Phison first quarter 2024 earnings call, this is Sunny Lin from UBS Research. Joining us today from Phison is K.S. Pua, the CEO and founder of the company. In the first part of the call, K.S. will provide some update on the first quarter results and business outlook, followed by the company's product introduction. After that, K.S. will also share some update based on the recent frequently asked questions, and we will then proceed to the Q&As. We will now pass it to K.S., you may start. Thank you. Thanks, S. Good evening, good afternoon, good morning. This is K.S. Welcome to join Phison with CQ1 2024 earnings. Okay, so let me start to share my presentation here. Anyway, we start with this page. Since four quarters, we collect the NAND, the memory companies' earnings for last four quarters, the accumulated loss is like more than TWD 20 billion. But the good thing is, move to the CQ1 2024, every NAND makers, memory makers turn to profit. I think this is a good sign. Actually, in the CQ4, we already predict the NAND memory makers is going to turn to profit. So the market is back to healthy, and we believe the new investment will happen very soon. Okay, this is about Phison's CQ1 2024 breakdown of our product line. Controller-wise, revenue actually increasing, but overall our CQ1 revenue improved, so the percentage is dropped around 1%. Industrial module, you can see for five quarters keep declining, and you're also able to read from the market. Industrial deployment is slow due to the high interest and also low demand. So industrial, I believe industrial already go through the most difficult period, and we hope to see the bounce back by CQ3, CQ4. Gaming is a slow season, again due to the inflection due to the interest rate. Embedded OEM we are doing good, is a big improvement, because we got many mobile and also PC OEM tier one design win. The one thing is surprising us is about the demand of enterprise storage. By CQ1, enterprise storage already 5% of Phison revenue. By the way, actually this also follow our, we wish follow our expectation because Phison invest heavy investment into our enterprise storage. We can see the enterprise storage is turned to very strong. The price from the NAND makers, they also keep increasing a lot. And since last year, Phison also made a lot of a design win into the tier one data storage supplier, data center customers, some medium-sized cloud company. So we believe enterprise business will become Phison's very key business starting from 2025. Okay. This is our CQ1 financial statement. Revenue-wise, we already announced is, TWD 16 billion, which is, year-over-year improved at 64%, quarter-over-quarter 4.9%. The gross margin, by year-over-year we improved 75%, by quarter-over-quarter declined 1.2%. So this TWD 5.6 billion gross profit is, second history high in our record. So, basically CQ1 we are getting, very good business and also very better, gross margin. So this is our revenue trend, since last year CQ1 each quarterly we are improving our revenue. Okay, back to the gross margin, a lot of information here, okay? By CQ1 2024, the gross margin is close to 34%. It's historically higher in the same period and also second higher in the history. In the CQ1, reversal improved 0.1%, means almost no reversal from inventory. And Phison, if you remember two and a half years ago, Phison we were buying back our treasury share, and I decided CQ1 to exercise 100% of the share. So then with this create the cost. This also affect the margin 0.47%. So this is overall close to 34%. Last quarter was 36%, but 36% we have a reversal close to 1.4%. So if we take out the reversal, last quarter would be like 34.6%, but this quarter with our take out reversal but we put back our the treasury share cost is 34.3%. So basically the gross margin CQ1 versus CQ4 is almost the same. The earnings per share for CQ1 is TWD 12, but we also like to disclose that we have a treasury share transfer impact on the EPS close to TWD 5.4, which actually is, to be honest, a big amount, but, I'm thinking CQ1 exercising this would be much healthier to the company and reduce the impact in the future. So, if we put the cost of a treasury share back, then the earnings per share will be back to like 17.5%. So this is a very good number to our executive. So this is the income statement. The gross profit is the second highest in history. Last quarter was the highest. And the R&D expense actually also increasing because we exercise our treasury share. And the non-operating gain. The non-operating gain actually is a very big number because our investment in China, they turn to profit. We are able to recognize the gain from the investment and also some gain from the forex and also their financial asset valuation. So this gain actually is offset our treasury share cost. It's almost offset. So, that's why I decided to take the CQ1, to exercise our treasury share and make sure no more impact in coming few quarters of our financial statement. Okay. So balance sheet-wise, cash-wise we still rich in cash flow. And for inventory-wise, it's increasing. The reason is the NAND Flash vendor within a year they raise the price 3x. The NAND price increased 3x within a year. So since last year CQ4 the market start increase the price, we need to secure material for our design win project such as for the PC OEM tier one, the mobile phone, the enterprise, the automotive. So in CQ1 the price is still reasonable, we decide to keep buying to keep for our inventory. But CQ2 the price keeps going up. CQ2 price already almost as high as before COVID. So we decide to slow down to buy inventory. So our inventory actually is good enough to consume for our design win customers. Okay. So this is about inventory. Okay. Phison volunteers