Good morning and good afternoon, everyone. This is Tiffany from Morgan Stanley. Thank you so much for joining us today for Phison's 2024 Fourth Quarter Earnings Conference Call. It is our honor to have CEO Mr. KS today to share about the financial performance, market monitor, and business outlook for Phison. After that, we will open the floor for the Q&A section. Now, I would like to turn the call over to KS. KS, please go ahead. Thank you.
Good afternoon, good morning, ladies and gentlemen. Thanks again for joining Phison, the 2024 Q4 and the whole year's earnings call. In this section, I will combine some pages with the Mandarin, which is easier for Taiwanese local investors to understand what we are talking about. This is our Q4 configuration. From this chart, you can see consumer is a little bit bad because it is affected by the traditional consumer season. Controller-wise, it is flat. Industrial is almost flat, but the enterprise is down just because of some inventory in the market. It looks like this inventory is slightly soaring. Overall, we are happy about our configuration, which is still over 70+% related to the non-consumer products. Financial report, revenue-wise, is at the bottom. It is the lowest in the last year, the whole year, the Q4. Gross margin is 3.8 million.
The worldwide is down 31% just because of low revenue and the slow in the market due to the inflation from the U.S. market and the worldwide impact by the worldwide inflation. Overall, we're still happy since the NAND market and the DRAM market was really, really pessimistic in the CQ4, but we're still able to get kind of acceptable gross profit. This is we keep working very hard on this. Revenue-wise, CQ4 is the lowest in the last year. We believe in these last four quarters, plus this year's four quarters, CQ4 is one of a slow season. We try to recover from our coming CQ1. The gross margin rate, I think we need to briefly to you. Overall, the CQ4 gross margin is 30.8%, which we have a write-down 4% back from our 4.7% from our inventory.
Q4, the market price is really bloody because of demand is low. And China module h ouses, they are dumping their inventory. If you read some report from China module houses , if they are public, you can see they also suffer their gross margin, and some of them suffer the minus net profit. Overall, the gross margin average is 30.8%. The whole year is 32.4%. The 2024, the gross profit actually will improve because of revenue growth. The earning per share by Q4 is 11.6. We have some explanation here. Basically, you may know we have exercised our investment in China. We brought back some extra investment profit back to our financial book. Okay. Due to that, we have extra profit. We also write off some profit to the R&D expense by the Q4.
So 2024, the whole year's EPS is 38.95, which is close to the 40. Let me use a Mandarin to explain more okay here. [Foreign language] 我们在CQ4去年因为有业外红心的转投资的这个受股的关系，那我们得到了一笔额外的业外利益。那因为有这笔额外利益，我们有提列了这个额外的奖金还有研发费用。所以在CQ4，我们的这个EPS是11.6，那整年是38.95，相对于2023年我们还是有成长。但是不管怎么看，CQ4是一个相对比较这个需求比较差，而且延长这个也很积极在去化库存。那中国的Motor House也很积极在杀价。那从他们报表各位可以看得到，他们的报表大概率还是赔钱的居多。那在现实情况底下，我们可以达到38.95，我觉得公司对这方面还是满意的。那我们更希望2025年是一个潜在更有发展潜力的一个年度，我们会持续的努力. The income statement, you can see here, I think you may need to know about our operating expense. We increased to TWD 4.8 billion due to we have extra profit about the investment. Due to that, we write off a big amount of the profit. In the OP, OP turned to minus just because income is from extra investment, but after we write off to our overhead, then this impact to our OP. Again, overall, the net profit in the CQ4 still close to 11. [Foreign language] 因为第四季的业外的这个收益，我们提列了奖金。那这个奖金变成是费用，那造成了营业利益的这个损失。但是不管怎么说，整年整个quarter还是赚11块多。那这个情境如果各位投资人清楚的话，2020年CQ4也发生一样的情况。当初合肥造新跟红心合并，我们也得到了一笔额外利益，我们也提列了一笔费用。但是不管怎么看，获利就是获利。那我们整年度还是维持在38.9的一个获利. In our balance sheet, revenue-wise, actually we grow quite a big amount. I think it's a good growth, 22%. Gross profit, we also improved by 18%.
The R&D expense, we also increased the R&D expense close to TWD 12.5 billion. This I think is the highest just because we have extra profit. By last year, we are taping out the 7 nm chips. Okay. You may know the mask cost is really high. This year we still have one or two need to be taped out. Last year, revenue growth, gross profit growth, by the way, the R&D cost also improved, increased. Anyway, the net profit, EPS after tax, we still deliver the good result to our shareholders. Okay. This is our total asset. I think by inventory-wise, I remember by 2023, the highest inventory was close to TWD 3.2 billion, TWD 32 billion. Now we are lower to TWD 24 billion.
