Hello, everyone. Welcome to the first earning conference of this year. Before we officially start our, I mean, today's conference, I would like to say Happy New Year to everyone. Now we will have our IR, Ronnie, to report the financial numbers for Q4 and 2025. As usual, we have announced our financial numbers yesterday, and we also uploaded our financial reports, financial report yesterday. If you need more details, then you can actually find the financial report on our website. Our Q4 revenue reached NT dollars 161.6 billion, marking a record high for a single quarter. This represents 42% year-over-year growth and 8% sequential increase, mainly driven by strong data center demand.
This sequential increase was actually above a normal seasonality, but just because the in terms of the contributions from the data center business has been increasing, I mean, for the last couple of quarters. The seasonality is... The normal seasonality has become a little bit different, I mean, from it was, from it was before. In terms of the GP margin, Gross Profit in Q4 was NT dollars 55.9 billion, up 59% year-over-year, and 7% quarter-over-quarter, marking a new all-time high. Shipping margin in Q4 was 34.6% versus 30.8% a year ago, and 34.9% in Q3.
Year over year, R&D and SG&A expenses increased 70% and 24% respectively, leading all paths up 21%, with profit surging 147%. Sequentially, all paths rose 7%, reflecting 2% and 11% increases in R&D and SG&A expenses, respectively, with profit growing 6%. As a percentage of sales, R&D expenses declined to 8.1% this quarter, compared with 9.8% a year ago and 8.5% last quarter. SG&A expenses were 10.2% versus 11.6% a year ago and 9.9% in the previous quarter. Benefiting from improved economics of scale, OpEx ratio in Q4 dropped to 18.3% from 21.4% a year ago and 18.4% last quarter.
Operating profit was up 147% year-over-year and 6% sequentially, bringing our Q4 operating margin to 16.3%, compared with 9.4% a year ago and 16.5% in Q3. In terms of the segmentation performance, Infrastructure delivered the strongest growth, with revenues up 94% year-over-year and 12% quarter-over-quarter. Power Electronics also posted solid growth, while Automation recorded modest but improving momentum. In contrast, Mobility remained under pressure, with sales down 31% year-on-year and 15% quarter-over-quarter. From a profitability standpoint, Infrastructure delivered the strongest performance, with profit up 287% year-over-year and 25% quarter-over-quarter.
Power Electronics also posted solid growth, rising 93% year-over-year and 1% quarter-over-quarter. In contrast, Mobility and Automation saw year-over-year profit declines of 27% and 83%, respectively. Quarter-over-quarter, Mobility recorded a profit drop, while Automation won to a profit. Operating income was negative NTD 800 billion in Q4, compared with negative NTD 900 million a year ago and positive NTD 2.2 billion in Q3. The negative income in Q4 was mainly due to the write-offs after a careful review of the assets at the year-end. In Q4, we had NTD 25.6 billion profit before tax. Our Q4 EBITDA was NTD 32.7 billion, up 92% year-over-year and down 7% quarter-over-quarter.
Q4 tax expense was about NTD 6 billion. The effective tax rate was 23.4%. Net profit after tax was about NTD 17.3 billion. The Q4 EPS was 6.67. Now let's have a look at the full year cumulative results. The revenue was NTD 554.9 billion in 2025, up 32% year-over-year. Gross margin was up 39% year-over-year, with a GP margin of 34.3% versus 32.4% a year ago. On the year-over-year basis, R&D expenses, I mean, increased 70%, SG&A rose 21%, OpEx grew 19%.
Operating profit increased 76% year-over-year. By segment, Infrastructure delivered the strongest performance with scale, with sales up 82% year-over-year, and profit surging 413%. Power electronics also posted solid growth, with sales up 25% and profit up 37% year-over-year. Automation saw modest revenue growth, while profit declined 39%. Mobility remained under pressure, with sales down 16% year-over-year, and profit turned to a loss. In 2025, we had about NTD 3.9 billion , I mean, non-operating profit. In total, we had 87.9 billion NT dollars pre-tax income. EBITDA was NTD 117.9 billion . Tax expense was about 18 point...
sorry, NTD 19.9 billion , representing an effective tax rate of 22.7%. As a result, net profit after tax was 60.1 billion NT dollars. Translating this number into EPS was 23.14. Yesterday, I mean, we, during the board meeting, we actually proposed a cash dividend for this year, which was NTD 11.6 per share. Now we are open to the Q&A session.
I have two questions. For the first question, I would like to know, would you have any guidance on the OpEx increase of this year? Also, do you have any guidance for the CapEx for this year as well?