to share the IFRS and non-IFRS results. So it's a difference by the operating profit, actually the share-based payment actually is most likely a treasury share, is an impact TWD 1.2 billion. So if go to the non-IFRS, the OP is 2.5 billion and the OP profit margin is 15%. Go to the net profit, non-IFRS net profit improved to TWD 3.5 billion equal to 17.4 earnings per share. So, if in my memory, this should be the highest in the quarter. So CQ1 we have a good revenue, we have a good product mix, we have a better gross margin and also our investment in China also turned out good. So that's why we combine everything together, we have a very fruitful CQ1 financial result. Okay. So the non-IFRS, the EPS increased approximately TWD 5.4. Okay. And this write-off on the financial book didn't impact our total equity and company will take this TWD 5.4 as a cash dividend in the next second half. So we are going to pay more dividend even though this write-off from the treasury share. So we expect the second half when we decide the dividend number, we will put the TWD 5.4 back to our base. Okay. April revenue dropping TWD 5.1 billion. Then you may feel surprised or uncomfortable, but I have to explain here why. The NAND price at the end CQ1 almost touched same number as before the COVID. Means the price is expected higher compared to the last few quarters. And since the price started to growing up from August, the system makers such as phone makers, IoT makers, PC makers start to accumulate their inventory. So basically the system makers by CQ1, they already have a full hand of inventory and CQ1, CQ2 basically for system is a slow season. So they keep developing improved inventory but the system not shipping that much. So they have no interest to keep buying the storage for their business. On the other hand, when price going up, especially the China module house which is popping in the market, their business is very slim. They are mainly in the retail consumer business. They're buying high price. In order to get revenue, they're dumping by minus. So make it short, the consumer market in April, May actually very slow and China module house keep dumping the modules and to be honest that is a turn to minus. But Phison, we have no interest to participate this kind of a bloody game. The reason we have no interest because we have a good quality of a NAND flash, I no need to waste this go to the consumer market. But Phison, we have a second choice, third choice such as we have an automotive business, we have an enterprise business, we have an aiDAPTIV+ business which able to help us generate much better gross margin. So I no want to waste my inventory to dump to compete with the China module house. So that's why I decided not to take some order from consumer which is a minus margin. But we also try to ask the NAND vendors, are they going to lower their component price NAND? Answer definitely no because enterprise is very strong. Means China module house when they dump their inventory, they buy higher, they dump, they dump low, they are not able to buy back low. So this is their decision. So we are not going to participate. We decide to carry our good quality NAND to move to enterprise, to move to PC OEM, to move to aiDAPTIV+, to move to the mobile and IoT. So we believe with our healthy inventory, we able to generate much better revenue by CQ3 and of course much better gross margin. And we also believe the system house, they carry inventory, they start to prepare for CQ2, CQ3 system shipment. Once the time moves to the CQ3 is a traditional season, good season, they start to ship inventory low, they have to back to buy again. So we are waiting our new, exciting business in CQ3 once they consume their inventory. So again, make it short, we are decide not to go to consumer to waste our inventory and we wait. And on the other hand, we have a very good, solution which able to help us generate a much higher gross margin business. Okay. Back to the business, we keep highlighting Phison's position in the storage. Client SSD, we have a good, percentage share in the controllers. In the module-wise, actually we are increasing as also embedded ODM, we are improving our, dollar amount. The reason is since a few quarters, hard working, we got more than 10+ tier one PC models design win with our PCIe Gen4 solutions. The demand by tier one is increasing. Especially we get the forecast CQ3, CQ4, the demand is improving. Okay. The reason we can get the improving business because we have accumulated some good quality of NAND, some price competitive of NAND. So this helping us to gain not only the share but also to gain the gross margin. We also gain the 5+ of tier one PC model with our PCIe Gen5. So in the coming years, we can see Phison able to get the share and dollar amount in the tier one PC OEM business. Automotive, again Phison, we gain a very good share in controllers but in modules-wise, we are still the almost nothing. So this picture, the orange color means Phison early to get design win shipping our modules. If we talk about controllers, talk about controllers, this top 20 brand, 100% orange color means our controller already in every automotive brand, the top 20. But in module-wise, in module-wise, we still have a few, we not happen yet to get the module business but by second half, some of the gray color will turn to orange. Okay. Actually some activities is ongoing and we can see more design win modules into the automotive business. So automotive revenue by 2025, 2026, Phison is going to see significant improve by revenue. Okay. So we happy to see our automotive business is turned to good very soon. Mobile storage. Actually mobile storage most likely is a MCP business which dominate by the three