In the coming CQ1 and CQ2, we are increasing quite a big amount of inventory due to we book a big amount of the low-cost inventory at January, February, and March. Okay. Later we can explain more here. [Foreign language] 库存部分，我们从高点320亿台币降到24亿。所以现金水位也回到了大概100多亿台币。那库存可以说是过去八个季度相对的低点。但是因应从去年12月后的减产，我们在今年1月、2月、3月，我们也从五家延厂已经有谈成了几个订单。那这些货物是从2月、3月、4月、5月会陆续的进来。那我们认为我们的这个相对的成本是在一个相对应很低的一个点。所以对于未来的营运应该会比较有帮助。对，241。那我们的整体的这个库存水位应该在半年度的时候应该会增加，因应下半年的需求。 Okay. Back to the Native. By CQ4, due to we have extra income, we write off overhead. The OP returned to minus. If we take out this impact, the real operating profit without the extra income and the extra R&D bonus and R&D cost, the OP is around NT 7,702 million. The gross margin profit rate is 6.1%. Back to the net income, it is around here. Okay. Basically, the business in operation, we are still in positive. Of course, CQ4, the demand is weak, the price is challenging. Overall, we still deliver the positive result to the shareholders.
[Foreign language] 因应我们把这个业外的部分把它先分开来看的话，本业的获利在第四季度还是维持一个基本的一个获利。那最主要是因为业外获利，我们提列奖金造成了所谓的OP的这个伤害。但是这个在2020年发生了，那已经我想跟各位股东来做说明。 Okay. 好，那对于整体的一个年度的影响，我们今年 Okay. Regarding the whole year's, I remember CQ1, we have the treasury share. CQ4, we have extra exercise investment. Overall, if we take out these two impacts, the OP is around NT 6.5 billion and the OP is 11.1%. Okay. The basic EPS is 35.3. Okay. This is just to deflate our operation. [Foreign language] 为了反映真实的营运所产生的一个利益，我们把在CQ1的库藏股的影响，还有CQ4的这个处分业外投资的影响把它拿掉的话，本业的获利是35.36的一个EPS。以上跟各位股东做汇报。 Okay. So this is about our 2024 financial report. February, our revenue is NT 4.5 billion and equipment is NT 8.1 billion. We believe, I believe, 2025 January by monthly is the bottom. Okay. We can see the demand month by month, MOM is increasing. We can see also demand turned to the strong. We also can see the consumer market, embedded market, PC, OEM, mobile, enterprise, industrial, they are willing to keep asking to buy more storage and controllers from us. Bottom line is the bottom month is January. [Foreign language] 根据整年度的观察，我们认为1月份应该是整年的营收的一个谷底。那有农历年的关系，也有因为去年CQ4这个通膨所carry到的影响。但是农历年后，因为中国DeepSeek的一个关系，再来年内的延长减产的效应，我们看到每一个月的需求确实是增加的。那我们也会努力的把这个业绩跟获利把它做好。 Okay.
We have some business highlight. Last week, we have a press release regarding our enterprise SSD, which launched to the moon. According to the updates, most news from our US office, this already landed on the moon, but they have some technical issues. Overall, the SSD is still working. Okay, still working. We are waiting the more testing report from our partner, Lonestar from the US. We also launched our first in the world, the first PCI Gen5 waiting line. We got some design win in the Tier 1, Tier 2 PC OEM already. We also start mass production for the Tier 2 PC OEM. This controller, we win several NAND flash controller outsourced to us. We can see this controller will start to bring more value to the PC industry.
We are also taking not only the TLC, but also going to deliver the QLC products to the market by the CQ3. We also started to release our UFS with the QLC solution to the China five mobile makers. Next week, there is a call, China FMS. We are going there to test together and to engage how to bring the QLC into the mobile phone. We have a full line, UFS 2.2, 3.1, and UFS 4, which cover the QLC for the China mobile market. One more thing is you may also like to ask about adaptive. After the, actually during the Lunar New Year, the DeepSeek, everybody knows about that. Yesterday, Alibaba also announced their Qwen, QwQ, which is the newest model. This just proves one thing.
To create a model, you don't need to use a multiple billion, just what they claim is $5 million. The new model is easy to be created. This is helping the AI LLM much easier go to the edge. Phison, we are working closely with the many system partners, server partners, and also GPU card partners. Week after in the US GTC, Phison is going to launch the newest aiDAPTIV+ solution in there. We can see, according to a lot of research and the market discussion, the pre-training is not need to cost that much anymore. Much easier to doing pre-train. Post-training, or we call fine-tuning, is necessary for this industry. Phison's aiDAPTIV+ proof, our solution is the best for the fine-tuning. We can improve inference.
