Welcome to Nanya Technology's 2025 Second Quarter Earnings Conference Call. All lines are in the listen-only mode. The conference will be held only in English for investors around the world. Today's conference will be approximately 16 minutes. Nanya Technology's President, Dr. Pei-Ing Lee, will summarize our operations in the second quarter of 2025, followed by our guidance for next quarter and key messages. Then Nanya Technology's Executive Vice President, Dr. Lin-Chin Su, Vice President, Mr. Joseph Wu, and Financial Executive, Philip Rao, will join us as we open our question and answer sessions.
Today's presentation materials are available for download at Nanya Technology's website at www.nanya.com. As usual, we would like to remind everyone that today's discussions may contain forward-looking statements that are subject to significant risk and uncertainties, which could cause actual results to differ materially from those contained in the forward-looking statements. Please refer to the Safe Harbor notice that appears in our presentation slides. I will now turn the call over to Nanya Technology Corporation's President, Dr. Pei-Ing Lee, for the summary of operations and current quarter guidance. Dr. Lee, please begin.
Ladies and gentlemen, welcome to Nanya Technology Investor Conference. I'm Pei-Ing Lee. Before I start today's agenda, let me make a brief comment. For DRAM, it's cyclical in nature. Therefore, the market is always up and down. For AI market, it's been recovered since mid-last year, 2024. For non-AI business, which represents almost 90% or more of the bit shipment, still the majority of the DRAM market continued to remain downturn until recently, until the end of Q2. For April and May, the market remained very bad until the end of June. The market started to recover for non-AI related market. Our Q2 result pretty much represents what I described the market situation. Today, in my report, we'll start with Q2 2025 revenue and result, followed by CapEx and bit shipment, and then market outlook, and conclude by business review and outlook. First, revenue and result.
Our Q2 2025 result on the net sale comes to TWD 10.526 billion compared to Q1 at TWD 7.188 billion. It is an improvement of 46.4% in revenue. However, the gross profit comes to minus TWD 2.165 billion versus minus TWD 1.075 billion. And the operating income comes to minus TWD 4.5 billion compared to Q1 of minus TWD 3.155 billion. And EBITDA also comes to minus TWD 743 million compared to TWD 797 million positive in Q1. Non-operating income minus TWD 606 million, mostly due to exchange rate reason compared to Q1 plus TWD 732 million, with income tax benefit slightly favoring in Q2. The net income comes to minus TWD 4.109 billion versus Q1 of minus TWD 1.941 billion. And the earnings per share comes to minus 1.32 versus Q1 of minus 0.63. And the book value per share comes to TWD 49.91.
For summary of quarterly result, first of all, revenue improved by 46.4% QoQ, and year-to-year also improved by 6.1%. For the shipment, QoQ improved by approximately 70% compared to QoQ, and then YoY also increased approximately 30%. ASP QoQ decreased by mid-single digit as I explained it in the very beginning of the comment, and to year-to-year decreased by mid-teens. Exchange rate decreased by mid-single digit, and year-to-year decreased by mid-single digit as well. For the QoQ result comparison with discussion and explanation, net sales TWD 10.526 billion versus Q1 of TWD 7.11 billion, and improvement of QoQ 46.4%, mostly due to bit shipment increased by approximately 70%. ASP decreased by mid-single digit, and exchange rate also unfavorable by mid-single digit.
The reason why ASP decreased by mid-single digit is as I explained that April and May, the ASP continued to slide down until June started to recover. Gross profit comes to minus TWD 2.165 billion versus Q1 of minus TWD 1.07 billion, and the gross loss increased by TWD 1.09 billion, mainly due to the lower ASP mid-single digit and unfavorable exchange rate. Operating expense TWD 2.336 billion versus TWD 2.081 billion in Q1 increased of TWD 256 million, mainly due to higher R&D expense. Our operating income comes to minus TWD 4.5 billion versus Q1 of TWD 3.155 billion. Operating loss increased by TWD 1.346 billion, and due to lower ASP, unfavorable exchange rate, and increase in R&D expense. Net income comes to minus TWD 4.109 billion versus Q1 of minus TWD 1.941 billion. The net loss increased by TWD 2.168 billion.