DRAM key makers. But Phison, we not only have controllers, we have a controller for the three NAND makers and the four NAND makers, using our eMMC. We also have our modules to the actually mobile most likely is a China phone maker. So Phison combine with our China investment, we have a very good share of modules into the China mobile phone. So this revenue also improving but we need to get the more design win. So we have a lot of activity to work with the Qualcomm, with the MediaTek to put our design win in our the area. Then we are going to take the more module business from China phone makers. Gaming-wise, Phison, we have a significant share there but unfortunately since last year CQ4, you may read from the market, the gaming business is not that good because of inflation, because of high interest. But anyway, we keep investing in the technology-wise. So we are waiting the next wave of a new generation of a console. So we believe we can create a new number of revenue very soon when the new console launch. Enterprise. By CQ1, we have a 5% revenue for enterprise and I can see by 2025, 2026 will be a significant number. And Phison, we get the five tier one server brand, data storage supplier brand, we, our PCIe Gen4 SSD and our PCIe Gen5 SSD will go into market by CQ3. So a lot of activity design in ongoing. So enterprise is a high value but also the high, number of investment and heavy investment. To be honest, running this business by within two- three years, I am really cannot see any third party survivor because the return is slow, the early investment is huge. But I believe Phison will be the only one third party capable to develop IPs, controllers, modules, and customization. So we keep investing because we believe this business will be big enough to Phison. So we are there and we keep developing our new solutions. Industrial modules, actually industrial business is terrible but we can see by CQ2 already touch bottom, the demand slightly coming back. So we believe after two-three quarters, the industrial storage will back to normal. So we keep also design in many, automotive, dedicated applications and a lot of, transportation and also some defense industry. Okay. So this is very stable but we can see some, growth after two quarters. Okay. High speed interface chip as a PCIe Gen5 Redriver, Phison, we gain the more than 50% share. Okay. So this amount is very small, it's a small business. PCIe retimer, we already got the several AVL from the server makers and we also try to put our retimer into the AI GPU platform is ongoing. We also working very close with the CPU and GPU supplier. So second half, by second half, we can see the revenue coming from retimer. So we are working very hard on this business. And Phison, by retimer, we have a few advantage. First, all the design team based in Taiwan, all the server AI makers, most likely Taiwan and some in U.S.. So we are here, direct support to ODM. The distance-wise is, within 90 minutes by car, no need taking airplane. So we have a much better service and the language benefit here. And Phison, we own 100% IP. So the ODM, they are happy to work with Phison. So we are getting their design in Windows with the ODM and the AI GPU customers. Okay. So back to aiDAPTIV+, I want to make the summarize again. Since April, Phison has many activity. We have a developer events, we join the AI Expo, COMPUTEX, we are going to many MOU with our partners. This was, back to the history 20 years ago, everyone saying the factory need to use a robot. The answer is yes but too expensive. After 20 years cost time, now a lot of factory full with a robot and no more manual, no more human there. But this spent 20 years. This same go to the generative AI. We believe the human need a lot of generative AI to help them improve productivity but unfortunately today, the generative AI is too expensive. Same as 20 years ago the robotic arms. Okay. But now we no need to wait for 20 years. Phison, we invent our aiDAPTIV+ which is only few% of the cost. Then the users able to go to their fine-tuning. So we give example, this is a cloud, in the world, they are able to only less than 10 able to spend few billion, few 10 billion dollars into their GPU infrastructure and ODM from Taiwan building their GPU system to them, they start to create the model we call it a pre-train. The users, they need to use inference but due to a lot of data privacy, users may not willing to share the data to cloud for training. So between the cloud, generative AI supplier to the users in the middle, they have a so-called the fine-tuning opportunity but the investment at the past was too high. But Phison, we able to create a new invention, we name aiDAPTIV+, the user just spend TWD 30,000-TWD 50,000, then you able to go to make your own fine tuning by yourself and this is proven and we already start to ship quite good number to the market already. So for the fine tuning, for the generative AI, it's more care about data privacy. Okay. But for data privacy, you want to train on prem, then the cost is a challenge. So thanks to Jensen, he created the generative AI. What his goal is, he, what this is what he said in the YouTube, he said you own your own data. He also agreed you own your own AI but he didn't mention you need to pay a lot. So most users, they own the data but unfortunately they have to go to cloud for training but if go to cloud, then data privacy will be the issue. So Phison's goal, I spoke to Jensen, he agreed you own your own data, you own your own AI but you are able to train locally with small amount of investment. So aiDAPTIV+, we able to achieve what Jensen saying that you own your own data, you own your own AI and Phison, we make it and we prove it to everyone. So we can see