For the OpEx increase this year, I think the OpEx the operate the OpEx will continue to increase. First, first of all, because we continue to have a lot of innovations and R&Ds need to be invested. Because we have, I mean, many new, sorry, we have many new products. We continue to need to, I mean, invest a lot into the R&D and innovations of our products. Secondly, Okay, for the CapEx, sorry, for the expenses, because we continue to, we actually are expanding our service team in other regions, because we are, our solution businesses have been increasing, I mean, in within the company.
We need to have more sales force and more FAEs and so on and so forth, to serve our clients in different regions. That is also part of the reasons we actually expect an increase, I mean, of the OpEx this year. Speaking of the CapEx, I think the CapEx this year is probably going to be slightly higher than the CapEx of 2025.
The next question is related to the GP margin in Q4. Could you please give us more, I mean, colors on your GP margin in Q4?
Because the GP margin in Q4 was actually softer compared to the third quarter.
My answer to that question is, I think because our GP margin is very much subject to the overall mix, product mix of the companies. There are actually, because we have, I mean, very diversified and many different product lines. The GP margin is always going to be a little bit, I mean, lumpy, but I think our GP margin currently, as I always said, I think it's actually quite healthy. I mean, it's at a quite healthy level, but we just can't really guarantee that we will always continue to, I mean, to achieve, I mean, the new record high of GP margin. I think that is actually our expectation, a healthy expectation for the GP margin.
Yesterday, according to the, I mean, financial report, I actually noticed that you actually mentioned, I mean, you actually mentioned asset, I mean, the write-offs of some of your assets on your financial reports. Can I say, I mean, is this kind of write-offs is actually one-off, or we will continue to see this kind of, I mean, write-offs going forward?
I think that is actually kind of a regular practice, I mean, at a year-end of every year. I can't really say that we will, I mean, actually expect to see such, I mean, write-offs every year, but we always, I mean, take this, I mean, cautious and conservative, I mean, approach to review our assets at year-end. That is basically, I mean, like that.
Next question: What percentage of total revenue does liquid cooling represent currently? What level could it reach in 2026?
In 2025, I think liquid cooling related revenue accounts for approximately 9% of our total revenue, with the majority coming from system-level solutions. We are not in a position to provide guidance for 2026, particularly in terms of mix, given the diversified nature of our business and multiple variables involved. That said, with the continued expansion of AI data centers, I think our AI-related businesses, I mean, including the liquid cooling business and our power-related businesses, or the liquid cooling market is actually experiencing strong growth momentum. We see significant opportunities ahead.
In, I think in terms of the AI investment, actually, the main demand driver for the AI deployment is actually still mainly coming from the major hyperscalers, especially in the US. If you look at the numbers and the investment, I mean, the CapEx numbers they have announced. For this year, supposedly, their investment, the absolute number of their investments are actually not less than the previous year. Of course, I mean, the CapEx, I mean, investment is one thing, but there are always, I mean, some other bottlenecks or uncertainties. For example, the lack of, I mean, labors and the lack of the materials. We can't be totally, I mean, we can't be totally sure about the pace of the deployment of their, I mean, investments.
Still, I mean, given that, I mean, given the commitment, I mean, from those, I mean, major CSPs, the commitment into the AI investment, I mean, from those major CSPs, we remain actually cautious, optimistic about the demand, I mean, for this year. I mean, although there might be some, I mean, uncertainties and or maybe just some ups and downs, I mean, in this long-term AI cycle. For the long run, we do believe that it is just a very early stage of this, I mean, the new era of AI, so we remain pretty, I mean, optimistic for the long-term opportunity.
The next question is related to the, to the potential bottleneck in terms of the power outage in the U.S.
I know that, I mean, actually, Delta has always been very ambitious in terms of the, your opportunities in the energy infrastructure business. Can you tell us more about, your plans or your opportunities, I mean, in this space?
I think that is actually a universal, I mean, issue, I mean, in many different countries and for many different types of our customers.... We do have this ambition to address, I mean, this issue, but, and, actually the architecture change, I mean, in the, in the, I mean, the power architecture change in the data center is always, is also somewhat related to, I mean, to this power efficiency issue. We will continue to work on this, and let's see that what we can achieve.
In terms of the, in terms of the outlook for this year, can you please give us some updates or just more colors for this, I mean, for main segments? I think as we, as the, as our Chairman just mentioned, this year, I think the main growth driver will continue to be the AI data centers. We actually expect to see a pretty healthy or solid, I mean, growth momentum for our power electronics and our infrastructure business.
Even though and then speaking of the Mobility, I mean, business, even though I think the market has been actually experiencing a pretty struggling and pretty challenging period, but still, in the long run, we still believe, this, I mean, the EV, I mean, is actually one of the key to address the carbon emission issues. We will continue to stay in this market. But just in terms of the, I mean, the clientele, I think, we actually, as everybody knows, as everybody knows, that in terms of the customers, I mean, we actually had or have, still have, like, much more customers.
