AI, HPC, as I mentioned in previous slides, our key differentiating factor on 2.5D is we can provide turnkey solution, because we could control the memories, we could control the ASIC, we could control 2.5D package altogether. This is not just within the Foundry Business Unit. We will work with our OSAT companies . We already have collaboration with OSAT companies to provide this turnkey solution. The key is our customers will not see the barrier between OSAT companies , us as service providers. We will be in charge of everything. To build those solutions, we built our own proprietary debugging solution, because without debugging solution, we cannot say we are the turnkey solution provider. We will let customers free from the debugging issue. We will work on it. We already have some experience.
The reason why we could provide this experience was we worked as IDM company. For mobiles, for example, when we provided memory and AP together, we worked together, we have good design methodology team who can find out the debugging solution. It can be used for 2.5D package as well. Okay. Thank you. Okay. Yeah.
One question on the slide, the IoT and FD-SOI.
Yeah.
What is the benefit of fully depleting silicon on insulator? Why was the technology reflected from the IoT first? As far as I know, generally, the benefit is high performance and low power, right?
What is the special benefits of those technologies at FD-SOI?
Yeah, I think you are right, high performance and low power. The advantage of FD-SOI is when you use FD-SOI, the difference from bulk CMOS is you can control the body bias. By controlling the body bias, you can change the leakage current significantly. By changing the body bias by, for example, one volt, the leakage power can be reduced by one tenth of the original leakage. Dynamic power is also controlled dynamically. You can control the dynamic power by adjusting the adaptive body bias. I think that is the key technology difference. Another thing is FD-SOI has better quality control compared to bulk CMOS. Quality control is better if you use FD-SOI. That's why some tier 1 customers early adopted FD-SOI technology. NXP was one example. NXP already announced press release a lot of times.
They were, I think, the first mass producer of using FD-SOI, and they are really enjoying the benefit of this one, and they like to stick to 28 FD-SOI and move to 18 FD-SOI. We also have multiple Japanese customers who like to use our FD-SOI technology, 28 FD-SOI. They already started designing. They are close to mass producing. I believe this technology can be expanded to tier 2, tier 3 customer, consumers as well. We have multiple, for example, printer companies who use 28 FD-SOI. On consumer side, 28 FD-SOI will be cheaper version, low power version, so it will be very popular. Thank you.
Okay.
All right. You mentioned earlier about building your customer base. You saw increase from China in a significant portion. Could you elaborate on what kind of applications are these for? My second question is also on cryptocurrency. What is your exposure to cryptocurrency? The third question is on GAA. Could you elaborate on why is this a breakthrough technology and what kind of applications do you see this for?
The third question was?
GAA.
GAA?
Yeah.
You mean.
Yes
gate-all-around?
Yes, that's right. Yeah.
Okay. First question was, what kind of area we expand the customer base in China? Our key area is mobile, consumer, automotive, HPC, and network. I think we expand in all those areas in Chinese market. I think one of the key difference in Chinese market is if they feel our process technology is attractive, they take it very quickly. Chinese customers move very quickly, and that's why we could expand, in China in a very fast way. I could not show the detail of the companies, but we are really expanding in Chinese market. Okay. Second thing is cryptocurrency. We work with numerous cryptocurrency, especially Bitcoin, Ethereum, and the Litecoin companies. I think the demand is still quite strong.
In a long term, we are carefully investigating how long it will go and how significantly the business will change from our customers and the said companies, mining companies as well. We are still studying on it. This year, we have pretty good business on cryptocurrency area. Okay.
I have a follow-up question on.
Gate-all-around?
On China. Can you help me understand what has changed now such that in this year you are seeing an increase in terms of the customers from China?
Sorry, I didn't-
What has changed now in terms of seeing that increase in terms of orders and customers from China?
Sorry, I didn't-
Increase in orders from China, what has changed? Previously, the orders from China was relatively muted, if I understand correctly. Now that has changed, and you are seeing a substantial increase, and you've doubled your customer base. What has really changed in this year?
I think one of the reason is since we prepared for the turnkey solution, total solution, originally, we had some limited capability in terms of the IP, in terms of the design solution, now we strengthen it significantly. I think that's one reason. Second reason is, I think since Chinese semiconductor business is very good. That is another reason, I think.
Okay.
How was it?
Yeah. Just a question on returns. It seems as if since 2005 you did spend quite a bit of capital deploying to the mobile internet era. By looking at the returns on those investments, were okay, weren't great. As we move into this IoT, AI, and autonomous, it's a step change. Maybe it's a more level playing field against the number 1 player. Do you feel that returns on those investments are different this time around as we move into this data era?
Yeah. We are enhancing the return on our investment, and it depends on the business volume. First of all, we try to increase our business volume, and I think the return on investment will follow. For more details, since we are going to have IR in one month later, I appreciate you to join our IR. We are going to talk about the financial status. Okay. Yep.
Yeah. Simon from Bank of America Merrill Lynch. Just a quick question on the EUV area. Do you really believe the EUV needed for the application process of chip, maybe from 2020? Without EUV, just multi-patterning is not enough for 2020? Your EUV fab for 2020, we guess, exclusively EUV fab, that fab will be only for EUV process or for other lithography process as well?
Our S3 fab. First, let me answer to your last question. Our S3 fab was the first fab to provide EUV, but S3 fab has mixture of full EUV and the non-EUV, like 10 nanometer and others. I mentioned about a pure EUV fab, it will provide the EUV process seven nanometer and beyond as well. That is my answer to you. Your first question is very hard to answer Whether you really need EUV only, I believe so, because EUV only can provide much less cost, because if you go with EUV only, you can reduce the number of masks significantly. It is more than just simple 2 digit compared to non-EUV version. Eventually, for top tier and advanced node, I believe that you should move on to full EUV like Samsung provides. Okay.
I think the time already passed. Thank you very much for your interest. Okay, thanks so much.
Thank you, Mr. Lee.
