Ladies and gentlemen, welcome to Nanya Technology Q3 2026 investor conference. Today, I'd like to present to you our Q3 revenue and result, and followed by capex and bit shipment, then market outlook, and then conclude by business review and outlook.

For Q3 revenue and result: our Q3 net sales comes to NT$113.6 billion versus Q2 NT$82.5 billion is Q-to-Q increase of 61.9%, and gross profit NT$114.4 billion versus NT$65.6 billion, an improvement of 74.5%. Operating income NT$106.8 billion at 80% versus Q2 60.8, improvement of 75.7%, and EBITDA of NT$110 billion, non-operating income at NT$2.4 billion, and we pay income tax for around NT$21.7 billion.

This comes to net income for Q3 NT$87.6 billion at net margin rate of 65.6% versus Q2 is an improvement of 74.6%. Earning per share for Q3 come to $25.4 per share compared to Q2 as $14.66 per share.

Our book value at Q3 $118.99 per share versus Q2 $93.45 per share. For comparison, our Q-to-Q comparison, revenue Q-to-Q up by 61.9% and year-to-year up by 611.7%.

ASP increased by greater than 60% Q-to-Q and year-to-year greater than 500%. Our shipment decreased by mid-single-digit due to depletion in inventory, and exchange rate favorable by low single-digit.

A little bit more detail explanation for Q2 and Q3 result comparison: the net income NT$133.6 billion versus NT$82.5 billion, increased by 61.9%, mostly due to ASP increase by more than 60%. If you look down to the bottom at the notes, you find that the ASP increase is primarily reflect renewed long-term agreement.

The bit shipment decreased by single-digit, as explained, is depleted inventory. Exchange rate favorable, low single-digit.

For gross profit, NT$114.4 billion at gross margin of 85.7% compared to Q2 is an improvement of NT$48.9 billion due to higher ASP. For operating expense, NT$7.588 billion versus Q2 NT$4.793 billion.

This R&D expense increased by NT$1.9 billion, and SG&A expense increased by NT$0.9 billion. For the note, most of this increase is mainly due to concurrent and future employee compensation.

For operating income, NT$106.89 billion versus Q2 NT$60.8 billion. Margin-wise increased by 6.3%, and operating income increased by NT$46.1 billion.

Net income NT$87.6 billion versus Q2 of NT$50.1 billion. The net income increased by NT$37.4 billion, and the income tax also increased by NT$9.6 billion.

For operating expense on the left-hand side of this chart, the SG&A expense for Q3, it comes to NT$2.443 billion versus Q2 1.5 and Q1 1.09, a substantial increase mainly due to concurrent and future employee compensation. On the right-hand side, R&D expense also increased from beginning of the year NT$2.1 billion to Q2 NT$3.2 billion and to Q3 at NT$5.145 billion.

This is again increase in increase, mainly due to concurrent and future employee compensation. For cash flow situation, our Q3 beginning balance at NT$217.6 billion and the end balance at NT$288.5 billion, with free cash flow of NT$75.2 billion.

During this quarter, the cash from operating activity NT$96 billion, and capital expenditure NT$20.7 billion, and with financial activity NT$4.3 billion. Okay.

In Q3, the cash activity -4.3, mainly due to dividend payout at NT$4.6 billion. We have dividend income also due to PCB stock sale.

If you look at the right-hand side, we have beginning balance of the NT$217 billion, end balance 288, this including all the cash from operating activity, capital expenditure, and financial activity, et cetera. For net cash situation, Q3 our net cash, which at the bottom notes, comes to NT$270 billion.

This is the cash minus debts, okay, cash of 288 minus debts of 18.5, okay. This compared to Q2, our net cash is NT$199.2 billion.

It's an improvement of around NT$70 billion in net cash. Our capex and bit shipment report here on the left-hand side: our capex for this year is target for NT$69.7 billion.

Far, Q1 to Q3 capex was NT$27.6 billion. This is the timeframe of starting to move in equipment for new fab.

This year is planned for NT$69.7 billion, with wafer capex accountable for around 50% of this amount. Our new fab capex will be NT$346 billion for 35.9k wafer start per month.

Currently, inventory level is low, bit shipment likely to stay flat from now on for a few quarters. For 2026 bit shipment, year-to-year comparison is about up by high-teen percent.

For market outlook: the overall market, the AI computation, that's including GPU, TPU, CPU, ASIC, dominate the DRAM demand growth. Memory supply to smartphone, PC, and general consumer were suppressed.