Week after, we are going to announce this solution, the newest solution in the GTC. Of course, we can see more our partner is going to engage with Phison from medical device, from the server business, from the industrial business. This we believe. Okay. Fine-tuning is coming, edge is coming, adaptive is the best solution. Okay. We also take the one example. According to the new series China Senior Year, they have a Professor Fei-Fei Li. She announced her lab doing the 1,000 training set, training under the DeepSeek R1, and cost is $50 in the cloud computing. Okay. Also they use 16 NVIDIA H100 GPU and cost is higher, but they spend 26 minutes.
Phison's team did the same things, fine-tuning the 1,000 set with the Professor from Yang Ming Chiao Tung University with the Adaptive, with the APIs of a GPU card, which is a 3% cost of what they use. The training cost is around TWD 500-TWD 600. This got to prove to fine-tune a model can be very, very low cost, affordable, and everyone, you get a server in your office, in your server room, you can do it by yourself. The trend is coming. We are also working with the hospital in Taiwan to doing the VLM, and this POC done and start with talking to how to move this to the clinic use. A lot of activity is ongoing. We see after this GTC, edge AI computing is coming. Let's have some summary.
Basically, last year, everyone talking about cloud AI, cloud AI, multiple billion, one full digit billion, three digit billion. This year, we believe we'll be talking about how many K dollars, which is go to the on-prem, go to the edge computing. Phison Adaptive, we work really aggressively with many of our industry partners. Hopefully after GTC, when the market start to award the edge is coming, we are going to deploy our solution to everywhere. On the same week, they also have an AMD GPU event in Beijing. Same week, one in San Jose, one in Beijing. Phison, we are going to join both there. Okay. 2024, we also, even though we tried to manage our headcount, the worldwide R&D headcount actually still increasing just because of we are eagerly to go into the enterprise SSD.
We are hiring more for the Adaptive. Due to the geopolitical conflict and customers asking to have R&D team other than Taiwan, the overhead wise is still around 21% expense from our revenue. To let you know where our distribution is, Taiwan still is a headquarter with about 3,000 plus engineers in Taiwan. In Japan, we have one subsidiary doing the customization for the local industry. In Malaysia, due to, again, asking by the U.S customers for Taiwan plus one program, we have around 60 engineers in Malaysia doing the enterprise SSD, automotive solution, and also the AI solutions. We also built another new entity in China Xiamen to deploy our enterprise business there. China is asking to use local content. Without our local designer, I do not think we are able to get the business.
The China, we have around 60 headcount doing the enterprise SSD for China home market. US in the San Jose and also the Broomfield, which is mainly for enterprise and also engagement with the cloud partners, system partner, GPU partner, CPU partner. No matter what, Phison already start to manage to control the headcount, growth of headcount. We are asking internal engineers start to use adaptive AI to improve the design, not to increase the headcount anymore. We can see some good progress. I believe second half, we can share more results to everyone. Let me have some summary. The NAND industrial start to cut production since last November. We can see the impact is happening. The whole channel, no matter in consumer or in industrial and the mobile phone, they are carry very low inventory.
Recently, we can see the good demand coming back. The DeepSeek proof, the AI, the LLM cost is not that expensive. Gen 1 is coming, it's only like 30B versus the DeepSeek 671B, only like 5% of the parameters, showing the same performance. It means you want to use LLM AI, you not necessary to use cloud. You can use the on-prem solution with affordable investment. [Foreign language] 这边做个总结，整年的这个去年的需求看起来毛利还是成长的。那也因应了去年下半年的通膨的关系，延长开始减产，那这个已经发生了。因为这个减产还有大陆地区的需求的关系，最近的整个需求对NAND的需求是蛮强劲的。我们现在的政策给客户，客户要求100万颗，我们大概只承诺给20万颗他的要求价格，另外80万颗，我们承诺会给他数量，但是价格必须要向上调整。这个是我们目前所看到的一个现实的情况。 Okay, this is today's report, and let me go to the Q&A collect from shareholders. This morning, we have a shareholders meeting. We decide to have a TWD 12 dividend for second half. Dividend is coming from the net profit. We took like 65%, so close to TWD 8.6. Last year, we have first half, we have a 20 share. We take out 55% back to the shareholders. Overall, it's a TWD 12.