Good morning, ladies and gentlemen. This is YS Choi, working for Samsung OLED Division, this is my great honor to present Samsung OLED display for the automotive market. Today's my presentation consists of three parts. Basically, briefly touch about the market dynamics of the small and medium display market. Secondly, why Samsung OLED could be a good solution for the automotive display market. Finally, I will talk about our Samsung Display's strategy for the automotive display market. This is the snapshot of the small and medium, overall small and medium display market. Basically, still a mobile phone. This is the talking about the just production area size in terms of production area, below 13 inches. Still, the smartphone and mobile phone market consists of more than 60% of the total production.
Secondly, the tablet, 10, 8, 9, 10, 12 inches tablet consists of the second biggest market, but definitely this market is not growing. It's decreasing actually. Interestingly, automotive market overall, this one is the third biggest market in terms of the small and medium display, but CAGR annual growth rate is almost a 2 digit, grow very close to the 2 digit. Basically, automotive display market has the three or four different applications. Basically, CID, center information display for the instrument clusters, and rear seat entertainment. Interestingly, head-up display, and e-mirror display, these kind of new applications coming in to the display market. This one is show or showing about automotive display market with the applications as I introduced earlier.
Still more than 90% of the display market dedicated for the CID and then cluster market. As I said, interestingly, new kind of applications, head-up display, rear seat entertainment, CMS, means the camera monitoring system for the display, is booming up now. If we talk about the display technology, as everybody knows that, this TFT display LCD has the theoretically almost 100% of the market. But according to the IHS, they say, two years later, maybe 1 million. Another, four years later, maybe 3 million OLED display will be in the market. This is the current market forecast and expectation for the automotive display market.
I just want as an OLED display maker, I want to more talk about the why future automotive display market need more OLED displays in the market. First, as everybody knows that the OLED display is much better optical performance than LCD displays. People, the normal consumers already experienced a very, very good display quality through their smartphone displays. Previously, when we talk about the just the display and then in the car, just simply we are talking about the size. Okay, you have a five-inch display and seven-inch display. Oh, you guys have a very good displays or 10 inches or something. Now people are talking about their quality. Every six months and one year, normal consumers see the very cutting-edge smartphones with the very beautiful displays.
They are frustrated when they go into the car buying shops, and the display is just outdated. That's the biggest challenge for our customers, OEM customers and the tier 1 customers. That's why they have a very aggressive discussions with us how they can close the gap. Between the consumer experience versus their consumers already experienced the kind of experience to their car experience. Second movement that we can address is that the evolution of the EV cars. Especially after the Dieselgate, all of the OEM customers are very aggressively trying to introduce the EV cars in the market. If we just talk about just one display and two display in the car with a combustion engine, the advantage of the power savings of the OLED display, yeah, that's nothing.
When we talk about the EV cars with the many displays under autonomous driving environment, this could be a very good differentiating factor going forward. The other point that we want to discuss is the OLEDs has a very thin and light form factor, basically. These car OEM designers kind of, always they have some kind of difficulties to just typically behind these display systems, many cables and air conditioning systems, and they have a very limited space. By adopting OLED displays , they can have more design flexibility. These new factors having together, we may believe that the new kind of, this kind of market expectation forecast will be changed if we think about this kind of new factor together.
Okay, for those of you who are not that familiar with the difference between OLED display and LCD display, I'm going to give you a brief introduction about the basic differences. Looks similar. This smartphone with OLED display, smartphone with LCD display looks very similar, but actual structure and then driving method is totally different displays. As you can see, OLED display has very simple structure, just two bottom glass and top glass, and then that's it. No liquid crystal in between, and then no backlight unit. Very mechanically, very thin form factor. Another interesting point is that OLED display is driven by the organic materials, which self-emitting the lights. LCD display case, always there is some additional help from the backlight unit.
In order to express the color, they also need this kind of color filter on top of that. Very, very complex structure. Another interesting point of the power efficiency. Regardless of the contents, LCD has to turn on 100% of their backlight unit. But if we just show the 50% of the display images, OLED case, we just turn on just 50%. If we need just 5% of the contents, we just need to turn on the 5% of the power. That's why we can significantly save overall powers of the OLED display versus LCD. As I said, it's very thin and light form factor. This is this kind of simulation of the display module. 12-inch, the 12.3-inch is very, very popular size for the cluster display.
If you compare the overall size and weight, OLED cases has just over 100 grams versus LCD case has almost 400 grams. If you think about it, just difference is, 300 grams, that's nothing, for the cars are almost 1,000 or 500 kilograms and then something. If we think about the EV cars and then times of six or seven displays, that could be a matter going forward. Next for the why we think that we can address why is OLED display to entertain in the automotive market. Before we set up our strategy, we just sat back and then thought about our future plans, how we can address customers and consumers' point with our OLED displays. From consumers' point of view, we thought about our kind of the, what's the difference of our displays.
First of all, as I said, consumers already experience a very nice displays with more than 10, almost 10 years with the smartphone displays. We believe that they want to have a seamless experience between mobile and automotive. Like mirroring, we can enjoy the mirrorings with our TVs and the mobile phones. Why not in the 5G era? Why not in the autonomous connected car environment? Haptic and the fingerprint on display, those kind of new functions they also want to enjoy. Definitely, they want to have a beautiful and then elegant interior design with the new displays. At the same time, the car. Car should be very. We always thought about the concern about fuel efficiency, also concerned about the safety, because of the car is always moving.
This is the summary of why and how we can address these kind of consumer's new wants and needs. Basically, as I said, the OLED display can give the very high optical performance with a wider color gamut and a higher contrast ratio. OLED can give them a very totally different form factor and then design possibility for the car design. At the same time, as I said, OLED can give a very benefit of the power efficiency, also OLED display can give them safety, enhancing overall safety. Finally, we can provide more values, additional values from the smartphone and the mobile experience to the automotive. Basically, high picture quality.
Because of the OLED has very high contrast ratio, we can give them a very true black experience versus LCD cases, some kind of the light leakages is inevitable because of the technology itself. When you think about some kind of a dark environment and the dark and rainy driving environment, because of the higher contrast ratio, we can easily recognize the object with the dark environment versus sometimes it is not easy, the kind of the environment with OLED or LCD displays. Additional point of the OLED display, beauty of the OLED display has the very wider color gamut. Coupled with the high contrast ratio, OLED can give very easily recognize the object in the display with the same brightness level or much better recognition versus other LCD displays. Interestingly, we can give another additional optical values. They have some kind of experiment.