Suppliers' long-term capacity plan based on LTA commitments. For the supply side, the new fab construction, equipment delivery, and process node introduction all require longer lead time.

LTA enable efficient product and capacity planning, supporting future market stability. Currently, we're seeing lean supply-side inventory.

From demand point of view, AI-driven: expanding AI application are driving the memory demand for higher content, for higher bandwidth, and customized architectures. AI drives DRAM demand beyond HBM and low-power DDR5 across networking, storage, and system management.

For general application, the supply shortage and higher ASP and result in down specification and extended life cycles. Also, recycle DRAM partially support low-end market.

For business review and outlook: for Nanya, our Q3 net profit NT$87.6 billion at EPS of $25.4 per share, and our gross margin at 85.7% and net margin at 65.6%. Our Q1 to Q3 net profit comes to NT$163.9 billion with EPS accumulated to $49.28 per share for 9 months.

Our shipment may stay flat until new fab wafer output at second half of 2027. Our operational result likely to be sustainable by favorable ASP may continue.

For operational update and outlook: AI infrastructure and server contribute to Nanya's 20%, more than 20% of our revenue. Flat shipment until new fab wafer output in second half 2027.

Our new fab equipment is scheduled for move-in Q1 '27. We plan to ramp up to 30k by the end of 2028, and target 50, up to 50k by the end of 2030.

Generally, the new fab plan and adding the equipment is aligned with future AI project under LTA. 3D wafer packaging and test fab is planned for Pingtung Science Park.

Our 1C customer sampling is targeted for Q1 next year. Our 1D, fourth generation, and 1E, fifth generation, self-developed process technology and EUV development are on track.

We continue to provide a broad product portfolio, including DDR5, low-power DDR5, DDR4, low-power DDR4, and DDR3, low-power DDR3. Also, we will support customized AI wafer-to-wafer bonding product for AI infrastructure solutions.

That conclude my report to you. Now we may move to question and answer.

Thank you.

Thanks for taking my question, and again, very strong results. I think I have the two questions.

First question is that very, very strong DDR4 demand still, right? I'm wondering if cosmos customers, like, for example, CSP or Nvidia, does it reach out for a more supply kind of security?

Also, I think it was a news report, right, talking about Apple may also consider to use company's low-power DDR products. Just want to get a sense about those kind of high-profile customers engagement with Nanya.

That's my first question. Thank you.

At this point, the DDR4 demand still quite strong. It could take more than 10% of the overall market demand.

In comparison, Nanya overall market size is in low single-digit. Okay.

DDR4 demand is quite strong. However, Nanya is flexible on adjusting our DDR4, low-power DDR4 with DDR5 and other product we have.

Okay. That's including our AI UltraWide IO.

Okay. Your comment about Apple potentially using competitors' DRAM, for Apple, we don't supply their major DRAM.

Okay. However, we may have supplied their modem DRAM, okay, and those is a special DRAM that we don't think that there are other supplier providing those supply to Apple.

Okay.

All right. Thanks, Dr.

Lee, for a very, very clear answers. My second question is actually about, kind of wafer-on-wafer kind of a technology, because when I see Qualcomm, right, they kind of use it for their high-bandwidth compute, right, and I think that's a one way to resolve the bandwidth issue or HBM supply issue for some AI accelerators.

I'm not sure if there is a kind of a potential or existing partnership with Qualcomm, or not just Qualcomm, right, but any kind of a progress for the kind of a wafer-on-wafer for the AI applications?

Charlie, I cannot comment on specific customer solution, okay. Mostly this is the customer's confidential and market trade secret, okay.

We have more than one customer working with us, okay, on the AI-related wafer-to-wafer bonding 3D IC projects, okay. The application would likely to be spend from, say, cloud computation for AI, as well as all the way to edge computation, okay, and physical AI as well.

Okay. Far so good.

Far, it's so good, yeah. We already have some revenue, okay, and not big, but the is a important business for Nanya to work on, okay.

Currently, even though the contribution is not big, we'll continue to engage with customer and support customer.

Got you. Do you foresee the end of the DDR4 price upcycle, or is there any kind of a telling or comparison for us to gauge when and how this DDR4 price hike will end?

DDR4, as I say, the demand still still quite strong, okay. Everything will would depend on the mark market ecosystem.

There are area that high end, they will proceed continue with higher ASP, okay, and there are area also they have to balance in between down-specification, extension lifecycle, as well as some area even using recycle parts, et cetera, okay. Overall speaking, DDR4 is still market is still reasonably healthy.