Exchange rate accountable for TWD 1.124 billion unfavorable, with income tax favorable by TWD 516 million. For operating expense, the left-hand side SG&A TWD 625 million, it is pretty much in normal range. And R&D expense at TWD 1.711 billion is also in the normal range, however, compared to Q1, which is relatively low, increased by TWD 256 million. For cash flows situation, the beginning balance for Q2 is TWD 62.603 billion, and the end balance comes to TWD 52.545 billion, with free cash flow of minus TWD 3.7 billion. The cash from operating activity is minus TWD 899 million, and capital expenditure is minus TWD 2.84 billion, and financial activities and others minus TWD 6.318 billion. Please pay attention to the bottom note.
In footnote one at the start, we have unfavorable exchange rate on cash and cash equivalent of up to $5.2 billion. This is as a result of the Taiwanese dollars being inflated. The net cash equivalent comes to 19.1 billion TWD. This is the net cash minus net debt, short-term debt and long-term debt comes to net cash of 19.1 billion TWD. For CapEx and bit shipment, in 2025, we targeted CapEx of the whole year 19.6 billion TWD, and what is already spent is 9.3 billion TWD. On the right-hand side, bit shipment, our Q2 bit shipment is increased by near 70% QoQ, and we are expecting the whole year increase by over 40% year-to-year. For market outlook, first of all, AI adoption in a cloud data center, which mostly HBM, continues to drive demand in 2025.
As I explained to you that non-AI market continues to decline until bottom of Q2, basically non-AI application bottom out at the end of Q2. Overall, DRAM market is expected to maintain relatively stable in second half 2025. From the supplier side, the major supplier, DRAM supplier, accelerate their advanced node and enhance HBM and DDR5 and low-power DDR5 positioning. So currently, there's a transition of DDR4 and low-power DDR4 to DDR5 and low-power DDR5. As a result, DDR4 and low-power DDR4 supply has been reduced. From demand side, for server, AI servers continue to drive HBM and high-end RDIMM to support DRAM growth. From the mobile side, AI smartphone upgrade boosts demand for high-density low-power DRAM. For PC, initial AI PC demand primarily coming from enterprise, while conventional PC demand remains conservative.
For consumer side, DDR4 and low-power DDR4 usage decline in PC and smartphone and increasing adoption in consumer application. For business review and outlook, for Nanya Technology's 1B node 16 gigabit DDR5 at speed of 5600 and 8 gigabit DDR4 is already in volume production. Our 16 gigabit DDR5 at 6400 speed is in engineering sample, and our development for 3D IC and TSV and DDP process is verified, and we are optimizing our DDR4 and low-power DDR4 supply to meet our customer demand. We continue to improve and expect across operation and shipment and inventory improvement in the second half of this year. We also received ESG recognition, top 5% in corporate governance among Taiwan-based companies. With that, conclude my presentation. Now we may move to Q&A section.
Yes, thank you, Dr. Lee. Before we begin the Q&A session, I would like to remind everyone to limit your questions to two at a time to allow all participants an opportunity to ask questions. We'll begin taking questions from dial-ins. For webcast participants, please message your questions with your name and company name to Nanya operator in chat box. Now, for dial-in participants, please press star one on your keypad if they would like to ask questions. To cancel your question, please press star two. As a reminder, it is greatly appreciated that you turned off speakerphone mode of your device to prevent possible echo effect. We thank you for your cooperation. Now, please press star key and number one on your keypad if you would like to ask questions. Thank you. Firstly, we'll take Charlie Chan, Morgan Stanley for questions. Go ahead, please.
Hi, Dr. Lee, and also Joseph, for the introduction and thanks for taking my question. So I think the first question is about the DDR4 kind of shortage issue. So how bad is the situation, and do you think it's going to be eased in the quarter? And also, I wanted to know how much can Nanya Technology benefit, right? Because Dr. Lee just said that the price was down in April and May until June started to pick up, right? But it seems like spot price or your industry in Taiwan saying that the spot price is essentially going up every month into the quarter. So I just want to get a sense of what could be the discrepancy between your price trend and your Taiwan industry peers.
Okay, Charlie, thank you for your question. Regarding the DDR4 and also low-power DDR4, as I explained, that is in transition from DDR4 to DDR5. And this has happened historically for every generation. Okay? That's including previously DDR3 to DDR4, DDR2 to DDR3, and low power as well. Okay? So that's the nature of transition of the market. And as I explained to you that the non-AI market is still accountable for more than 90% of the overall market in the bit shipment point of view. So still quite important for all the suppliers. Okay? And as a result, therefore there must be some adjustment in terms of the future product, etc., HBM, low-power DDR5, DDR5 to position in the marketplace. Okay?