It is known to be that hazardous blue light is very, very harmful for the human eyes. That's the experiment of the mouse. After 24 hours of exposing the RGB light, 20% of the retina cells survived. Basically, hazardous blue light is very, very harmful for the human eyes and then the animals. When we compare with the LCD display and the OLED display, thankfully, OLED has a very limited, very much less harmful blue light versus LCD. Which means, if you think about the automotive car interior design environment, when we have a very long drive with a dark environment, in the dark we can easily feel some kind of eye fatigue. OLED case, we can significantly reduce that kind of the issues. Another interesting point is that this one.
It is known to be that 6% of the human capital of the man, not the woman, 6% of the men are color weakness people. They have difficulties to recognize between the colors red and green. OLED case, we can control each pixel by pixel. That means we can intentionally increase the red pixel performance. In that case, even color weakness people can easily recognize these kind of signals through the OLED displays. That's another interesting optical values can give consumers and then customers. Next one is talking about the mechanical difference, basically. This one is fundamental difference between LCD and OLED, because OLED has very simple form factor and then even flexible substrate, flexible display is possible. We can give, not easily recognizable, this one is actually a flexible display in the S curve. We call it S curve, double curved display.
We can implement this kind of just curvy designs with a flexible display easily. This could be a very, very wow effect for the consumers and then OEM customers. OLED case, no liquid crystal, as I said that, no backlight unit, and then no color filter, and easily bend, easily make some curvy designs possible. This one is another kind of the example. For rear seat entertainment case, because of the thinner light form factor, we can give them different kind of design ideas. Right-hand side, this one is the demo, was introduced last week, the SID show in L.A. This one is basically rollable displays based on our flexible display. This one can be rolled up and rolled down. Basically, this one is a 14-inch display.
If we use the two-third of the display, it can be 11 inches, 10 inches, 11 inches of display. If we roll up a little bit more, we can use this one as a nine inches, eight or nine inches of display. At the same time, we can use this kind of additional space behind the display. Let me show you some actual videos of this display. One-third, two-third, and the full displays, because of the flexible substrate, we can roll up and roll down. We believe that this kind of new form factor can give a lot of new design ideas for the OEMs. This one, we can show you how these displays can be rolled up and rolled down. Finally I'm talking about the electrical difference. As I said that OLED is the content-based driving.
That means we call it Own Pixel Ratio. When we need just about the dark, close to the menus, we need just 7% of the power. Regardless of the contents, LCD has always 100% power should be turned on. Just the audio case, we need the 11% of the power needs to now turn on. Video, just pictures, is 17%, camera monitoring system, just around 30% of the power. The highest power needed content is about the navigation map, and that's only just more than 50% of the power. All things together, we can simulate a user profile. Usually, it has the 30% of we need, the 30% of the OPR, Own Pixel Ratio. If we compare the overall power consumption versus 12 inch, simulated at 600 nit, OLED case has just less than 50% of the LCD power.
If we think about just one display in the combustion engine, that kind of the power saving is minimized. If we think about the EV cars, with the connected environment, with a more autonomous driving environment, which requires six displays and eight displays and 10 displays in the car, this could be a matter. This one is a simulation of the six displays with the, yeah, eight inch, 10 inch, 12 inch, power displays itself, power consumption is 26 watt, versus LCD is just 46 watt. The module difference is just one kilogram versus 2.5 kilograms. This one is only just 1.5 kilograms difference. With this kind of flexibility, we can give power, the OEM car designers power management system, we can give them more flexibility and more rooms to install more kind of electronic devices.
When we think about the connected cars and autonomous driving, many, many sensors around the cars, in that case, the overall power consumption should be multiplied going forward. In that case, if our customers use OLED displays, we can save additional lead acid battery. All things together, we can give them, for example, 20 kilograms and 30 kilograms more kind of savings for the car designs. Those kind of benefit can combine, if we combine those kind of benefits, overall benefit will be much better, much bigger than typical one or two display comparisons. This one is another kind of a new category and new applications. Now, some customers, we are thinking about talking about eliminating physical side mirror.
If we eliminate physical side mirror, we can have a very, very simple, very small form factor with the camera modules. They say sometimes 3% or 5% or more fuel efficiency with these kind of new designs. By eliminating the physical mirror, they have to install the displays inside the cars. Just below the A-pillar, there is a very limited space. OLED is a very, very thin form factor. That's why they can easily implement, install the OLED display in the car, versus LCD case, there's a basically much thicker form factor. A lot of heat is coming from the backlight unit. They also need a heat management, heat sink, heat management system with the OLED or LCD cases.
It's not easy to install those kind of the camera monitoring system with the e-mirror in the car. Another basic difference is that, this one, we took the videos and played on the one with the LCD display and the one with OLED display at minus 20 degree. In Singapore, in the very hot, humid area, we don't need to worry about minus 20 or minus 30 degrees. This one is critical in Norway, Finland, Northern Canada, and Michigan area. Minus 20 and minus 30 degrees in their winter times. Because LCD case, liquid crystal doesn't move fast at very low temperature. This one is very critical moment. 1.1 second, there could be almost 3 meters of difference of the recognizing the object that's passing by me. OLED case is, no response time issue.
Also, we can increase our safety with the OLED displays in the car. Next one is another interesting new concept by utilizing OLED display. As I said, some car accidents happen when drivers try to change their settings in their audios and videos and some CID controls, and when they lose their eyesight. Oh, that happens. That's why some kind of customers. We have very concept stage to what about installing the display on top of the steering wheel? Typically, display, just below the steering wheel, there should be airbag should coming out. Because of that, some size, the weight limit is. That's why LCD cannot make even happen. Even our rigid glass-based OLED cannot meet their requirement, usually about 40 grams and 50 grams. Only the flexible display can meet their requirement.