Again, DDR4 is not Nanya's only supply. We are flexible to adjusting DDR4 to DDR5 and other product as well.

Great. That's all my questions.

Thank you.

Thank you, Dr. Lee, for taking my question, and also congratulations on great result.

My first question is, within your 60% or above ASP increase in last quarter, can you quantify how much of the impact was from the renewed long-term agreements? Is this mainly bi-annual repricing mechanism that you started reflecting, or did it include also annual or other quarterly contracts?

I'm just trying to see if we can also expect another round of this type of renewed long-term agreement-related pricing hike into the next quarter in Q4. Thank you.

That's the mix of all those you mentioned, including renewal of 1-year or 6-month LTA that was mostly negotiated in either first quarter or second quarter last year, or third quarter last year, okay. During those timeframe, the ASP was relatively low, and Nanya has make our commitment to deliver.

We have fulfilled our obligation on all the delivery at the price we have agreed upon last year, okay. That's the situation for Q3, where most of this long-term LTA has expired.

On top of that, they are Q2 and Q3 quarterly update as well. I would say contribution is mostly due to those long-term agreement for those 60%.

Just a small follow-up is, you said large amount is from these renewed the pricing contract, then can we say definitely, more than half, maybe even 2/3 or maybe 70% of what is...

I cannot quantify precisely, okay? I cannot quantify precisely.

I would say more than half, yes. Yeah.

More than half. Thank you.

Just a one second question is about your summary, you has mentioned AI infrastructure and servers now over 20% of revenue. Does it include mainly DDR5, DDR4 for server customers, and also for enterprise SSD solution, does it have any low-power DRAM solution that you're selling to server as well?

It's a mix of everything, okay, including DDR4, DDR5, and low-power DDR as well, yeah. That's including chips that used for communication, for storage, and for the system management, as well as enterprise SSD, and also RDIMM as well.

Understood. One small follow-up here is, because these customers are very longer-term order duration customer in our understanding, if you have any expectation about how big this segment could grow from currently over 20% to it could even grow to 25% or 30% or even higher number within next 12 months, if you could give us your high-level thoughts about next direction, that would be really great.

Thank you.

We plan to increase that percentage 2027, '28, and '29, okay, for our goal. Our target is by '29 or 2030, we can grow to 40% to 50%.

Oh, 40% to 50%.

By '29 or 2030, yeah.

Can I interpret... For all these...

Sorry to interrupt. Can I interpret your message as the firm's strong commitment to allocate increase in capacity, 30k by end of '28, and also 50k by end of 2030, majority will be used for is the server application?

AI infrastructure, including, okay. That's including all kind of different AI application that I just described to you.

Great. Thank you very much, Dr.

Lee. Thank you.

Thank you.

Thank you very much, Dr. Lee.

Sorry for you may hear some background noise, sorry for that, but congrats on your great results. Maybe very quickly for your long-term view, your ASP, DRAM price has been rising very notably up to Q3, but would you share your view, maybe December quarter still some room to see the double-digit increase?

Then maybe consensus indicates very stable or marginal decent increase for next year. Would you share your view?

Thank you.

Okay. For December quarter, we expect that the market shortage will continue, and likely ASP may again improve gradually over December quarter.

As of next year, the situation will depends on continued shortage in each of different sectors, okay. It's important to continue to observe the continuation of this sector application.

For instance, for HBM, the negotiation for next year is about to be completed, okay, and depends on HBM the ASP situation. That would likely also change the capacity arrangement, distribution around different capacity.

Knowing that the new capacity introduction and the new construction, everything, will take longer lead time, okay, next year is kind of tough situation for the supplier to improve substantially in their output, okay. Likely, next year will remain tight as well.

Very clear, Dr. Lee.

Another quick check is DDR4 and below, including the low-power, do you believe still the demand, the volume, will grow, or you expect maybe a little bit declining trend of the peak growth? Thank you.

Likely, DDR4 and low-power DDR4 collectively, in terms of the demand size, will continue to get smaller, okay, likely will get smaller. However, the remaining demand is still quite strong compared to the supply, okay.

Taking an example, for last year, 2025, when the industry has started to end of life DDR4 and low-power DDR4, at that point, the DDR4 and low-power DDR4 in combine, in terms of demand-wise, still almost 30% of the market size, okay. That percentage, even though it's maybe coming down a little bit gradually, okay, still a substantial number.