during COMPUTEX, a lot of our Phison's partner, our ecosystem, SI, hardware, software, they are going to demo to show the use case of aiDAPTIV+ to everyone. So we believe after the COMPUTEX, aiDAPTIV+ will going to be very popular. Okay. So aiDAPTIV+, we name this is a home computing on prem, secure, reliable and one is called affordable. Affordable, you no need to spend that much of investment. So what's aiDAPTIV+? Actually aiDAPTIV+ is Phison exclusive patent, our invention. We use a NAND, we use a NAND to replace, to reduce the DRAM memory in their systems. Just give a number, HBM, HBM 1 GB is $10-$20 per gigabyte. NAND, enterprise NAND is a $0.10, $0.10 versus $20. Okay. So you can imagine how much we are able to lower the cost of the training machine. So nobody believe NAND can go to the generative AI platform but we prove it, we make it, we also putting the NAND to have a more opportunity and share in the AI ecosystem. So Phison, what we are doing is we are providing our this so-called, this is not SSD, let me clarify, we name this is AI100. Actually it's aiDAPTIV+ cache. Plug in this into the AI system with Phison's proprietary middleware, then you can start to have your fine-tuning with affordable investment. Okay. Now we give you some number here. Okay. We're doing the, we work with our ecosystem partners, in Taiwan to train the Chinese Taipei and the Hakka reading comprehension. We use a two model, Llama 2 and the Mistral 7B. This is using Phison aiDAPTIV+, the GPU card is for NVIDIA and this is not the H100, this is a reasonable price with the two piece of a Phison aiDAPTIV+ AI100 and this is our training condition. Now we see what's the difference. This white color is a natural Llama 2 70B. White color means nature, nature. Means you download the Llama 2 or Mistral from the website, you can immediately inference, you can get the 81.6% accuracy. Taiwanese, you can 69%. But with the Phison, the aiDAPTIV+, with this affordable GPU, with Phison's two piece of Ai100, after three approach, three approach, taking this token, improve to 81%-92%, 69%-85%, 68%-87%. Then the question is how much, how much of this equipment? This equipment is around $50,000. Not $5 million, not $5 million. Okay. Not $2 million, it's $50,000. So this is a number to prove aiDAPTIV+ is good to use and affordable in the edge applications. So this also can help to a lot of small, medium size, corporate, single, individual users, you can take this to improve your productivity. So who need the aiDAPTIV+? Actually we already have more than 100 use cases, lawyer, judge, accountant, engineer. Phison in-house engineer, now we have a seven use case to use aiDAPTIV+ to improve our knowledge management, to improve our copilot, our firmware design. Now we are doing very low design. We are doing document by our aiDAPTIV+ systems. So a lot of POC ongoing with Taiwan's, U.S. and Malaysia system integrators is ongoing. So we, Phison, need to build an ecosystem. So today we have many service provider, the system integrator, the software company, the hardware company, I think this is you know who they are and more partner coming from Southeast Asia, from the U.S., somebody from Europe. So now we are helping them to build the whole ecosystem to help aiDAPTIV+ go to the popular in the market. So COMPUTEX, we believe a lot of media here, through the a lot of report, we believe after COMPUTEX, aiDAPTIV+ will be popular and we believe a lot of, system integrator from the world will come here to take this and to promote, deploy into their country. So this is going to be very exciting in second half of this year. And we already start to ship by CQ1, actually we ship like close to 100+ sets and CQ2 we can see a few 1,000 sets of assignment and we expect by second half next year, we hope, we hope go to a few 10,000 pieces of assignment. So this is very hot today and very exciting. So again, we repeating aiDAPTIV+ is a home computing on prem, secure, reliable and affordable. Okay. Back to Phison, CQ1 we still have a heavy R&D investment is 21% of the revenue but we still able to get a good gross margin. So the reason we can have aiDAPTIV+ within six months just because we have many years heavy investment into R&D. So we keep a lot of IP portfolio, a lot of knowledge on hand. So when market come, something happen, we can immediately react and create the good products to the market. So let me have some summary about our CQ1 and CQ2 expectation. CQ2 is a slow season, traditionally slow season in the consumer market. So I am not willing to go jump into that broadly competition. I want to keep my good quality NAND for our good quality business and we have a good solution and we can decide to move the NAND to the aiDAPTIV+, to enterprise, to automotive which we able to get reasonable market. NAND price already risen to relatively high level. We don't see the NAND vendor willing to lower the price and we also don't see the NAND price will keep increasing. I think it's already good enough and healthy to the market. For Phison's aiDAPTIV+, we can see very good movement, a lot of design in, a lot of POC ongoing, a lot of partner coming to us. So we expect second half this year, we can get very good return. For enterprise, demand is strong actually. We are also facing some NAND flash shortage, but we keep talking to our NAND partners, asking the supply from them. Our PCIe Gen 5 were launched by the CQ3 so we hope this also can help us to gain the business. By the way, Phison also putting our enterprise SSD into the cloud computing system, also into some AI systems. So even though we are small player but we get some significant design win with our ecosystem partners. Automotive, we have very good design wins there and we can see more