If you think about when actually some display on top of the steering wheel, head impact happens, it shouldn't be breakable, because some kind of glass breakage and then can damage to human body. That's why we thought about, "Oh, what about the plastic cover window on top of the display?" We make some kind of the demonstration. Let me show you the demonstration. Interesting concept. Basically, display is plastic, and the cover window is plastic, so basically not breakable. Very light form factor. That's why we can provide additional innovative display concept for the automotive display. Additional values. Mr. Lee also talk about some kind of haptic feedback with our OLED display. The reason why OLED display can easily implement is that, this one is the mechanical form factor.
As I said, the LCD has a liquid crystal in between, and the backlight unit, and a very thick form factor, versus our rigid display and the flexible display is a very thin and light form factor. That means if we apply the haptics and fingerprint sensors behind the display, it can easily transport that kind of signals to the sensors. That's why basically, haptic feedback and fingerprint on display, we can easily implement with the OLED displays. Until now, I have presented why OLED display can address new demand and innovative concept to the automotive market. Still, OLED display has some kind of intrinsic challenges. This one, basically, OLED display is driven by the organic material, which means over the time, with the higher brightness, organic material can be degraded over time.
Until now, until recently, we had some kind of brightness versus lifetime issue, especially for the automotive requirement. We recently overcame that kind of basic automotive reliability requirement with the acceptable brightness and acceptable reliability. Still, some customers, and for the certain usage of their cases, especially for the bright sunlight, direct exposure to the bright sunlight, then they want to have higher brightness. We are working on it. For the most of the cases, we overcame the kind of basic requirement, but some certain usage cases, they pushed us to increase the more brightness. If we just simply increase the brightness, overall lifetime can be degraded. We are working on it to achieve the higher, to meet all those kind of requirements now. Another point is that, image burning.
As I said, the organic material is organic. If we turn on the simple, certain images for a long time, it can be degraded, and then it can show the less brightness. This, in automotive cases, we tested that this one, over thousands of hours, if we turn on the same images without changing the fixed images, then this area become, looks like this kind of very, a little bit less brightness, bright area. That means image burning issue. This one is actually happening for the think about that. 3,000 hours, 5,000 hours without changing your manual. Theory, in reality, that's not possible, but automotive customers always ask us to be perfect. We are working on to minimize. We already developed that kind of minimizing algorithm, our technology to be shown like this.
In real cases, we don't have any concerns and the OEM customer's already okay, they are still pushing us to upgrade more, to avoid that kind of the potential issues. These are the kind of our technical challenges. Still, we are working on it. Another point is that this one shows what mobile phone display market, sales revenue or market share. This year, we expect to have more than 50% of the market share in terms of the dollar value. It took us more than 10 years in the smartphone and the mobile phone market. This one is challenged in terms of the commercialization of our OLED display into the automotive market. Last year, we introduced our OLED display in the real car, according to the technology adoption life cycle, it is very early stage.
We need to closely work with our customers, we know that that's the long way to go, especially in the automotive market, we are going to support with a very innovative OLED display going forward in the automotive market. That's still a long way to go, we will go there. This is the summary of our overall strategy for the automotive display market. We think that we have created OLED display market in the smartphone. Our mass production of the OLED display in the actual usage happens 11 years ago. Cumulative, we have expected to ship 1.8 billion pieces of this OLED displays in the market. This one is actual product we have worked with our customers.
We introduced a very innovative technology with the OLED display for the mobile market for the past 10 years. We have proven track record. Why Samsung OLED in the market? We have proven in the market more than 10 years. Also believe that we have created the OLED display market with the smartphone and the small size display market. We want to continue this kind of the proven track record, our history in the mobile. No, I'm sorry. In the automotive display market. Later this year, we think that there is another kind of introduction of our OLED display in the market. Later, as I said, that unbreakable and rollable and transparent and the new additional values on the OLED display, we will differentiate with this kind of innovations into the automotive display market.
That's our strategy in going forward. Thank you for your attention. This is end of my presentation. If there's any questions? Yep.
Thank you very much for the presentation. You talk a lot about the technological advantages. Can you talk a little about the cost advantage? Where that current situation right now, the cost between the OLED and LCD from the OEM perspective?
It is not easy to talk about just simply cost focused, and that cost always involved with the volumes and overall project volume and blah, blah. I think that you are talking about when OLED pricing can be matched with LCD displays. That's the basic questions. We think that if we deliver the very acceptable and the meaningful values to the customers, the price will be decided by market, not by us. I think that we are closely working with our customers to have a right value point with the right pricing. That's our strategy to work with our customers. Okay.
Thank you. Just to continue a little bit on the cost side. From a manufacturing perspective, is it more efficient to dedicate lines potentially to automotive OLED within your Gen 6 capacity? Is mix and match actually absolutely fine vis-a-vis mobile in particular? Secondly, could you elaborate a little bit on OLED versus micro LED, especially for applications like HUD displays? Thank you.
The market and the commercial dynamics is totally different from the mobile phones, which has big volumes, and then OLED cases has a very design-oriented, maybe need more custom displays in the smaller fabs. Basically, you already answered the question. We have a basic right manufacturing fab with the right kind of applications. Mix and match is basically our strategy for the manufacturing. For the micro OLED, micro LED displays, and then we also saw some very nice demos and then prototypes from our competition. My personal opinion is that it takes time. It takes time to be used in the automotive market, we will see how they perform in the market. Basically, now we are just focusing on the OLED display for the automotive.
Any other question? We'll end up the second session here. Thank you, Mr. Choi. We'll have about 15 minutes break. We are about to start the last presentation for today. Please, no photos during the speech, and only the questions related to topic. Okay? Please join me in welcoming Mr. Chun to stage.
Can you hear me? Good morning in Singapore time. Good afternoon in Korean time. Very nice to meet you. My name is Sewon Chun. I'm very happy to be here today to make a presentation about data center era. My name is Sewon Chun, but my initial, SW Chun. When I dispatch to the U.S., my colleague named me, "Your S, Sean is better," and Steve, otherwise, named me Steve. No. They are saying, one day suddenly, Sancho is right nickname to S. Before I join Samsung, SW stands for stupid and wide. After joining Samsung, I try to smarter and wiser. Today, I have studied about the data center, and this may help you to understand the data center market and how much impact this data center to memory business and IT industry. Our start was this.