We, I mean, there are the European OEMs or the American OEMs, but as everybody knows, that the Chinese OEMs, they actually have been really, really dominant in the global market. We will continue to try to, I mean, approach the Chinese OEMs, hopefully our exposure to the Chinese OEM customers can continue to, I mean, increase, I mean, going forward. Speaking of the automation business, as we just, I mean, reported last year, we actually had very modest growth for automation segment. Still, we can't really say that we have already seen very clear signs of, I mean, recovery in the China market.
The only thing I can say is the market is probably, I mean, has already been bottoming out a little bit, but still not a very strong signs of, I mean, recovery yet. Hopefully, of course, we do believe, we do hope to see the growth, I mean, for this year can accelerate. That is something actually out of our control. I mean. The first question is actually related to the capacity. In terms of the capacity issue, I think we still have a actually pretty tight capacity in order to fulfill the customer, I mean, customer's demand. Last year, actually, at the year end of last year, we had 3 new factories, I mean, in Thailand.
3, I mean, 3 of our new factories in Thailand have been going online at the year end of last year. We continue to have, I mean, some new capacity plans for other regions. Not just, I mean, not just Thailand. Actually, this year, we even, I mean, visited Mexico, but still, we are still in the process of, I mean, evaluating whether the, I mean, the environment, the overall environment, I mean, in Mexico is suitable for us to, I mean, to further expand our capacity in, in the America, I mean, regions.
The next question is related to the, I mean, what is the timeline and shipment scale for PowerRacks? Could this become the next growth driver?
Actually, assuming no major disruptions, we expect to see initial shipments of PowerRacks this year. However, the actual contribution will depend on customer demand and broader supply chain coordination. Visibility remains limited. Conceptually, PowerRacks integrate additional components such as relays, breakers, cabling, PDU, ADS, BBU, PCS, and even liquid cooling systems, which meaningfully expand the addressable revenue opportunity for us. I want to circle back a little bit to the to your opportunities, your opportunities in the energy, I mean, infrastructure market. I know that you have this, I mean, hydrogen energy business, which aims to, I mean, address this power insufficiency, I mean, issue.
Do you have any other, like plans, or do you have any other new product lines which might, I mean, actually help with this, I mean, power insufficiency issue in the background of, I mean, increasing power usage due to the AI deployment?
I think we just, I mean, actually briefly mentioned that the, I mean, the new architecture, I mean, in the data center, which is actually, I mean, refers to this, I mean, 800 VDC, is actually part of the solutions, I mean, to solve this, I mean, energy issue.
In terms of this new power architecture, actually, the AI data center, they are all using a lot of, I mean, energy and electricity. Something that we can actually do to help with this situation and or issue is that we help our customers to try to save or save, I mean, as much as, I mean, energy as possible by saving or reducing the energy loss, I mean, during this, I mean, energy conversion process. That is actually one way that we help with this, I mean, energy or electricity insufficiency, I mean, issue. The other one is actually the one you just mentioned, which is related to our hydrogen energy business.
As we previously, I mean, we explained in our earnings conference, this, I mean, Hydrogen energy, actually, compared to the traditional, discrete power generation, I mean, solutions, actually, in terms of the power conversion efficiency, is, maybe higher, much higher than the traditional solution. It can, I mean, up to maybe 65% conversion, I mean, conversion efficiency for this Hydrogen energy. In terms of the contribution or the shipment, I think we are actually expecting to have some, I mean, very initial shipment, at the end of this year.
If we want to see more slightly, maybe more meaningful contribution from this hydrogen energy business, we may need to wait until maybe next year or even maybe for the longer run. You just mentioned that you already have some clients, they are testing your products. They are testing your fuel cell products currently.
What kind of customers they are? Are they the like the CSPs or other types of customers? For the hydrogen energy, the fuel cells, the clients currently, they are the utility companies instead of the CSPs.
I think there is still, I mean, for the CSP clients, penetrating into the CSP clients, I think, still, I mean, it's still going to take some time to penetrate into the CSP clients. The next question is, can you please give us some maybe breakdown or split in terms of your power products, I mean, for maybe GPU servers or maybe, like, ASIC servers?
Actually, we don't really have the, we don't really provide the details, I mean, for, I mean, for this kind of split. Actually, we provide a lot of different products, I mean, including our AC products, AC powers, and DC converters for, I mean, for different type of data center clients.
It's actually not that, I mean, simple to really separate different type of our products for different type of the platforms that work for different different customers. Speaking of our, I mean, DC-DC converter business, we usually do not comment on specific specific, I mean, customer progress. However, I mean, given the expected meaningful, I mean, increase in the market demand, we actually remain cautiously optimistic about our overall DC-DC business this year. I think, I mean, generally speaking, the whole industry or the whole, I mean, architecture is still, I mean, in a stage of, I mean. It's still evolving.
just like, I mean, many Silicon Valley guys are talking about, maybe we can actually set up the data centers in the space. I think there might be some, I mean, advantages of, I mean, if you, if you try to put the data centers or set up the data centers, I mean, in the space. there is still some other, I mean, some other problems that you have to solve. for example, in terms of the cooling, how do you plan to cool down the equipment? I mean, if you put those, I mean, equipment or, I mean, data center equipment in the space.