the top 20 of automotive brand we are able to put our module in. Last is about the PC OEM. Actually since last year, we got a lot of inquiry from the tier one PC OEM asking for customized our SSD for them. So they trust Phison good not only in the IC but also firmware system and the module quality assignment. So PC OEM will become also our key business second half this year. So this is summary about today's earnings call. We call it some Q&A here. So let me clean up this, then we can go to the online. So is it Phison going to keep expand the ecosystem? The answer is yes. So we use Taiwan as our home market, training our partners here. Now we start to move this to Southeast Asia. We start some engagement in the U.S., next will be Europe and also South America, Middle East, Africa also came to us to ask about aiDAPTIV+. Unfortunately, we don't have that many resources, human resources to support worldwide but now we need to engage with many Taiwan local system integrator partners. We need to work with them, then we go together to everywhere to support the human to have a much better productivity. Yes. Let me, okay, sorry. So is it Phison's aiDAPTIV+ compatible with the GPU? I need to say something this is quite sensitive to answer, okay? But we are not writing here, we are not writing here but you can listen what we are talking. For today's current AI training, I think you know what's there, right? H100, MI300, this is a so-called high price GPU system. High price, it's a million dollars, at least million dollars. But Phison aiDAPTIV+, we can adopt the NVIDIA, AMD, Intel GPU, okay? And POC, almost every NVIDIA, AMD, Intel GPU are able, compatible to Phison aiDAPTIV+. The difference is a performance. So now we are working with the individual CPU, GPU supplier to improve the performance. So for the end user cost, performance is very important, okay? So we already prove every GPU we are able to use in our aiDAPTIV+ but we need to keep investing a lot of software engineer to improve the performance on other GPUs. Okay. How Phison will sell our aiDAPTIV+ solution? We are building ecosystem partner. We have a profit sharing not only with the hardware ODM partner but also software developers, system integrators, service partners. Okay. So this is we are building a new ecosystem which not happen in the market yet but with this new ecosystem, we believe every participant, they are able to enjoy good return. Why? Because we lower the fine tuning equipment from few million dollars to 2%,3%,4% means they're able to gain the value from these systems and user, they are more willing to pay because it's only 2%,3%,4% of original budget. So we believe they can get the good return together with us. Of course, most of investors, they are concerned about the price increasing. First of all, I disagree the price will keep increasing because if price keep increasing, they will turn to bubble, okay? We believe the price already in the relatively high, now it's healthy, then waiting the consumer market back by CQ3, then everything will be like regular, okay? So we are waiting there but on the other hand, we can take our inventory, good quality NAND inventory, go to the high profit business, okay? So we think most likely will be scenario B happening in second half this year, okay? And this about inventory, we believe by CQ2 this year, we can see some inventory reversal from our write-off. So this can help improve some of our gross margin. Okay. So this is a question we raised from investor. Now we go to the online Q&A. Yeah. Thank you so much, K.S. That was very detailed prepared remarks and highlight of your business and products. We will now kick off the Q&A session. Attendees can use the raise hand function or type your questions in the chat box. As a reminder, if you would like to ask a question, please state your name and come to the table. Thank you. I think we have a few people raising hands but if I may, maybe I can start with a few questions. So K.S., I think my first question would be regarding your view on the NAND market because you just mentioned that you think the profitability for the memory makers will return to a more normal level in second half this year. What is your observation on the supply and demand for second half this year? And I think based on your current inventory on hand, will you still consider to take more high quality flash with the pricing upward trend continuing at this stage? Thank you. Okay. Basically today, if talking about supply and demand, right? We go to the NAND vendor, they said they are under supply and this is true. Why? Enterprise, data center, cloud, they ask big amount of petabyte to build their inventory, to build into their systems and this is true. But today if you go to the mobile and the PC OEM, they are more conservative because they already built their inventory. They are not hungry. Overall, supply demand according to the NAND makers is under supply but I think the gap is not that much. The main drive of their demand is enterprise. But if time moves to the CQ3 this year, we believe mobile phone, PC OEM, some desktop AI PC may have to go to the market means the system company will start to reduce their inventory. So they have to go back to buying some more components for second half and next year what they want to use. So if this happens, consumer market will turn to normal, normal means healthy. Today consumer market is very, very weak which that's why I don't want to participate, okay? So overall by this year, second half, NAND, to be honest, I think will be tight. Price most likely may still increase little but not much, not much, okay? Because increase much, they will kill the market. So Phison now we have a good enough of inventory. It's good