Do you know who is this guy, this gentleman? Oh, you don't know this gentleman. This gentleman, Victor Hugo. What he said, "The cloud, the only bird that never sleep." My topic today is the cloud. In the modern age, cloud are where data never sleep. Continuously fueled by more and more memory. This cloud is very, very important in our memory business. Let me start here of this Victor Hugo's one quote, "The cloud, the only bird that never sleep." Okay. Let me start with this slide. This slide shows the memory revenue trend among all semiconductor industry. When you look at this PC era, there is a lot of up and down, up and down, but still keep growing. Only in that time frame, only memory revenue portion is only 15%. 15% only.
You can see here, after launching smartphone, then mobile portion is getting important for memory business among all semiconductor industry. Once smartphone launched to the market, after that, Android OS tried to catch up for the leading company for the smartphone, then smartphone business has been booming. That impact to mobile total memory business, we call it mobile era. Mobile phone market right now, long time before, in this time frame, skyrocketing. Penetration rate and the growth of smartphone shipment number has been skyrocketed that time frame. However, smartphone shipment number is mitigated nowadays. Fortunately, since 2016, data center market is booming. Means that the world data center market boom is, the server market is booming. Server market impact to memory business a lot in total semiconductor industry.
In the process, the memory, the role has been changed From just a passive component to being at the core of the IT revolution. Memory is going to be the very important role from now on for data center era. That's why I define, from now on, the data center era is just started. As you can see of this chart, the cloud industry, as you can see, right now we are here, the rapidly growing since 2017 and 2025, KRW 400 billion. The main role of this data center revenue growth is data center cloud. Cloud is going to be a main role of this public and hybrid cloud revenue impact. Okay, let me switch the topic. Data center is the core of IT industry, as you can see here. Processing, storing, seamlessly connected for multiple application.
As you can see, when the connected auto, connected car started, and the data generated by huge amount generated by this application. IoT, Industry 4.1, gaming, mobile, PC, to have this multiple connection, 5G infrastructure is required. You know why the 5G is very important for this data center and this semiconductor industry? 5G provide hyper-connectivity. When you look back to the 4G era, when the 4G launched to the market, that timeframe, nobody knows 4G, how much impact to this semiconductor industry. In the end, we knew how much impact to 4G is going to the impact. We already have seen the impact. 5G, the one very important message is hyper-connectivity. Right now, 4G infrastructure only connected PC, tablet, smartphone. Very limited application is connected. Once 5G infrastructure deployed to the market, after that, many IT gadget connected.
You call it IoT, Internet of Things. Many IT gadget connected through 5G. 5G is going to be big impact to this semiconductor industry. Okay. Nowadays, data center business is booming, as you remember. Amazon AWS is leading this momentum. Why on-premise to public cloud? Why the company is trying to move their data or their infrastructure to public cloud? There is a two reason. Customer point of view and data center competition dynamics. First of all, I would like to emphasize customer side. Why the dotcom company or Fortune 100 company try to have IaaS services from data center guy? Definitely, they have some benefit. First one is TCO benefit.
As you can see, all the companies, big company and small company, they try to have their own data center, but cost-wise, it's too much, big money to maintain, to sustain, to service that one is very difficult. They realize that. Also, even small company, we call it SMB company, they are saying, "To build my own data center, it cost a lot." That's why they are looking for some efficient way to have data center by their own, but they found out that TCO benefit a lot. Data center provide that company's economy scale, they provide. Hyperscale size, optimizing, infrastructure standardization allow lower cost. That is the first priority why on-premise guy is moving to public cloud.
Also, the other benefit is that the guy who try to build their own data center, but they cannot have the most advanced technology or the most advanced computing architecture. Data center has a lot of capability to deploy their new technology, their new services, their new infrastructure, their new platforms immediately. Companies can adapt. That is the reason nowadays why the other one is the competition dynamics for the data center guys. Amazon is leading this industry at the moment, and as you can see of this slide, number of services is skyrocketing. CAGR of 45% here. Amazon leading this industry a lot because they have a lot of capability, they have a lot of resources. How about the other guys? Microsoft and Google, and nowadays, China data center guy. We call it BAT, Baidu, Alibaba, Tencent.
That is the reason why on-prem guys to public cloud. Based on my understanding, data center is just a study. When I look back 2016 and one year before, 2015, memory demand has been down significantly. Data center guys told us that why. I asked them, "Why?" Memory demand has been down significantly in 2015. They are saying, "We did optimization." Means that hyperscalers, they build their whole computing servers in their own, scale out, and they optimize that one by adapting the most advanced software technique. That time, memory demand has been down. I study about the data center since year 2015. What is the memory consumption a lot from this data center? Maybe a lot of memory consumption is going to be in data center.
What they are doing before 2016, they try to have the most advanced, they develop the most advanced technology, most advanced computing architecture. They want to try to find out what kind of services to deploy to the market, then we will hit the market, we earn the money a lot from our customer services. They realize that we need to build our own data center by adapting the most advanced technology. IaaS, we call it IaaS, Infrastructure as a Service. PaaS, Platform as a Service. They build many kind of services on their own, they realize that, "Oh, I have level resources to support this kind of stuff." They deploy IaaS and PaaS to the public, they promote that IaaS and PaaS to their customer start interesting to join data center.
All the customers, their customers is very nervous about to move their data, own data, to public cloud. They are thinking, "Okay, I understand that the benefit what we earn by using data center." However, they have to keep their own secret information in their own. That's why there is a 2 track for data center deployment business. Is one track is on-premise, strongly rely on all the information, rely on the data center. Other one is the big company, they want to keep their own data in their own, normal data has to transport to public cloud. We call it hybrid cloud. Right now we are very early stage of data center business at this moment. What is going to be the next?
Still very high growth potential for the data center from dotcom and Fortune 100, Fortune 500 company try to move their data to public cloud. Amazon, you may heard about Amazon's Snowmobile. They are providing some Snowmobile services to their customer. They visit their customer, they download all the information from their server to Snowmobile, they dump that on their Amazon data center. That things happen. A lot of momentum to grow. On top of that, the more important thing, I already mentioned before, 5G. 5G may impact this data center a lot. I already mentioned before, 5G infrastructure provide hyperconnectivity, means that all device connected to each other, they generate a bunch of amount of data has been generated by those gadget. Telco company point of view, they have some new services they're looking for.