By 2024, the whole dividend is 13 plus 12 is 25. Inventory, TWD 24.6 billion, which is a distribution here. We still try to accumulate more components for enterprise. We can see the demand is there. The embedded OEM for PC and mobile also getting stronger. This will help us to increase our revenue in this year. Controller-wise, we got the Gen 4, Gen 5 PC OEM, which recently we are a little bit short in the Gen 4 controller supply due to the earthquake here and due to the sudden increase from our NAND customers. Overall, we believe this year we can grow the whole business, but one only worry, one worry is second half, we may not have enough flash, even though we purchased a big amount of the NAND flash by Q1.
ESG-wise, Phison, we keep getting a lot of good improvement. We can read here, our ranking is keep improved. In the OTC, Phison now is a top five in the whole industry. We keep improving our ESG. We invest a lot using the green power, doing the society supporting, and also improve our internal transparency. This is our report today. We open to the Q&A. If anyone, you can ask by English or you want to use Chinese, I will reply accordingly. Thank you.
Yes, thanks KS for sharing the insights. Now we'll open the floor for Q&A. If you'd like to ask a question, please use the raise hand function in Webex or directly send the questions via the chat room. First, I would like to start with some questions for KS. Do we have a target for 2025's full year revenue and also the margin side? As we already see margin pick up in fourth quarter, do we expect it to continue to go up in the coming quarters? Thank you.
Phison, basically, we did not give the financial guidance to the market. I think it may be hard to me to answer. After answer, the OTC may call us to give the guidance. We are in the market. We agree NAND is a volatile market, but this whole year, we believe the NAND is very positive. We can see the demand in the, I mean, demand from a PC, from a mobile, from server, especially from the edge AI, getting stronger. Of course, we have a goal on revenue side, but we keep doing our best. Only worry is we may not have enough inventory by second half if the production cut starts to impact the CO2 supply. In gross margin wise, of course, last Q4, I think is a bottom gross margin. We can see the gross margin start slightly back due to the demand stronger.
We also are able to collect a big amount of wafer supply from suppliers at the Q1 with a much lower price in the market. Gross margin wise, we are also happy to see once the Adaptive demand is increasing. This definitely will help us to improve our gross margin rate because this includes the software licensing. Software, you may know, this is 100% gross margin. Overall revenue wise, 2025 will grow versus 2024. Margin wise, definitely will grow versus 2024. Okay.
Yes, thank you. Yes. Now we will open the floor to the analyst. Jason, feel free to unmute to ask questions. Thank you.
Thank you for taking—hello? Can you hear me?
Yes.
Thank you for taking my questions. My first question is in terms of the price in coming quarters. We know that some of the players, they are cutting the capacity or utilization rates currently. How do we expect the price impact in coming quarters? Do we see some of the wafer price or die price or even SSD module price or even finished good price can reverse in second quarters or even continue into third quarters? Thank you.
Pricing wise, after Lunar New Year, we have two NAND supplier increase the price around 9-11% on the wafer. What I like to say, this is happening. In the spot markets, spot market, the price already for some low density products already increased 30%. For the middle capacity, middle capacity still like only 5%. We believe in the highest density because high density will consume more flash memory. In the Q2, definitely price will increase more than two digits. The answer is the NAND supplier already increased the wafer price and the spot market deflate the price. In the retail SSD, we can see the leading key player already increased 5-7% last week. We believe the price trend will keep growing until Q3. Of course, again, we keep talking to the NAND supplier.
They have to increase the price back to their reasonable margin. If price going that fast, that much, same story may happen by 2023. This time, I think one thing is the assumption that 2023, they have no AI edge demand. In 2024, you can look to the China. China now start to push the edge server for DeepSeek. What's the AI? AI input is a power supply. Output is data. When AI start to use in the edge, the SSD in the edge demand will definitely increase. Actually, since last week, we can see the China demand on the high density SSD demand already happened. I still personally more optimistic. The price in the March, April, May definitely significantly increased. Hopefully, by Q3, price increase with a single digit, not that much. The whole year, I think still very positive. Okay.
Got it. Got it. Thank you. My second question is in terms of the impact from the U.S. new rules. I think a few months or a few weeks ago, U.S. government's restricted some of the NAND supply to China. That means they need to package it in a wide list vendors. Your competitor, especially in China, they are still using some of the 12 or 16 nanometers. We believe that some of the capacity for backend are in China. Do we see any positive new orders because of that kind of new rule happening this year, especially for your SSD controller business?
Again, this is a hard and sensitive to make a comment in the public. The answer is yes. The whole market worry the supply chain uncertainty, which is why I'm just saying that our Gen 4 controller was in high supply due to our NAND customer increase the controllers. Our US module house also increased the controllers. On the other hand, we also try to gain some share in China retail business. The answer is yes, but that's retail monitoring.