enough. We are thinking to keep our inventory, not to waste our inventory but put our inventory go to the good quality business. Maybe by end of CQ2, if a NAND partner needs some help, we may consider to buy more from them for our future use. Okay. Thank you. And my same question would be a longer term question regarding your generative AI solution for edge applications which you mentioned that you price it at around $50,000 per set and it is very attractive. So I think could you actually share can your chipset be adopted in the existing system or will it need to be adopted in the new systems the customers build going forward? And how should we quantify the potential opportunity with the runway you just mentioned? First quarter, a few 1,000 sets in second quarter and 10,000 of shipments in second half. So how should we quantify that as you get more mature on that part of the business and how is the business margins compared with your corporate average? Thank you. Okay. You look to this picture. This is the workstation we are now working with our hardware ODM partners. This motherboard, the chassis, the power, the fans, the DRAM, everything is a legacy component. Means ODM, there no need to invest new motherboard, new chassis, new power supply, nothing, nothing. This is already happen in the market. So without the new investment, just plug in Phison AI100 with Phison middleware. Then this equipment immediately turn to the fine-tuning machine with very low price to the market. So means this is going to can be popular very soon because you no need to have an extra investment. And second, we need to prove, we need to prove this is going to use, I mean very useful to the market. So that's why we have many system partner, the software partner. We need to tell everyone, every single users, every small, medium size, you need document, you need everything, this machine can help you to make it. So we need to collect a lot of use cases. So now we have already over 100 by COMPUTEX, we can see many booths to demo different use cases, okay? Once this happened, this will turn to like 10 years ago PC desktop. So what's the difference of this equipment? You can take this is a PC desktop. This equipment you no need to put into the so-called the chamber, you no need to put the air-conditioned, you no need to give a big power supply. This is that desktop you buy is, you put on your table, then you can run your fine-tuning. So we call this is a home computing. Easy to use, easy to deploy, easy to sell, easy to assembly. The only difference is you just plug in Phison ai100 with our middleware. So once the COMPUTEX, after COMPUTEX, a lot of use cases in the market published and the media, the press, the worldwide will realize, oh, this is a mature product. If this is a mature product, things is way easier, okay? The system integrator, they just promote, they consult, they're helping the users to integrate this, then Phison easy, we just sell it. So I don't see basically we already spend the six months to educate the market. We see, we can see the second half this is going to popular very, very soon, okay? Okay. Yes. Thanks, K.S. That is very helpful. I think we have a few people raising their hands. Just a reminder, please state your name and company name before asking questions. Jason from CLSA, you may unmute and start. Thank you. Then we Have time. Heads up, I'll move to second one, then next we're back to Jason. Okay. Charlie from Morgan Stanley, you may start. Thanks. Thanks, KS. And KS, congratulations for very good results. My first question is very simple. It's about your second half gross margin trend. Since you mentioned that the NAND module price cannot go much higher and I suspect you kind of run out of some low cost inventory, right? So can you give us some sense about the gross margin trend in second half? Yeah. Actually, I talk to my customers, module house from China, they may suffer minus in second half because inventory, low cost inventory gone, they buy high, they sell low. But again, Phison is a different, okay? Since yes, we have a good quality of a NAND, good quality means very good cost structure. We need to use this into the meaningful products, okay? But on the other hand, Phison is not like China module house, buying component, making modules, use a cost up model to sell the products. We create new applications such as aiDAPTIV+, such as automotive, such as enterprise which the customer willing to pay premium, okay? So I agree the low cost inventory you know to the Q4 may gone but during that period we are able to generate the new products with the attractive gross margin to the market. So we keep investing new technology. aiDAPTIV+ now is a first generation. By next year first half, we have a second generation coming. So second generation is more powerful. Powerful means we are able to take much better premium. So I agree gross margin second half is a challenge but I don't see it's a big problem to us because we keep investing, we get the good product and aiDAPTIV+ can be a good example. Yeah. Thanks, K.S. Actually, our next question is on this aiDAPTIV+. So a couple of questions. So just a very kind of top down question. If you have to provide us some revenue contribution percentage guidance, what would be the revenue contribution this year and the next year? And also some technical question, right? Your edge server is a $50,000 but NAND 1T price is at $2,000. So for Phison to book revenue, do you book the system revenue or just the storage revenue? Thanks. Okay. First, Phison is hard to give guidance about aiDAPTIV+ because this is not happening yet, right? We are still in cooking promotion but I believe after COMPUTEX, after COMPUTEX, CQ3 we have a more clear picture. Maybe