So DDR4 and low-power DDR4 end of life or gradually reducing its supply in certain markets is happening naturally. Okay? And it may prolong for one quarter or more, etc. However, eventually, this is going to naturally happen. Okay? And from Nanya's point of view, Nanya has been a longevity supplier for all the product market. That's including DDR4, DDR3, and even DDR2. We are currently still supplying DDR2, low-power DDR2, and low-power DDR3, etc. Okay?
So Nanya will continue to provide our customer whoever has a need for DDR4 and low-power DDR4. And currently, our customers are quite enthusiastic. Okay? And then for those customers may be looking for even longer supply. Okay? We will do our best to meet our customers' demand. I don't know if that is a specific answer to your question, but DDR4 and low-power DDR4 recovery is making a big help to the overall market. The non-AI-related market, including DDR5, DDR3, low power, they all have some degree of recovery as well.
I see. Yeah. So maybe a quick follow-up is that do you think that it's going to be a shortage in the quarter as well? And do you think it's because customers are there double booking, triple booking, or is the real demand bigger than supply?
Well, for DDR4 and low-power DDR4, what we see is that the demand is real demand, okay, for Q3, Q4, for Q3, Q4 confidently.
Okay. Okay. Gotcha. And next question is more financial-related. So maybe Joseph or Philip, if you can help us. Because from your financial results, the 2Q EBITDA margin is negative 11%, and that compares to 1Q was a positive 7%. So if I try to break down the potential impacts, currently maybe 5 percentage points impact and ASP 5% impact, but that still cannot explain why your EBITDA margin is down so much. Especially demand recovers, fab utilization should also improve, and that should be positive to your gross margin or EBITDA margin. So can you help us to understand why EBITDA margin was declined so much in 2Q?
EBITDA margin, I think two reasons. One is our cash flow operation activity. Okay? It's not as good. And also, very importantly, the exchange rate also made a pretty big impact.
Hi, Charlie.
This is Philip. I think the main reason is if you look at our operating income, it's -TWD 4.5 billion. Okay? And if you add back the depreciation, it's still negative in the second quarter. Okay? Okay. But Charlie, how about I'm sorry. Please go ahead.
Okay. Then I understand. But how about fab utilization? So usually when you have an idle capacity, you have some penalty to your gross margin, but how about 2Q? I would assume fab utilization should be getting full, given customers' rush orders.
Yeah. In terms of the capacity cut in Q2, actually, it was gradually reduced, and at June, we are resuming full production. And in April and May, we have some minor reduction.
I see. Okay. And so just one very quick one. I think people would care about your third-quarter financial outlook. Would you be confident that the third-quarter EPS can turn positive?
I would say the EPS will make a very significant improvement. Okay? And gross margin-wise, confidently will be turning positive. Net margin-wise, we still have some room to work for. We need to make our effort. Okay? So for Q3, the situation is that we will see very substantial improvement in gross margin. Okay? In terms of net margin, more likely in Q4, okay, that could be more possible, more confidence in Q4. But Q3 is totally without possibility, okay, Q3. But still some room to make improvement for. Still a lot of effort needed.
Yeah. So gross margin to turn positive, meaning 20 percentage points improvement. And in two-Q effect, right, continue to impact gross margin still. Does that mean that your ASP can grow like 30%-40% in the first quarter? Third quarter. Yeah. Third quarter.
At this point, if you look at the market price trend, okay, that's not totally impossible. Okay? However. Oh, okay. Great. However, I cannot give you a specific number for those forecasts.
Okay. But that inference directionally is right, right? You have effects going.
So as I say, gross margin turn positive is very confident. Okay? And beyond that is a lot of effort that we need to make.
Sure. Sure. Okay. Understood. I will be back to the queue. Thanks.
Thank you. Please press star key and number one on your keypad if you would like to ask questions. Thank you. Next question, Jay Kwon, JP Morgan. Go ahead, please.