As I described to you just now, Nanya's overall market share is low single-digit, okay. That's the reason why we stay in supporting DDR4 and low-power DDR4, okay.

On top of that, I keep on mentioning that Nanya do have flexibility of move to DDR5 as well as other product if the market were demand so, okay. If I look into DDR4 and low-power DDR4 market situation, just DDR4 alone today is probably almost over 12% to 14% of the demand today, okay.

That number, even though it may be slightly coming down at time, but still substantial big number for supplier like Nanya.

Did you say low-teen, 12% to 14% of the

total demand,

DDR4? Yes.

Yeah.

Low-teen. That includes the DDR3 or only DDR4?

Just DDR4.

I see. Oh, would you remind us your Q3 revenue mix, maybe DDR5 including low-power DDR5, and then DDR4, low-power DDR4, and then DDR3, low-power?

What is the revenue mix?

Our DDR5 remain around 10%, okay, and low-power DDR4 and DDR4 combined may contribute to likely around I would say more than 60% of our product, yeah.

You mean the more than 60%?

60%.

DDR4 and low-power, yeah. That does not include the DDR3, right?

That's not including DDR3.

Not included.

Not included.

Oh, okay. Oh, sorry.

One investor some investor asking any update for payout ratio, because it's very profitable, I know you have to spend a lot of money for the new fab construction, but any rough idea, the payout ratio, and yeah, I finish my questions. Thank you, sir.

Okay. That's a dividend distribution you talking about, okay.

Dividend.

Considering two factor. One is that we need to spend quite a bit of capex for future equipment procurement, okay.

That's capex could go up to NT$1.6 billion, just I'm sorry, NT$16 billion, right? NT$16 billion NT$16 billion US$16 billion, just I'm sorry, a little bit mix in the number.

US$16 billion just for one one fab, okay. We need to have a lot of capex upcoming.

Also second, very importantly, Nanya has a consistent dividend distribution policy before. We will continue to pay out our dividend for next year, okay.

We will balance between these two factor, capex requirement and dividend policy, and we'll balance in between. Sorry for the number mixing for you.

Yeah. Very clear, sir.

Did you say $16 billion for the one fab?

Yeah, $16 billion for one fab. That's including construction, including cleanroom, including equipment, as well as some EUV for the future.

That means 35k, as you pointed out,

16 billion? 16 $16 billion is for 50k.

Oh, 5-0.

Yeah.

50k.

Yeah.

Thank you so much, sir. Appreciate.

Thank you.

You're welcome, Simon.

Thank you.

Okay. They come up with the webcast question from Capital Securities, Lien Chen, okay.

Here's three questions. First question: What is the capex financing plan, production timeline, and potential customer demand for new fab in Pingtung?

Okay. Let me answer one by one first, okay.

The question is a little long. Let me answer one by one first, okay.

Our capex planning for Pingtung fab is not finalized yet, but we need to go through the first phase of construction, then the equipment move-in. Likely, this will be totally up to $10 billion capex in the long run, okay.

Production timeline, we target equipment move-in by second half 2028. Of course, the planning is mainly for customer demand on AI application.

The next question is: How it's mainly for wafer-to-wafer production? How do NDC and FATC collaborate in backend production?

FATC mostly in the wafer assembly and test. This means that after wafer is finishing processing, finishing what we call wafer level testing, and then FATC was started with grinding, dicing, packaging, assembly, okay, and then testing, okay.

This is the more of the very backend process, it's not a wafer-level processing, okay. The wafer-level processing will be mostly process by Nanya Technology.

The question two is: Could AI lead to a long-term structural DRAM shortage, and what's your view on consumer demand? Okay.

Actually, I have answered some of the questions in my presentation, okay. AI has been leading to some structural change in DRAM industry, okay, and in the short term, of course, it's a in shortage.

In the near short term, though, the new capacity introduction, new equipment delivery is long lead time. Solution in the near short term is unlikely, especially in 2027 or early 2028, okay.

What is your view on consumer demand? A consumer side, the supply has been suppressed, and consumer side, as I described, the high-end consumer will continue, they will persist with high cost, okay, and the low-end consumer may likely to down-specification or extend lifecycle, or even use recycle parts.

Q3 from Lien Chen is that: Can double-digit contract price growth be sustained, with gross margin remaining at 80%? You're talking about for the future, okay.

It's potentially possible. The another question from Michael Shan, KGI Securities.

Michael's first question is: What is the current LTA coverage rate and proportion of contract with price floors and ceiling? Okay.