Okay. Thank you. My last question is in terms of your OPEX for this year. I know you never give the guidance for your financial numbers. Can you give us some trend or outlook that whether you will expand your R&D headcount or you probably will set up a new office or R&D center for this year? How do we expect this kind of OPEX for this year? Thank you.
Okay. Since early 2024, we start to manage the headcount increase in Taiwan. Basically, Taiwan whole year, the headcount increase is only like few 10, below 50. We are well managed. Taiwan office R&D start to use Adaptive to improve their design efficiency. In Taiwan, we do not see we are willing to keep increasing the R&D. In Japan, China, Malaysia, I think Japan is okay. We do not need to increase. For China market, especially for enterprise business, China, we need to expand because we need to take a share in China. Malaysia-wise, I think we are also going to increase because pushed by the US customers, they wish we can move some more design to Malaysia. Malaysia, we also can be able to cover the business not only from US, but also from Middle East.
Personally, I believe Malaysia team may go to a big event either this year or next year. We also got some business there. I do not much worry about increased headcount in Malaysia. China, we have to. We also got some local partner to support us. The answer is in Taiwan, we are controlled our headcount. We have no any interest to increase more R&D here. We try to use our AI to improve. By the way, we also have an India JV. India is going to hire more engineers, but it is under JV, not under Phison headcount. We are going to support them training the R&D in India doing the India business. That is all. This year, I do not think we have planned to build other R&D center.
In terms of the whole headcount in this year, how many % do you expect to increase for this year?
Taiwan, do not expect. I do not see Taiwan we are going to increase, but Malaysia and China may double. Double means from 60 to 120.
Got it. Got it. In terms of whole, I mean, employees or headcounts, how is that impact for this year?
Taiwan, I don't see Taiwan going to increase. It's low two digit.
Okay. Got it. Thank you. I have no more question. Thank you very much.
Thank you, Jason. Next, we'll have David. David, please unmute and ask a question. Thank you. Hey, David, we cannot hear you. Okay. Maybe we'll move to the next participant first. Hi, Michael, you can now unmute your mic and ask a question. Thank you. I'm sorry, there seems to be some issues with the microphone.
Okay, wait. Technical issue. David said he cannot he can't unmute. David, try again, David. David. Since not yet we have one on the chat board, let me answer here, okay? Talking about we have MOU with Huawei here, what is a potential future opportunity? Actually, we see a lot of medical research companies came to us to ask using AI improve the hospital efficiency, the clinic efficiency, also the medical new development, something like that. We can see the edge AI is ready to go to the edge because the investment for them is only like NT 3 million, it is affordable. The software-wise, it's only NT 1-2 million. They can use AI to improve their research. We can see more POC coming. Yeah. David still cannot unmute.
David, can you please send your question to the chat bot and I will help read it out. Maybe we can go to Michael first. KGI, Michael, you can now unmute your mic. Thank you.
Hello.
Yes.
Yeah, please.
Can you hear me?
Yes.
Yeah. This is Michael from KGI. Thank you for taking my question. I have two questions. The first one is related to the demand momentum the company is seeing on the customer side. We hear that you have recently raised price. I wonder, does that mean you are seeing a real big up in customer demand? If so, do you think that it is an organic recovery or are customers just restocking demand ahead of expected price rebounding? Thank you.
Okay. Demand-wise, I think the weaker happened by December and January. After the Lunar New Year, turned to really aggressive. We can see from mobile, from PC, from industry, also from some enterprise application getting strong. From AI, edge AI in China is crazy. They are trying to pull in more. The price increase is not pushing by us, it's by the NAND supplier. NAND supplier already announced to increase at 10%, 11%. We believe the price will keep going up because the demand turned to good. Again, I want to share with you this year, they have one demand which you may not figure out is the edge AI. If edge AI going to deploy in the market, then every one month, two months, three months, they have to buy the new SSD because AI output is data. I think this is organic.
Okay.
Got it. Got it. Thank you. I have another question is related to your inventory. The company's inventory days are still at historically high levels. Right now, as you mentioned, the industry is mainly seeing supply cuts. Since that, the market also expects the first quarter to be the worldwide peak for logic IC. If logic isn't doing well, memory demand probably won't be so well either. Given this, do you believe that your inventory can be reduced to a much more healthier level?
No. Actually, I'm worried about second half. You know the NAND supply start to cut production from November, right? It means the output will impact happen by March to May. Meaning supply in March-May will be declining. Second half, we believe the AI, edge AI demand, enterprise demand, and other gaming demand may come in back. We still try to secure some more supply now. Our position is if we buy in the second half, the price will be much higher.
Yeah.
Okay. Got it. Thank you. Thank you, KS. Very helpful.
Thank you.