Thank you for taking my question, Dr. Lee. I have two questions. The first one is, relating to the initial discussion point, can we split the specific impact from the FX headwind? So NT dollar has definitely appreciated against USD, especially into June when your ASP started to pick up. So I do think it impacted June quite a bit. So can we have a separate FX impact instead of just a single-digit down remark, possibly, just to help us understand the gross margin impact better? That's my first question. Thank you.
Okay. The impact by the exchange rate comes from different areas, including in the revenue, but also including the asset. Okay? And also including the cash. Okay? So from the cash side, we try to minimize the impact by the exchange rate by having most of our cash is in US dollars. Okay? There is some impact, but cash impact has been minimized. It's mostly in the net value, and that is shown in our cash flow situation. Okay?
And as I explained, our cash flow due to the exchange rate impact is up to TWD 5.2 billion. That's the reason why exchange rate impact is big in all areas that I described. First of all, in revenue, second in asset, and third in cash. Okay? So what you see, the 5% is mostly just in the revenue side. Okay? And then it comes to the bottom on the net value, on the net margin point, I mentioned that from quarter to quarter, the exchange rate actually hurting us by TWD 1.124 billion. I don't know that explanation is good enough for you.
No, it's fairly reasonable, sir. Thank you. My second question is, you used to share us DDR5 wafer input, and also the bit shipment mix regularly in the past. So could you please share what was the result in the second quarter? And also possibly probably recap the first-half result. And also, could you please share your year-end plan or full-year plan? Given your bit growth has been revised up from 30% or more to 40% or more, I just wonder if this also includes a higher output from DDR5, or does it mean you're selling out more DDR4 and DDR3 from your inventories? I want to know a little bit more details on this. Thank you.
Yes. We now since have a portion of our capacity shipped from 20-nanometer to our second generation of processing node we call 1B. Okay? And that capacity on 1B is currently producing both DDR5 and DDR4, and we'll be producing low-power DDR4 and low-power DDR5. So that capacity for our second generation is going to be multiple products. Overall speaking, we're still expecting our 1B node, the new technology, will account for more than 30% of our output by the end of this year. We're still going to. We are still expecting that. Okay? And the DDR5-wise and DDR4 ratio in our new generation of product is adjustable. Okay?
Which means that we have a flexibility of making adjustments according to our customer's demand. At this moment, likely the customer may be more demanding on DDR4 delivery than DDR5. Okay? So we would make certain adjustments according to what our customer demands. That is the new technology, processing node technology. On the existing technology node, the 20 nanometer, we still have a number of DDR4, 8 gigabit, and 4 gigabit DDR4, as well as some DDR3, some low-power DDR4, some low-power DDR3, accounting for the rest of 20 nanometer. And again, those capacities also adjust according to customer demand and the market situation.
Thank you, sir. Just two more follow-up. May I ask then the first-half output of these 1B nanometer in terms of your bit shipment output production or bit shipment output? That's my first question. And second is, you said customers may demand for more DDR4 over DDR5 in light of this tight situation. So does it necessarily mean that Nanya Technology may be prioritizing DDR4 instead of DDR5? When it comes to that decision, would you be willing to take that decision even if DDR4 gross margin is low? Or does it mean that you are guaranteed with somewhat similar margin by the customers to that of a DDR5, that you are producing DDR4 more? May I ask about just the opportunity cost and also your return on metrics from your production selection? Thank you.
At this moment, based on the market situation, okay, the DDR4 looks slightly favorable than DDR5. However, DDR5 is still a growing market. Okay? And so we have certain commitment to our DDR5 customer. So we have certain commitment to our DDR4 customer. Okay? Therefore, as I said, for those we are committed, we will deliver. For those it's not committed, we will have flexible to make adjustments.
Thank you.
Thank you. Next we'll have Simon Woo, Bank of America, for questions. Go ahead, please.
Okay. Thank you, Dr. Lee, and Joseph. Number one question is, your ASP still maybe half quarterly basis contract, the other half maybe monthly? And then how about your maybe daily spot sales, spot market sales? Would you share the rough idea?
Okay. That's quite interesting question, Simon. We have some quarterly as well as monthly. Okay? And of course, some spot as well, but not as much. Okay? Small quantity. Okay. The ratio of quarterly and monthly very much depends on the market situation. Normally, when the market is uptrend or downtrend, customer will come with a little bit different concept. For the downtrend, customer may want to have more opportunity of adjusting the price. So likely the monthly will be more than the quarterly. But still, they are long-term customer. They require that make a commitment in a quarter or make in a half year.