next earnings call we can share something, okay, to the public. Second, talking about the Phison revenue, first of all, Phison we are not selling workstation. Workstation is not my business, okay? Workstation is their business. We are not selling service. Service is owned by the system integrator and the software company. We just selling Phison's AI100, AI100 with our middleware. So our revenue coming from our storage, we call AI100 and our middleware licensing, okay? Yeah. Okay. So let's assume the storage cost is like $2,000 but you also charge the service. Sorry. License. Like middleware license. Yeah. Yeah. Yeah. Yeah. Okay. So what will be the markup then? Okay. This is something like business confidential but I always talk to my customer, my SI very simple. Hey, since you want to go to Llama 2 70B training, you need to put around $2 million for DGX but now you pay only 30K, 30K to run Llama 2 70B. How much money you are willing to share with me? It's simple, right? From million dollars to 30K. So actually the users, they are willing to pay more. Then more is not that like a million dollars, it's a few 10K. So we think this is attractive business to Phison. Yeah. Okay. And next one, you said you invent this technology, right? But do you think outside of Taiwan, those Western Digital, Micron, whatever, right? If they see the benefit of this solution, do you think they will come up with a similar one without infringing your patent? Definitely yes, right? But let me make it simple. In the market, there have a NAND company, controller company, and module house. Module house, without the controller, they cannot make it. Controller house, without the middleware, they cannot make it. This is a system development, it's not any just a single SOC. So we believe module house or controller house, when they let to do this, they will take time. Phison, of course, we invest a lot of patent there. We are standby. Anyone come, we're happy, we're happy and we are standby to take action. Back to the NAND company, we believe NAND company, they know they are capable to make this. It's not much secret. But all the NAND company, actually they are Phison customers. Phison also their customers. And Phison's role to the NAND company is getting important because they have many key controller come to Phison, they also ask Phison to use their NAND flash. So we have a very close and deep relationship. If this is true, why they need to go to build this and come to Phison for dedication? It's nonsense. But Phison, we already have a business model. We're doing design service, design ODM to the NAND company. They can either consign the NAND flash, we build the module to them, or we license our products to them. And we already have aiDAPTIV+ program with one of a NAND company through design service and consignment, okay? So we don't see the NAND company interest to go to this because to be honest, this is not able to consume high petabyte. It's not high petabyte, okay? But we are willing to help them because the cloud player, they won't come to Phison. Phison still too small compared with them. So they may go to the NAND company, NAND company come to Phison, NAND company happy, we're happy too. Okay. Okay. That makes sense. Okay. Thanks for your answers. I'll be back to look here. Thank you. Yeah. Thank you. S. Okay. Yes. Thanks, Charlie. Simon, you may go ahead. Thank you. Oh, yes. Thank you very much, Mr. Pua. The question is regarding the enterprise SSD market. So, number one, do you think the QLC NAND is really critical for the AI server, even your new AI in the NAND solution business? If so, why? Define. Define Define what critical. I cannot catch what's critical. Critical means difficult to use or what? Maybe more cost competitive versus TLC and also the older controllers are ready because I'm hearing that the enterprise SSD demand for AI servers these days more and more QLC NAND based and also the related controllers. Actually, you can ask the market who has a QLC enterprise today. My understanding, only two companies able to supply, okay? So the Phison no plan to supply the QLC NAND based enterprise solution because? No. No. No. Ongoing. Simon, by Q1 next year. Oh, next year you can supply more cost competitive QLC NAND based SSD for the enterprise solution? Yes. So, in that case, your controller will be okay next year? Currently the R&D process? Yes. Definitely. Yes. Then, how about if any competitor promote the TLC NAND for enterprise solution? What could be the QLC's relative advantage because QLC sounds a bit slower than TLC but it's cheaper than QLC is cheaper than TLC but slower than QLC. So how you want to optimize overall function with your maybe controllers? This never happened from SLC to MLC, MLC to TLC. Same question keep repeating, right? But humans solve it. So now I can tell you QLC should be no problem. Controller is getting powerful. We have a solution. So any rough idea what percentage of the cost saving versus any chance to see the better performance with the QLC versus TLC? In a lot of application, most likely from nothing to yes, you care about performance. From yes to big volume, you may not think about performance. You think about the cost. So a lot of application, they don't care much about performance but they care the cost. Yeah. Okay. And then the last question is when do you expect your balance sheet to show maybe a lower inventory level? And then no need to worry about your value of the NAND inventory, overall inventory value along with maybe potential flattish NAND price or some softening NAND price. I mean, my question is potential valuation in the reduction but the lower NAND price. Yeah. Okay. I give you a scenario, okay? If I buy a NAND wafer from NAND company