Thank you, Michael. Now we have Simon from BOA. Hi, Simon. Please unmute and ask the question. Thank you.
Okay. Great. Thank you. Hi, KS. The first question is February revenue, $4.5 billion looks very strong, up 25% month to month. What was the key contributor? Is it price hike driven or just a volume increase?
Volume increase. The price impact is not happening that much. Price happened mainly in March, April, and May. February order, actually, we get order from December, January.
I see. Very clear. Any rough idea for the month of March, April, what percentage of the price hike? Media reported 10%, quoting your some comment.
Depends on the different capacity. For the low capacity, we see already 30% up. For the middle capacity, 64, 128, around 10%. Higher is not yet. Not yet. We believe may start to deflate the price increase by April.
Yeah. Yeah. If the NAND price recovers and then the demand recovers, don't you think the NAND chip makers will increase their fab utilization ratio? And then again, maybe NAND chip will be more supplied?
I believe so. This may happen by CQ2, then the supply increase by CQ3.
In that case, again, NAND chip price will fall back again?
Definitely, this does not happen in one night. One exception, potential exception, is CQ3, CQ4, edge AI deployment. This never happened in the market when they said if they have 1 million AI servers go to the market with 1 million extra enterprise SSD, you can imagine how many SRPs there.
Yeah. Yeah. Very quickly, you emphasized some edge AI. We truly understand the demand content increase along with maybe billion parameters, NPU or AP chips. We wonder why the OEMs need to input more NAND chips per box for smartphone or PC along with the edge AI progress.
Okay. This is a little bit complicated. If you award what Phison saying, the Adaptive. Okay? Adaptive is a solution to use a NAND flash as a complementary versus the HBM or VRAM. We believe the AI server to the edge, they need to have a fine-tuning function. You need fine-tuning, either you buy a lot of DRAM or you use a Phison Adaptive. DRAM price is much higher than the flash. Also the dimension. You want to plug in the DRAM up to 1 terabyte, you have to imagine how big the area of motherboard you need to occupy. Go through Phison Adaptive is simple. Just a single M.2, U.2, we can cover unlimited of a capacity. The price can lower a lot. AI related to the NAND, they have two portions. One is the Phison Adaptive.
Second is if using Adaptive to make the server price much lower than others' AI server, means easier to sell more server to the market. More people use AI, create more data, they need more SSD.
Yeah. How about the smartphone or PC? Why we need more Phison solutions for the smartphone?
No, no, no. No. Smartphone, storage in smartphone is just a storage, okay?
Yeah.
Nothing to do with Phison. If they use AI, I don't think smartphone need AI because the smartphone go to the telecom, telecom, the data center can provide the AI service. Smartphone is since last two quarters, price is so cheap. The new design in phone, they use a 256 configuration.
Yeah.
Okay? That means the total, the average of the carrier of a smartphone increasing. Nothing to do with us. Same as a PC. PC now, what we get the full case is mainly 1 terabyte and 512 gigabyte because price is so cheap. If go to second half, if price increase, they may consider back to the 512 on PC and 128 on the smartphone. This will be another way. Phison contribution on NAND is nothing to do with the PC or smartphone, but on the edge AI.
Yeah. Lastly, sir, would you recap your revenue mix for the direct export to the U.S. versus China sales?
First, I don't have this number in my hand, but I can see China business most likely go through our partner, go to some country, and we have a subsidiary, right? We believe our revenue distribution still in the U.S., mainly U.S., Southeast Asia, and go through other countries, but China is mainly go through our subsidiaries.
Subsidiary. In that case, you have to worry about the U.S. tariff on the U.S. product or no concern?
I don't know if I need to concern because if the tariff everywhere, what's the difference? If there are tariff for every country, so what's the difference?
Is your product considered as made in Taiwan mostly?
We are 99% in Taiwan or 99.9% in Taiwan.
Yeah. Taiwan.
Yeah.
Yeah. Okay. All clear. Thank you very much, KS.
Yeah. Okay. I see they have two questions in the chat bot. Let me. Hua, I'm ready. From David, please could you help understand your customer adaptive? I think China is an early mover in on-prem AI software. Explain how. Okay. China today, they are crazy on the DeepSeek on prem, but mainly is for inference. They're still not yet to ask the fine-tuning, but we already have two local partners to educate, to promote Adaptive to the China market. Next week in the China FMS event, I'm going to have a speech going to tell them we are ready for the on-prem fine-tuning. Okay? That is where you see. Our partner also engaged with some local partner who provide the server, who provide the service, who are the system integrator. We believe the second half Adaptive will be in China and very popular. Okay?