There are many different, I mean, conceptual ideas, I mean, are being raised in this stage, especially, I mean, during the innovation of technology. The only thing I can say is we actually, we actually must, I mean, continue to innovate in line with market trends and customer needs in order to sustain our competitiveness.
My next question is, actually, can you please give us maybe some rough breakdown or split between your AI servers and your traditional servers? We think that actually it's actually not easy to, I mean, to separate the AI server revenues and AI server power revenues and traditional, I mean, server power revenues.
Can you give us maybe some guidance for the Q1 seasonality? Because traditionally, we all know about the first quarter historically should be the lowest season for the whole year. Just because the product mix, the business mix has been changed over the years, over the quarters. Can you please maybe just give us some ideas or some clues about the first quarter seasonality? Second question is related to the high voltage DC. As I recall, that you actually not just have, you don't just have the 800 volt DC, you also have the ±400 volt DC.
Can you please tell us, I mean, the adoption rate of, I mean, by the CSP clients for these 2 different types of, I mean, high voltage DC power racks?
I think as we, as Chairman said, different customers, they actually have different preference for these 2 types of, I mean, HVDC power racks. We do provide, I mean, both, I mean, solutions, so it's not a problem, I mean, for us. Speaking of the seasonality, because, as everybody knows that in the first quarter, because, I mean, there are actually fewer working days, I mean, in the first quarter because of the Chinese New Year holidays.
Theoretically, the second quarter is very likely to going to be, I mean, better than the first quarter, given that reason. Speaking of the 800 VDC and ±400 VDC, because, in terms of the deployment, in terms of the pace of deployment, because they are all based on the, I mean, the needs, I mean, of our customers. It's actually hard for us to forecast, I mean, the pace, I mean, of their pulling, our merchandise.
Still, I think of, generally speaking, in terms of the sales contributions, I mean, from this high voltage DC, I think, next year should actually be the, the main year in terms of we start to see some more meaningful, I mean, contribution from the power Rack business instead of this year. My next question is still, I mean, related to the capacity, I mean, issue we just, I mean, discussed at the beginning of the conference. Because you mentioned that you actually are considering to maybe expand your capacity in your, in Thailand or maybe the U.S., or maybe even Mexico. Are if you have, I mean, any new capacity or new factories in those regions, are they only, I mean, related to the, your power products?
Maybe you have plans for both power and liquid or the cooling products as well, I mean, for those new factories in different regions. I think we always plan our capacity based on our long-term, I mean, long-term plans instead of, I mean, any short-term needs. It always takes, like, a few years at least, at the very least, maybe 2-3 years to complete the whole construction of a new factory. We can only have some rough, I mean, estimate. Let's say that we may need, like, more new capacity, and then we want those semi capacity to serve maybe the clients in different regions.
I think there are always some dynamics and always some flexibility in terms of the capacity planning, especially for different type of products. We can actually, I mean, decide which factory for which product later. We need to just think of, I mean, or think about which region we might want to increase our capacity, I mean, at this stage. Now I think we can still have, like, one last question before we call it a day. The final question, I think, is actually related to whether Delta has any plan on the AI robots. Currently at this stage, I think we have actually still more industrial, I mean, robots.
Which we really have the mass production for this type of products. Speaking of the AI robot or service robot, we actually just, I mean, set up, establish a robot, I mean, research center. I think it was last year. This is actually for the long-term, I mean, innovation and the long-term plan for this, I mean, area. I think in order to achieve, I mean, or to, I mean, for those so-called service robots to be more reliable and to be more realistic for the actual use, I think it's going to take maybe a few years before that.
I think there are actually, I mean, some areas, I mean, or some areas or some problems have to be, I mean, have to be tackled before we see that really happen. Including the sensor or the sensing technology, including the communication capability, or, and also including the communication between the edge and cloud. Then also you need to, I mean, consider the safety issues if you want to really deploy the service robot in the in the live, I mean, environment. It's not like the factories, I mean, environment. It's actually much more simple and it's much more simple. I do believe it's actually going to take, I mean, some time to achieve that.
I think, as I said, I think, what we are, I mean, what we have been seeing now, I mean, they are still mainly more for a demonstration purpose instead of they, can be actually used in the, I mean, in the real, I mean, environment. That is my perspective. Okay, thank you for joining our, I mean, earnings conference today. I will see you on the next quarter. Thank you. Thank you.