today, I can turn to revenue only after 90 days because NAND need to go to packaging, testing, assembly, burn-in, okay? Means the turnover of flash is 4x a year, okay? Then if I need to make TWD 1 billion revenue easily, I need to TWD 250 million, right? I want to make TWD 2.5 billion. So this is easier mathematics. Of course, today my inventory cost is I keep saying it's a good quality of a NAND. Good quality means good cost structure because we accumulate it, right? And I want to take this to get since the flash is tight, I want to take this as a good weapon to support my products. So of course, NAND price is related higher now. I have no much incentive to buy more. So in coming few months, you can see inventory was slightly going down. Then we're waiting by quarter end if a NAND company call help, then we may go back to buy again. So I don't much worry about my NAND inventory because NAND makers already suffer five quarters minus. Finally, CQ1 turned to profit. I don't see an yone willing to back to minus again. So in coming few quarters, the NAND will be very healthy. Healthy means in terms of—sorry. Price is stable. Price? A stable price to your point. Yeah. Yeah. So overall, you had a very exceptional management experience. There are lots of upturn, downturn. So 2023, first half very tough, second half well recovered. And then 2024 year to date, so far so good. So how do you assess overall NAND industry overall cyclical trend? It will be still? No, I'm not going to comment because this is not my business, okay? The feeling that the NAND price is healthy. Happy. Happy. NAND maker happy, customer happy. So NAND makers happy and then you are the NAND user who are distributor, developer, also happy, you mean? Yes. Yes. Okay. Yeah. Yeah. Thank you, Simon. Thank you very much. Yeah. Thank you very much, K.S. Appreciate. Okay. As we are running out of time, I think we can take one more question from the attendees. We will try Jason from CLSA again. Jason, you may unmute yourself. Thank you. Jason, he's writing. Actually, okay, the question is my understanding is yes, the price increased a lot means the demand by gigabyte actually declining because users, they may not willing to buy their expensive products. So that's why you can see the gaming declining, gaming, okay? But on the other hand, if Phison start only in the retail consumer business, now I'm in trouble. But good thing is we have enterprise, we have aiDAPTIV+, we have automotive. They're helping us to get the new opportunity with much reasonable gross margin, okay? So second half, CQ2 consumer is weak but enterprise is strong. But we expect CQ3, CQ4 aiDAPTIV+ will turn to very strong, consumer will back to normal, industrial will pick up. So we expect the second half we have to working very hard to get our revenue and also our gross margin, okay? So I hope I'm answered Jason's question. Yes. And I think we also have one question in the checkbox from Ambian if you could provide some detail on. Thanks. Basically, within two years, within two years, I believe I can't see any third party able to developing enterprise controller anymore because it's no way to get the return and this proving, okay? This proving. But Phison, we're still investing in our controller because Phison, we are not making controller, selling controller. Of course, we're selling controllers but most of our return is coming from the modules which is a customized modules. And Phison, we have a good invention aiDAPTIV+. We can take our controller, our NAND flash built to the special module. We can get a much better premium from the market. So we keep investing. Gen5 enterprise, I think, is going to popular by 2025 and we are getting the design wins and opportunity. And I believe by 2026, enterprise products will be Phison's biggest share in the revenue. So we keep investing, okay? So I hope I answer Ambian's question. Yes. I think that's very helpful to the attendees. And before we wrap up the call, if I could squeeze one more question regarding your AI business. So K.S., I was wondering if you could share your strategy on AI product roadmap and outlook for the business. In addition to your current rollout, what can we expect from your roadmap on AI products going forward to drive more synergy in your AI strategy during the next couple of years or the next few years? Thanks. Okay. Our aiDAPTIV+ is very, very new, very new. Of course, definitely we have a second roadmap but we think better not to go to public. We keep this as confidential. What we try to do right now is to build the ecosystems of our first generation AI, okay? To help the products go to the market popular, to go to popular. So this is what we are doing with first generation. So in-house, we have a second generation ongoing but again, we keep this as our confidential. And by first half next year, we are going to disclose our second generation and this will be more powerful, okay? And not only saving the cost but also improve the performance, also reduce the power consumption, okay? So that's what. That's what. I think the first wave, Phison's first wave is we want to put aiDAPTIV+ popular, we get our return, we help our ecosystems to get very reasonable return. They're happy. They promote it. Then second generation will coming, pick up to make them taking the much better gross margin business, okay? Okay. So due to time constraints, we will wrap up the call here. Thank you so much, K.S. and Phison team, for having this call today to share the business update. And thanks everyone for joining this call. If you have any questions, feel free to reach out to Phison IR team or your sales representative at UBS. Thanks again and have a nice weekend ahead. Yeah. Thank you. Have a nice weekend. Bye-bye.