That's still there. So our Q2 result is also very much impacted by two factors. One factor, as I explained in the very beginning, that April and May, the market price continued to slide down for the non-AI market. That's including every product, DDR4, DDR3, DDR5, everything. Okay? Even low-power. Okay? And then that's one reason. April and May, still pretty bad and even worse than Q1. Okay? And that's one reason. Second reason is that we still have some long-term customer, committed customer. They require quantities and long-term commitment in quarterly or even half year. Those we still have to make commitment and we still deliver our commitment. And those also impact on our Q2 result. Okay? Basically, that was negotiated at a market downturn. Okay? So therefore, those pricing is not necessarily very good. Okay?
And on the other hand, upcoming for Q3, because of market upturn, we were able to negotiate more long-term, more quarterly contracts as a result because upturn customers want to fix their volume more than the downturn. Okay? So that's the current situation. Okay? And there's no clear cut on when quarter more than month or month more than quarter. It's mostly market-related. I don't know if that answers your question or not.
Yeah. Yeah. Yeah. Very, very clear, sir. And then quickly, yes, 1B nodes can be used for DDR5 and DDR4. Sounds good. But the 1B is still what? As of July, only 10%, 20%, and then 30% for the end of 2025? Is your point?
In terms of the capacity-wise, 1B, in terms of wafer capacity, is already over 30% of our production. I see. How about? It will produce more bits right away. Yeah.
So you are saying 1B nodes currently 30% of the wafer capacity? Yeah. Wafer capacity around mid-50, 55, or 60K, or?
Yeah. Mid-55. Yeah. Maybe 55 to 60 depends on the product portfolio and technology portfolio. Yeah.
That means that I remember you don't have much 1A nodes. So after 1B, the majority is 20nm nodes above, right?
I'm sorry. I didn't understand your question. So 1A nodes around the top? Oh. Oh. Oh. You mean the 1A nodes we already phased out end of life?
Yeah. Yeah. No more 1A nodes.
No more 1A. Yeah.
So that means about 70, 70% your capacity is 20nm nodes above old one.
Yeah. However, that's enough account for 1B will produce more bits.
Oh, I see. That's the way it works. So in terms of the bit, when the 1B node can be more than 50-50% of the total your bit production?
Likely in Q4, the bit production will be about even.
I see. Very clear, sir. And meanwhile, your capacity is very fungible. So if a customer asks more DDR4, low-power DDR4, you can keep using 1B nodes, which can be more cost competitive versus the 20 nanometer nodes.
Yeah. There are some flexibility for us to make adjustments, both in 20 nanometer and 1B nodes. Yeah.
Yeah. 1B means easily what? 50-50% better cost versus 20 nanometer nodes or?
I'm sorry. I didn't quite follow your question.
About 60%. 1B nodes must be very cost-effective versus the 20 nanometer nodes. So per bit basis, 1B nodes should offer easily 50-50% lower cost per bit.
Oh, no. No. I think cost-wise, it's not that much. Okay? Cost-wise, it's only slight advantage when it's in a mature situation. This is very similar. This is very similar in all suppliers because 1B node, it produces more. However, it also has much longer process steps, and cost is also higher. Okay? So as a result, you get some cost advantage, but not very much. Okay? And I would say this is very similar in all suppliers. However, 1B allows you to build higher density. For example, you can move into 16 gigabit. Also, you can build DDR5. Okay? And for DDR4, you also can build smaller die, which may, in terms of the floor plan, they can fit into much smaller floor plan. Yeah. For PCB, for customer requirement.
Yeah. Very good. One last thing, sir. Second half outlook good. DDR4 demand getting stronger. So that means your overall bit basis production volume, the DDR4, including low-power DDR4, DDR4 can be higher than the 50%, 50. So second half. If you include low-power DDR4, it's already way over 50%. Yeah. DDR4 and the low-power DDR4 altogether,
more than 50%. Yeah.
For second half this year.
Yeah. Even in June, we already achieved that. Yeah.
No. Sorry. I think maybe opposite way then.
Yes. Yes.
DDR5. DDR4.
You are right. In the second half, the DDR4 and low-power DDR4 will be over 50%. Yeah.