Meta trade] you already heard about in the U.S. They have their own business like Amazon, like Microsoft. This 5G infrastructure, the base station, the core part and edge part is going to be, because 5G deploy to the market and the real-time communication is very critical. When you drive autonomous car in the road, then there is no real-time communication, in that case, human life is going to be terminated because of the miscommunication or misinformation, and delay of information. Very critical impact to the business. Lastly, autonomous car company. Autonomous cars era started, I don't know when the autonomous car era, level 2, is started. I don't know, actually. If once that autonomous car started, bunch of information, very critical human being life, critical information, is going to be treated in real-time.
Auto manufacturer, they also try to have their own data center to have their own services to their customer. They don't want to rely on big data center. If they give their data to data center, then that information has to be analyzed by them, then they could make some other services for the data, other services for the automotive market. They try to keep their own data center by utilizing their own data center, then they want to service this to the market for their customer. I could darely say today, data center market is just a study. From here, as I will touch more detail about the data center, what is the biggest enabler to grow this data center market for the memory point of view? Any other questions? After my presentation, you guys need to go to have a lunch.
When I was in military service, when I was a student in university, I don't like professor, longer lecture just before lunchtime. I already spent 20 minute, I have bunch of slide. I prepared 20 slide here. Only I have 27 minute, 50 second. Okay, let me start here. Data center beyond 2018. The biggest enabler of data center is AI and in-memory. You can recognize in-memory means that in-memory database, the leading company, SAP HANA. 5G already mentioned before, 5G provide hyperconnectivity and high bandwidth that enables IoT easily. Means that, in other word, bunch of information, structured, unstructured data, is gathering to some places. That is data center.
Data center handle this data as a value, immediately they have to treat this one, then they have to response this information to their customer, their client, to give them some other opportunities. As you can see here, data created and stored exponentially grows. That is, the data is very important in this data center guy. Let me start AI. When I firstly listen this AI is going to be the very good application to consume more memory. When you look back the long time before, the AI machine learning algorithm already exist. Why the long time before, AI machine learning is not popular to all of this audience today? Even I didn't know about the AI back to 2015. My colleague traveled to the U.S. 2015, I traveled to the U.S., May timeframe, together, what the memory demand is going to be.
We promote SSD to the market aggressively. October timeframe, I need to go to the U.S. together for surveying the market with our the engineer, and one engineer, after travel, he call me. "Hey, Sewon. You know what happened? Compared to last May timeframe when you visit and when I visit this time from October, everybody talk about the AI and machine learning." I start study about AI and machine learning. That timeframe, nobody knows NVIDIA, what he's doing. They launched DGX-1 to the market, and what this DGX-1 is going to be. Our product planning guy, fortunately, they already knew about some kind of high bandwidth memory requires certain timeframe. That timeframe is just from our product planning to marketing, and I handle HBM memory. I realized that, wow, this DGX-1, the HBM, is critical role in this DGX-1 machine.
After that, I study about AI a lot, the machine learning. What is the machine learning? I have some question, because I'm not an engineer. Why? AI machine learning algorithm already existed. Why these things is just now popular, everybody talk about AI and machine learning? Before that timeframe, computing power limited, and so algorithm exists, but computing power is not enough to support that one. Machine learning is critical. One is the how fast they get the conclusion. What happened after that? What happened? CPU power is getting upgraded by applying the Moore's Law. One specific machine learning area engineer was looking for some other accelerator, some additional computing power they need.
When you look back long time before, when you buy a PC and you are the gamer, in that case, normal graphic engine inside of a CPU is not enough to support their requirement. The same story happened to this AI. All the engineers look for some other accelerator to enhance how could they reduce machine learning algorithm as quickly as possible. Efficiently, they could get the result based on that. Finally, they found that one GPGPU. Right now, that's why NVIDIA is. I need to buy NVIDIA stock long time before. Unfortunately, I didn't know at that timeframe. I said to my son, he's in the U.S., "When you earn the money, when you get the money, please buy some stock." Anyhow, just kidding. GPGPU is the main role at the moment.
This is revenue-wise, but in the future, this is very expensive solution, but very powerful. However, all the engineers in data center guys looking for some other alternative choice for replacing the expensive GPGPU solution. Finally, they got FPGA and ASIC. This FPGA and ASIC is not only for the machine learning area. Data center guy, the infrastructure, it consists of network fabric, computing fabric, and storage fabric. There is some limitation of network bandwidth. How could they enhance effectively network bandwidth? That is their core competition. That's why they are looking for some FPGA and ASIC solution. Google, ASIC, TPU, TensorFlow is kind of ASIC, and FPGA is programmable. Based on their workload, they could change, they could modify that one. FPGA business here and ASIC is, the FPGA facing more faster growth than ASIC. You can follow me?
My explanation is too difficult to understand? My parent was a teacher. My son is going to be a professor, but my parent was the elementary school teacher, but my parent insist me, "You have to be a professor in the university." But I failed. I ask my son, "You are my son, you have to be a professor." But once again, my son also failed. Okay. Just kidding. You can follow me? I have 20 minute right now from now on. I have, right now, half of them is finished. What is the revenue growth for this AI application, machine learning application? Cloud service providers are expanding AI service rapidly. Iris data center business is just starting, but this AI business is kind of infant stage. Infant stage. When you look at here, we are here.
Very small revenue, AI hardware revenue here. This one, $115 billion up until 2025. Among that cloud is the core of role of AI hardware. 40 times bigger than your 2017. AI service revenue point of view, all the area when we face every day. For example, healthcare, advertising, automotive, retail, energy, sports. You remember, in CES, Intel announced that some AI functionality, their Nervana system. The U.S., basketball and NFL, national football, is very popular in their country. When I was in the U.S., I was big fan of basketball. The football, they analyze all the guys moving and everything, they analyze for that. That's why AI service revenue is going to be exponentially gross, 17 here and 2025, 20 times revenue gross. All area, AI is aiming for all area.