How to keep customer from backing out if market turns? Okay.

The question is, for Nanya, our current LTA is around 50% to 70% range, okay, and the term is ranging from, say, quarterly to all the way to 3 years contract, okay. Overall speaking, currently, 3-year contract probably 20% of our output, more than 20% of our output.

We plan to increasing this long-term contract according to customer demand. Customer is very proactive in terms of getting longer-term supply.

We are working with customers to extend our LTA to longer-term supply. There were discussion of the floor price, ceiling price for the upcoming new contract as well, okay, and, of course, there will be some liability from both sides, okay.

From the supplier side, we will commit to our capex, okay, we'll commit to our future capacity grow and equipment purchase. From a customer side, they will likely to commit to some degree, okay, on their secure of the long-term supply.

Michael's second question is: Can the potential supply-demand gap be quantified based on DRAM silicon area requirement for AI chip? This is actually excellent question, yeah.

I think it's a good idea, and we have been doing that. Based on the result from analysis that likely DRAM's supply shortage will persist for a long time, especially for AI application.

The key index that we have to pay attention to is the AI investment. Is it the return of those investment is sustainable?

Okay. That's key question and key index to pay attention to, okay.

Such that if you're looking to the key investment from all US CSP provider, as well as the AI large language developer, okay, including Anthropic, OpenAI, and many others, okay, and also another regional AI development in mainland China, all this thing combined, if it's the return of investment is making sense, then this shortage is likely to continue for some time. Next question is from Hua Nan Securities by Yi-Hua Lu, okay.

Yi-Hua has three questions. Question one: What percentage of NTC shipments are wafers, package, and modules?

Okay. We ship around 20% or so as a high quality what we call known good die, high quality wafer, that customer can take that wafer directly to dice and apply to their multichip package, okay.

That's around 20%. The rest of 80% is either device or module.

Which product are primarily required for private placement customer? Okay.

That's already described in previous investor conference. The product could include enterprise SSD, could include KB cache, or high-bandwidth flash, and also include communication, switch, and management chip as well.

Given recent share buyback by Kioxia and SK Hynix, does Nanya have any buyback plans? Okay.

The answer is, so far, we don't have any buyback plan. Question three is: How could GlobalWafers GlobalWafers is a company, is a wafer supplier.

Wafer price increase impact Nanya's cost and margin? Okay.

Overall speaking, we have multiple supplier for wafers, including our sister company, okay. Overall speaking, this wafer price increasing is not the only topic.

Overall speaking, all the supply side, including equipment, material, construction, labors, cost is also been increased, okay, not just wafer, okay. There will be some impact on the cost side, but so is ASP of DRAM as well, yeah.

The balance between that for the margin. Yuanta Securities from Peggy Wen.

Peggy has two questions. Question one: With AI-related revenue accounting for over 20%, what is Q3 revenue breakdown across other application?

Okay. The other application, including our low-power application for automotive, for industrial, as well as for general consumer application, all compute all together, and also with the PC as well, okay, but PC mostly using DDR5.

What is the growth outlook for AI-related business in late '26 and '27? As I described just now to other question that Nanya plan to increase our AI infrastructure business from currently more than 20% to more than 40% or even 50%, year by year gradually, by 2029 or 2020 2030.

With the demand shifting from DDR4 to DDR5, what's the DDR4 outlook, supply-demand dynamic and pricing for 2027? I think I have commented this question just now over Charlie's question, as well as Simon's question as well, okay.

As I said, DDR4 shipping to DDR5 is gradually, but so far DDR4 demand is still quite big, and supply still quite in shortage, okay, for certain sector. Of course, PC sector, they don't use DDR4 as much, okay.

It's not representative for this common of DDR4. The next question from Yuanta Securities by Yi-Chen Su.

Two questions from him. Question one: What is the company's current order fulfillment rate?

Okay. For the long-term order that we committed to, our fulfillment rate, of course, we will maintain 100%, okay.

That's a commitment, that's the signed contract, okay. Of course, there are many other inquiry for wafer for DRAM supply.

The fulfillment there usually is quite low, okay. Question two is: What is the company's bit shipment growth guidance for 2027?

That's interesting question, and actually, that is a embarrassing question for Nanya. We don't have much big growth in the beginning of 2027, and plus, we have a very depleted inventory, likely our bit shipment will remain flat for a few quarter until second half of 2027.

Thank you. That ends of all question asked online.

Thank you. Goodbye.