That means DDR5 is still lower than 50%. DDR5 will be much lower.
Yeah. Much lower. Okay.
How about the second quarter result then? Roughly, the DDR4, around what? 70-80%? What's the percentage of the DDR4?
Over 40%. Yeah. Over 50%.
DDR4.
Yeah.
Sorry, sorry. DDR4 only 40% or sorry, I didn't get it.
Over 50%.
Oh, over 50%.
Right. Right.
So DDR5 will be maybe what? 30% range or?
In the teens.
The teens. I see. In teens. And the rest around 30-40% still DDR3 then.
DDR3 plus low power
Yeah. Plus low power around 30%.
Low power is around 15%. 15%. Yeah.
So DDR3 and the low-power DDR3 together mid-teens, 15%, 15.
15%. Basically, low power and DDR3 are about same.
Okay. DDR3 plus low-power DDR3 together, altogether, 15% or 30%?
All the low power together. All the low power together around 15%. And all the DDR3 together. DDR3 including 4 gig, 2 gig, 1 gig, etc., is about 15%.
And that's. And then the DDR4.
Yeah. And Simon, that ratio may be adjusting also month by month.
Okay. Yeah. Yeah. Yeah. Anyway. For the future. So as of the second quarter, the mainstream product still DDR4, including low power.
Yes. And some DDR5, some DDR3, we still maintain as a longevity supplier to customer who require DDR3 as well.
Yeah. Yeah. So it sounds that the DDR5 around the teens, and then DDR3 also mid-teens. So similar ratio, DDR5 versus the DDR3.
Yeah. Yeah.
All clear, sir. Thank you so much. Hopefully, you can deliver very strong Q3 results. Don't get stressed, but thank you, sir.
Thank you, Simon. And thank you for your nice comment. Appreciate it.
Thank you, sir.
We're now in question and answer session. If you would like to ask the question, please press star key and number one on your keypad. Thank you. Next question, Charlie Chan, Morgan Stanley. Go ahead, please.
Sure. So thanks for taking my follow-up question. And just like the last quarter, definitely can we take your wafer and wafer these developments? And if yes, any of the contribution percentage in 2026 or 2027? Thank you.
What do you mean by wafer to wafer business? You mean for the AI-related business?
Yeah. Oh, okay. Yeah. Because there's a high-speed bandwidth.
Yeah.
Yes. Low-density type of, yeah, wafer stacking.
What? Demand. Okay. What we're targeting is not low-density. Okay? We're targeting is high-density because for whatever AI application, it will require a lot of data computation and data communication. Okay? So that requires very high-density. For very low-density to do the AI function, it's going to be very limited. And as a matter of fact, it's not necessarily the full AI function we're talking about. Okay? Partial AI is possible. Okay? So your question is about when that's going to be happening. We are working on that.
Okay? And as I mentioned in one point in my presentation is that we already have 3D IC, okay, DDP process verified. Okay? And that not necessarily is wafer-to-wafer bonding. It's chip-to-chip bonding. But that's the beginning process of what's preparing to get ourselves prepared for the future. Okay? Started with chip-to-chip bonding and then gradually moving to wafer-to-wafer bonding. Okay? And we are expecting that likely we will start to have some business, okay, likely and hopefully by the end of next year, okay, and into 2027.
Great. Yeah. So that's the only question I want to follow up. Thanks, Dr. Lee. Thank you.
Thank you.
We thank you for your questions. Let's move on to the webcast questions. And Dr. Lee, please begin.
Okay. I got the first web question coming from KGI Securities by Michael Shen. Michael has two questions.
The first question is, what is the ratio of the company's pricing based on the contract and spot? And for contract, how much is the monthly and quarterly? And actually, I explained this question in previous many questions by Simon, okay, and also mostly by Simon. Okay. The quarterly and monthly is basically what mostly our sale. Okay? We also have some half-year and yearly contract as well. However, the spot, we only do very limited spot. Okay? And at this moment, we can see that more customers demanding longer-term contract instead of short-term contract. And that is evolving, changing. And your question is, second question is, is the company concerned that with the DDR4 pricing higher than DDR5, there could be a major shift in PC and server demand towards DDR5 leading to the sharp drop in DDR4 demand and the price collapse in 2026?