Not for healthcare, not for retail, not for legal, logistic, gas, all area of the human beings touch, they are looking for. They are cultivating new business model for this AI. This AI is very important role for the data center guy. The value of data center data is changing, as you can see here. Time variable, analyzed data, and real-time data era is going to be started. We are here, bunch of amount of data exponentially grows here. Traditionally data era, data is just the data. From now on, the data is handled by real time. Data is the money. You cannot imagine how much important the data. This graph shows you the time value of analyzed data, but data itself is meaningless. Time value of analyzed data.
Data has to be analyzed to get the value, when the data analyze as a value, very early stage of that, the money is bigger than ever. Time pass by, that data value, analyzed data value, is going to be diminished. Mission-critical data, here, real-time data has to be handled immediately. We call it mission-critical data. That makes the money a lot. The value of data is changing. Very important. Fortunately, you are typing the keyboard right now. You are sending email to your colleague or your boss, "Oh, that guy is the data center, his screen, but he's right or wrong, please check it out." That data, wherever you generate the data, that goes to data center. Data center is watching you, surveillance camera. They are watching you. Whenever you go to the other place and then Google track you.
You have a cell phone here, smartphone here, they know S.W. Chun is in Singapore. George Orwell's "Nineteen Eighty-Four" is already started. The important thing is that 5G, I many times emphasize 5G and autonomous car era. 5G is very important for us. Autonomous car more important to us. Autonomous and 5G important to telecom company and data center guy, but autonomous car is especially important in memory side. Why? Autonomous car started. In the car, you are watching TV, you are watching video clips, whatever. The data is real-time transferred to your car, and car has to receive the most updated one. You're going to be dead because of the latency. When you go through the road, but there is a block, big block there, accident happened, but you don't know because you don't watch the outside of your car.
That data is immediately transferred to your car, and then there is a block, you have to stop. There is a latency issue, a little bit delay that information to the car, and then car crash. Very important. Also, the quality and performance is very critical for autonomous car era started for the memory point of view. Semiconductor point of view, very important. Why I am emphasizing data, the exponential growth of data, because the data has to be treated real-time, and otherwise, the data is not valuable. This is one of a good example, the SAP HANA-based one. This guy's goal is to convert 100% of SAP product to SAP HANA-based by 2025.
SAP HANA-based means that in-memory database system means that all the data dump to DRAM side, and that is handled in DRAM, because DRAM is the fastest latency they have and provide the high bandwidth, too. That's why they're looking for DRAM solution for that one. This in-memory is very memory DRAM intensive application, by 2025, their goal is all the product is going to be the SAP HANA in-memory DB system, means that they need more DRAM. Also, the calculation of the DRAM, and then that has to store to hard disk drive. Nowadays, they are using hard disk drive and SSD, based on the workload. In the end, latency, I already mentioned before, latency is very important to handle the valuable information and then may goes to HDD to SSD. The SSD portion is getting bigger and bigger, time pass by. Okay.
When this all phenomena, AI and in-memory database for the server area and data center area, I convert that one to our memory business. 2018, AI and in-memory database portion is very small at the moment. That's why I'm saying data center business is just started for the memory consumption point of view. 2023, AI and in-memory portion in server memory total revenue is going to be huge portion. That's why data center guy is focusing to AI and in-memory database. In other word, in-memory database, business intelligence require it. That's why they need more high capacity, faster latency, high performance DRAM solution they need. Also, they're looking for some very fast latency SSD they're looking for, because of this reason. I express in other expression for worldwide memory revenue by application.
We are here now, server portion is going to be here, and 2023, server is going to be the main role for memory business industry. Mobile smartphone area is the main driver for consuming the memory, but in the future, server is going to be the main role for consuming memory. This is the conclusion slide. You can follow all the stuffs today I explained to you. Sang-Hyo Kim , you follow me? Why I knew your name? I read your report. Let me drink water. It's very familiar, the name only. Nicolas, when I was sitting in the IR question, Nicolas, blah, blah. This is the first time to face-to-face meeting. I'm very happy to be here to make this kind of presentation. This make you happy. This presentation may help you to understand the memory business in the future from now on.
This is the conclusion page. Data center is at the center, right? I already explained public cloud, hybrid cloud, AR and VR, SAP in-memory database, AI, 5G, edge computing, IoT, all the information they generate, multiple application, they generate a bunch of data that goes to data center. The more important thing is that it's kind of a snowball effect. Long time before, one application demand has been down, and then that application is on impact. Based on my analysis, all demand is connected. Once one demand is picking up and that influence to the other area, and then kind of is a snowball effect. Memory demand is getting more and more. Negative side, down significantly and negatively impact to the market.
Based on today's presentation for the data center, the high growth potential AI and in-memory DB, two enabler already explained, if my guess is right, in that case, memory business is not so gloomy. I cannot say not so bright, but anyhow, high growth potential I could anticipate. All demand is connected. I would like to emphasize this message to you. Any other questions by here? Because next page, I will explain about the Samsung strategy. How could we react based on this big business opportunity. What is the Samsung strategy from now on? You can follow me. Thank you very much. Let me start the Samsung solution for data center era. This is the kind of snapshot. Conceptual data flow in data center. This is not the engineering session. This is, kind of IR session.
You need to understand, what is the data flow in data center? Where is the memory requirement? Where is the opportunity from memory business point of view? Here, I heard about the data scientist. Oh, yeah. One small joke. My son study in U.S., and he asked me a question, "Dad, I'm going to have a major in university. One major is better." Computer science is the future of your life. My son selected computer science. I listened that one from my son, data scientist. That was the very first time, data scientist. What data scientist doing? Data scientist, in archive, focus of a HDD, there are a bunch of big data exist. All the data generated only 10%, 5%, I don't know, but only valuable data has to be stored to archive.
These data scientist, they're looking for some business by data mining from that big data lake. They try to find some other business model based on that. They find out some good business model from the data archive. Data lake, we call it data lake. They mining the data, that is stored to be treated and handle the meaningless data to valuable data. The video clip, surveillance camera information, stored here, that is we call it unstructured data. That transform structured data to be treated in computing system here. You well know about the von Neumann computing architecture. I was an engineer for a long time before. I had been an engineer for 12 years in Samsung. I start my job in Samsung as a assembly engineer. That time, DIP, actually, package I did a lot.