I will share in the enterprise SSD business outlook. Overall, we can see, okay, the cloud investment in the AI slightly slowed down. I mean, you wait for Microsoft, you can see, right? Again, second half from edge, from edge AI enterprise SSD demand definitely increased. We are also working with the field storage system supplier. We are designing the controller and the SSD for them. We believe second half the enterprise SSD demand and the portion investment will slightly increase. Brenda asking about China edge AI demand strong to corporate use close to edge AI. You know the so-called close loop. Actually, close loop defined by Phison, close loop AI. Yes, we are promoting our solution to them, but unfortunately, Phison do not have a good enough engineer to support them.
Next week, we are going to engage more partners in China to help make our solution popular in China. GB10, same positioning as Adaptive. No, GB10 is not. GB10 is using on the Project DIGITS , which is mainly for inference. As my understanding, they have 128 gigabyte DRAM in the system, which is fine-tuning may can only up to 13B or maybe less, maybe 8B. With the GB10, Phison is going to make the POC on top of GB10 with Adaptive. Let's see what happened after GTC. Okay? One more question from Kevin. Required installing Adaptive. Okay. Adaptive is a solution which is good for any legacy PC platform, workstation platform, server platform. Doing nothing change. Nothing change. Just adapt with the Phison Adaptive, install our middleware software that immediately can use. Nothing changed on the hardware. Okay? It is seamless and easy to use.
The only our challenge today is we don't have enough engineers to develop the software. We hope, we wish more software companies come to us to work with us, deploy their software on our Adaptive. We already signed up a few software companies in Taiwan. They started putting their software on Adaptive to serve the bank, to serve their medical service, to serve some office use, once to serve the financial report. We hope more software companies, no matter from Taiwan, from China, from Malaysia, from U.S., from India, to come to us. Okay? How to project revenue contribution for Adaptive enterprise SSDs in 2025? Again, we don't have the focus on our financial report, but we believe enterprise this year will be our main revenue generator. Adaptive will be our gross margin generator. NAND pricing recover start March and H2.
What will be gross margin from Q1 to Q4 this year? Again, we believe the gross margin definitely will significantly increase, but it's hard to say, right? Because we are going to increase our enterprise SSD revenue, which may a little bit lower margin, but Adaptive, which is a much higher margin. Again, it's hard to say, but I think everything can be more clear and better visibility by end CQ2. We can tell more at the end CQ2. Okay?
All right, KS. I think still some speakers would like to ask questions by microphone. Is that okay?
Yeah, it's good. We still have 10 minutes.
Okay. Sure. Simon, please have your questions asked. Thank you.
Oh, yeah. Yeah. Thank you very much again, KS. Very quickly, some investors are saying in China, there will be many memory chips. Do you expect lots of NAND chips will be supplied by Chinese local chip makers going forward?
Okay. I have no comment on DRAM market. I'm talking about only NAND. I have nothing worry about China NAND. Their market share is 8%. If they increase the 50% this year, their market share may go to the 12%. China occupies the consumption close to 40%. If China local content insists to use only the China flash, can fulfill only less than one third. I have no much worry about the China NAND supplier. By the way, if you go to the research in the market, the China NAND supplier, their component price is the highest in the market than any other Korea, U.S., or Japan supplier. Means they have a very strong market to support their NAND flash, and this is helping them to have a much healthy financial report, and they have no interest to cut the price.
You are saying China's maybe YMTC and NAND can represent only one third of the China demand as a revenue?
Based on the number, what I read from the book here. I have no much worry because China is still asking for local content, right?
Yeah. Their local price is also higher than international price.
Yes. Their offered wafer price is higher than any other one.
The China, a lot of the memory chip packaging testing finished good done here in Taiwan. Do you see China's big operation for packaging testing for the NAND product there or?
I'm not in this field, so it's hard to comment.
I see. Do you feel any impact from Japanese guys, I mean, the Western Digital NAND split, etc.?
I think SanDisk, right, SanDisk is now running independently. Hard drive is a good gross margin, not NAND flash, right? That is why SanDisk now turned to very conservative, and they announced a better price. They have to.
Your NAND chip sourcing is more Kioxia and Micron, not much?
No. No. No. We buy from every six supplier. Yeah. Korean, non-Korean supplier may around 35% of my total supplier.
35% from Korea, you mean?
Yes.
The rest maybe U.S. and China guys. All right. Lastly, yeah, so glad to hear around 80% payout ratio. Would you remind us the annual total dividend, second half, $12, what was the first half? Sorry, annual total?
13, 25. 25. Year 2024, total 25, 13 plus 12.
Oh, yeah, yeah, yeah. 13 plus 12 means 25 out of total EPS is 30 something.
38.