I don't expect that to happen. Okay? At least confidently, Q3 and Q4, certainly we won't be seeing that happening. Okay? And one thing I have to explain is that for PC and server, the customer are shipping to DDR5 and low-power DDR5 as well. Okay? Mostly DDR5. Okay? That's already happening for a long time. However, still some remaining customer require DDR4 supply. Okay? Mostly in server side as well as in PC side as well. There will be some customer continue to use DDR4 for the reason of their system. Okay? That's still will not go down to completely zero. And on top of that, I explained in my presentation is that DDR4 and low-power DDR4, now they're becoming more adoption, becoming more mainstream usage in the consumer.
One thing we have to notice that when you use one DDR4 8-gigabit versus when you need to move to DDR5, you need to move to 16-gigabit, and one 16-gigabit DDR5, even though with 8-gigabit DDR4 pricing being going up, still higher than 8-gigabit DDR4, so from the BOM cost point of view, there's still advantage of 8-gigabit DDR4. Okay? However, customer will continue to look for the solution as well. Okay? So there are some certain applications will require 8-gigabit DDR4, 4-gigabit DDR4 as well. Okay? So I don't expect the DDR4 pricing to collapse. Okay? On top of that is that, as I say, DDR4 and DDR5 transition is happening. It has happened historically always from DDR3 to DDR4, DDR2 to DDR3, and those transition is not necessarily seeing the more legacy product collapse. Okay?
It may adjust, okay, moderately, okay, but not necessarily collapse. And the next question comes from Yuanta Securities by Michael Hsu. Four questions from Michael. Michael's first question is, could you elaborate further on COGS of 2Q25? Why GM is go down due to Q? And actually, I explained it very clearly just now a couple of times already. The GM goes down for several reasons. One reason is price go down. Second reason is exchange rate not favoring. And the price go down, as I explained also in my previous explanation discussion, is price go down because of in Q2, April and May, price continued to go down further compared to Q1. And the trend only start to recover at June. Okay? So in average, Q2 average price still not as good as Q1. Okay? So the price marginally go down Q2 versus Q1.
And that situation will be improved substantially. And the second reason for that is some of the Q2 ASP was negotiated in Q1 when Q1, the market is pretty bad, downtrend. Therefore, those long-term contract price, quarterly price is also not good. Okay? And that situation will be improved in Q3. As I explained, the quarterly price negotiation, the more customer demand, longer demand for longer supply. Okay? That's already happened. And we are expecting the gross margin has confidently will be turning positive in Q3. Okay? And the net margin that will require some hard working, some effort. The second question, what is the revenue contribution of DDR5 production in first half? In Q2, as we already discussed, the DDR5 contribution is in teens. And this is the result of the more demand on DDR4 as well.
In 2025, again, the second half will very much depend on customer demand and customer commitment on either market, DDR4, low-power DDR4, or DDR5. Whatever commitment we meet, we met, we will deliver. Q3, your question three, what is the DDR5 verification progress of new customer? That's going on quite nicely. Okay? We are currently deliver our 16 gigabit DDR5. We are sampling, and the current delivery of DDR5 is at 5600 speed. We are sampling DDR5 also in 6400 speed. Your question number four, CapEx and total capacity planning for 2026. We had a CapEx on 2025 of TWD 19.6 billion. 2026, we're still working on it and still subject to our full approval. How much wafer capacity are expected to increase in 2025 and 2026? Basically, wafer capacity will remain pretty much similar between 2025 and 2026. Okay.
We have come to the next question also from Yuanta Securities by Joyce Tang. You have one question. How does the company view the increase in the contract price in 3Q? Or what would be the percentage of contract would be the company's negotiated in 3Q? Okay. As I described to you, the long-term contract has been increasing. Okay? However, unfortunately, I cannot disclose to you the detail of the percentage. Pardon me for that. Okay. That pretty much ends all the questions on the website. Okay? And thank you so much for attending today's investor conference. And thank you for your support to Nanya Technology. Bye-bye.
Thank you, Dr. Lee. And ladies and gentlemen, that concludes our conference call today. Please be advised that the replay of the conference will be accessible within three hours from now, which will be available through Nanya Technology's website at www.nanya.com. We hope you will join us again next quarter. Thank you for your participation and have a wonderful day. You may disconnect online at this time. Thank you and goodbye.