I was process engineer. I moved to the product engineer, kind of analyzed the yield and how could we improve the yield, how could we make a qualification. I moved to the product planning. Finally, right now, I moved to the sales and marketing. That time, I taught the sales and marketing guy as a product planning guy, von Neumann computing architecture here. CPU requires memory, this is the DRAM and storage, SSD or HDD. I already mentioned before, accelerator here. Long time before, the accelerator is just the graphic card, but nowadays, graphic card has a different role for machine learning area. The accelerator here. Accelerator is goes to GPGPU and TensorFlow and FPGA, whatever. This computing system calculate that one, finally, data scientist find some services. Okay, this services is good for our business.
That deploy to the services from data center guy. What it means to us? Memory point of view, what it means to us? We have to provide high performance product and high density, low latency product with very high quality of product we have to provide to our customer. If there is no HBM, the machine learning, it takes a long time than before. Samsung is the first company to provide HBM to the market. Now we have a huge market share for that helps a lot to develop many kind of AI by utilizing AI and machine learning, they found some services to their customer. Memory role is getting critical in the future. What is the memory solution in Samsung? No matter what, we are not negotiating quality. That is the base of Samsung strategy for memory business.
Quality is the key because, the DRAM side, I already mentioned that in-memory database, all data has to be handled in DRAM side. There is one single bit error, system crash. Our customer looking for very high quality of DRAM, we have to provide. We have a responsibility to support that one. High performance point of view, Samsung developed HBM world first, we launch this product to the market. Nowadays, this demand has been skyrocketing at the moment. They are looking for more and more HBM memory. The issue is that HBM is expensive That's why our product planning guy recommended to me, okay, the GPGPU can use HBM if their workload is so critical and very important, has to be handled very quickly, then use HBM. Otherwise, for more raw bandwidth, GDDR6.
Otherwise, they have no money, and they have to look for some machine learning solutions, then we could provide GD5. Those three products, very low power solution has to be. The machine learning system, very power-consuming product. A lot of power consumed by AI machine learning. Already I mentioned 5G, whatever, low latency is very important for storage area as well. DRAM definitely, very fast latency we have, and we developed very low latency product. We call it Z-SSD. Very low latency. Also, I already mentioned, in-memory database requires high density, 256 gigabyte, 128 gigabyte, and 64 gigabyte, and SSD like this. We provide a I do not want to touch about detail and NF1, what is the NF1? Next to form factor to provide a very high capacity of SSD. Let me touch what is the Samsung’s technology leadership.
Our engineer understand the limitation of planar NAND. Then internally, we discuss about what is the next step if there is no migration for the planar and there is a technology barrier. In that case, what is the breakthrough technology long time before? Then our engineer take a risk. Okay, let's move to the V-NAND. That timeframe, even SW, stupid and wide, this guy doesn't trust the engineers, didn't trust our engineer's decision. Wow, we have to put the wafers with the planar NAND, then we earn the money. However, we have to take a risk. We are the leader in this industry. We have to breakthrough technology, we have to show to the market. Then we decide goes to the V-NAND here. Look back year 2010, that timeframe, D1x, Y, G, DRAM technology is getting difficult and more tough and tough.
I'm very proud of Samsung engineer. They show me this clear roadmap. I darely say Samsung, in memory business point of view, Samsung is the technology leadership. Still, we keep going. Now we have keep going that technology leadership continuously. What is our solution for the AI acceleration? I already mentioned many times HBM, GD6, and GD5. Memory performance equal AI performance. We keep providing a high-bandwidth memory, and we keep providing GD6, very high bandwidth memory, continuously. GD5, based on our customer's choice. We line up not only HBM, but also GD6 and GD5 to this AI acceleration. High-density solution. Intel Skylake, 2 CPU-based, 1 6-channel, 2 per channel, Samsung is only 1 company to provide over 3 terabyte.
Because we developed the 16 gigabit monolithic DDR4, world first, 1 and a half years before than our competition, and we strongly promote this product to the market. In that case, high capacity of SSD, we keep providing to our customer. World first product is this. Only time clock is stopped. I have to finish my presentation today. What is the very strong point for the We could say to the market, Samsung has a leadership for the memory industry. Samsung is not only component provider. Samsung is the total solution provider. Means that we directly goes to market than the other. Why? We have all solutions, the vertically-integrated. All the things vertically-integrated. For example, SSD case, we have a controller. We have a NAND flash. Inside of that SSD, there has to be have some DRAM.
The many package technology we have to provide to our customer. Samsung is only one company to have a total solution and go to market rapidly than the others. That's why Samsung is the number one in the world. Okay. I have only two slide. You must be hungry, and I explain all the stuffs. If I deliver my message properly, then there is no question today. Okay. What is the Samsung Memory's mission? Already I mentioned that AI and in-memory database. Here, left side is human being. In the long term, AI and machine learning has been done properly, and then some robot is going to be take over some timeframe. What Samsung Memory's mission from now on, Samsung Memory is going to the bridge between human being and AI robot. Why? Think about that. Robot. I watched Eagle Eye. I watched that movie.
I have very, very impressed because the Eagle Eye is a kind of AI, total control system in the U.S. They are watching all the stuff. They order all the stuffs. To destroy that AI system because they try to destroy some issues, and then the guy try to extract from the system the memory. If they don't have a memory, in that case, that very sophisticated AI system is meaningless. There is no memory. Through technology revolution, we are bridging between human and AI robot. This is going to be last one. Samsung keep providing leading-edge technology for our customer. By utilizing that Samsung memory, our customer provide their customer, means that their society, and that make this industry healthy, and our human being life is getting more easier, getting more happier.
Samsung mission statement is to devote its talent and technology to creating superior products and services that contribute to better global society. This is end of my presentation, and you guys must be hungry. I am also be hungry. If I explain about my presentation agenda properly, then there is no question. Let's go to have a dinner, a lunch, please. Okay, let me stop here my presentation today. Thanks for listening my presentation. But this scenario is right or wrong, and maybe two years, 10 years after, you may realize that one. I hope you, this going to help you to understand the market. Thank you very much