Such a high payout ratio, I mean, how much? 80%? Yeah, 80% payout ratio. That's your kind of the ongoing policy, 80% payout ratio?
No, it's 64%. 64%. 25 by 38.9 is 64%.
That's the.
Yeah. Phison, actually, we have confidence that we can run much better business in the coming years because we have other new business such as Adaptive, we're going to medical, we have automotive. We are also able to generate more cash flow. We are also happy to increase. In the past, we mainly asking 55% dividend from the net profit, but we hope we can keep slightly increase the dividend to our shareholders. Again, this we need to follow the market. Coming years, at least coming two years, we are quite optimistic to our business.
All clear, KS. Thank you very much.
Thank you. Yeah.
Yes. Thank you, Simon. We have another speaker. Jason, please feel free to unmute your mic. Thank you.
Thank you for taking my question again. My question is, I know that Phison's maintained a very good profitability or gross margin in last quarters and even during the NAND price downtrend. I think can KS provide us how you can maintain such a better gross margin compared with Phison's profitability in the past, even during the pricing downtrend? I think the major factor is because you already have a higher product mix in terms of the enterprise. I just wondered that if price is reversed in the coming quarters, but you already have a higher contribution or better product mix from enterprise. Will this kind of gross margins cannot improve just like pricing in the future? I mean, because you have very related stable margins because of the enterprise contributions.
If pricing is reversed up or improved, will your gross margin can only improve less than or increase less than the spot price or contract price?
Okay. If Phison today is still mainly with the USB memory card business, then forget it, okay? Every year, that's why Phison will keep have a high R&D investment. Invest about what? Invest on the new technology, new product for futures. Okay? To this, we enjoy the gross margin just because three years ago, we start to develop something new for our customers, which they're able they're also happy to pay the premium to us. By the way, Phison has a very good configuration of our product's high mix, okay? We cover almost every NAND application, almost every. I can share with you we have some modules which is 75% gross margin. Of course, some consumer module, maybe single-digit margin.
With the new business model from Adaptive AI server, we believe this will help us improve a lot of our gross margin rate because there we cover the software licensing and software development is 100% gross margin. In the future, Phison will keep investing more in the software development on top of Adaptive because we can see this is a real trend and this is a real business demand. By the way, enterprise SSD will help us to improve our revenue, but the gross margin wise is not that much attractive because it is a high dollar amount. Revenue will be from enterprise SSD. The gross margin mainly will be from Adaptive and other configuration industrial. The answer conclusion is how we keep our gross margin because of heavy R&D investment.
We are going to keep continuous R&D investment to make sure in the future we have all the new products that we can enjoy the higher premium from the market.
Got it. Got it. It means your gross margin is sustained by your new technology or innovation instead of the product mix or from enterprise.
Yes.
If you keep releasing the new product, you can keep your gross margin or even get it better.
Yes, very true.
Yeah. Got it. Got it. Thank you. I have no more questions. Thank you very much.
Okay. We still have some questions at the chat room side. The first one is from Mark. He asked, can the 2025 CQ2 revenue turn positive year on year this year?
Why? I don't know why last year. Let me see. What is our CQ2 last year? CQ2, Chinese or [Foreign language] Sa? CQ2. CQ2, sorry, I am okay. Let me check. Sorry, let me check with my CQ2 before. I think CQ2 is I think we have a confidence. Based on today's demand, I think we have a confidence. Okay? Yeah.
Okay. We have another question. Do you still see the structure uprising gross margin to 30% or even higher in the midterm? What would be your target mix percentage of gross margin for AI Adaptive in the next three to five years?
Of course, this is our goal to improve our gross margin rate, definitely. Okay? That is why we keep investing, heavy investing into the R&D. Again, this is relative to the whole market. Again, we are moving our products to the much higher premium market. We wish and we do the best to make it happen.
Okay. One last question is, how much does AI Adaptive contribute to total revenue?
Still small. Still small because more still under the POC. I wish this GTC, we can show something new to the market, then the Compute x is a time to call that order. By March GTC and the end of May Compute x, we can see how strong the demand on Adaptive.
Okay. I also have a question sent to me from the investor. Do we see any rush orders in fear of the US tariff concern? If so, do we think this will be sustainable into second half this year?
From Taiwan, I mean, the rush order from U.S. to Taiwan, I didn't see yet.
Got it. Okay. In the interest of time, I think we do not have any further questions now. We will conclude the earnings call here. Thank you for all your questions and your participation today. If you have any additional questions or require further information, please do not hesitate to reach out to Phison's investor relation team and us at Morgan Stanley. Thank you for joining us and have a great day ahead.
Thank you, everyone. Thank you.