BizLink Holding Inc. ($3665)

Good afternoon, and good morning, everyone. Welcome to BizLink First Quarter '26 Earnings Call hosted by UBS. I'm Ally Chen, covering the industrial sector in Taiwan. It's our pleasure to have BizLink management joining the call today. Now let me hand over the call to Mike Wang, Senior IR Manager, to introduce the management team and host the earnings call. Mike, please?

Thank you, Ally. Good afternoon, everyone, and welcome to BizLink's First Quarter 2026 Earnings Conference Call. My name is Mike Wang, Senior Investor Relations Manager. Joining me today are Roger Liang, our Chairman; Felix Teng, our CEO; Florian Hettich, our COO; and Charles Tsai, our CFO. Our earnings results were released earlier today and are available on our website where you can download the latest earnings materials and access the call through MOPS. Today's call will begin with Charles, who will review our financial highlights, followed by Florian for operational highlights. Felix will then conclude with corporate updates before we move to the Q&A session. You may submit your questions at any time through the public or private chat function, and we will address as many as time permits. Before we begin, please note that today's discussion may contain forward-looking statements based on our current expectations and are subject to risks and uncertainties. Actual results may differ materially. Please refer to the safe harbor notice in our earnings materials for further details. This call is being recorded and will be available on our Investor Relations website within 24 hours. With that, I would now like to turn the call over to Charles.

Thank you, Mike. This quarter reflected a combination of product mix transition effect, early-stage, infrastructure program ramp and varying recovery conditions across portion of the portfolio. Within HPC, next-generation platform continued ramping while portion of earlier generation product normalized, creating temporary mix utilization and operating leverage effect during the transition period. In addition, newer platform typically begin with lower initial volume and higher upfront engineering and operational costs before scaling efficiency improve over time. Outside of HPC, industrial and automotive conditions continue improving gradually, supported by selected customer activity and pilot program engagement, while electrical appliances remain competitive across selected product categories due to industry overcapacity and customer competition. At the same time, we continue expanding operational readiness across several manufacturing sites to support future customer development deployment requirement and increasing infrastructure complexity. While this effort may temporarily impact near-term efficiency as operational move up the learning curve, they're intended to support long-term capability expansion and execution flexibility across the portfolio. More broadly, this quarter also reflects an ongoing evolution across the AI infrastructure ecosystem as deployment cycle become more synchronized and rack level complexity continue to rise. Customers are placing greater emphasis on execution capability, interoperability and deployment reliability. Importantly, platform transition are also becoming less sequential and more overlapping. Earlier infrastructure upgrade cycle often follow relatively clean transition path. Today, deployments increasingly coexist across multiple infrastructure generations architecture simultaneously. As a result, deployment timing is no longer determined only by silicon availability, but also by rack level validation, power infrastructure readiness, thermal optimization and interoperability requirement. At the same time, supply chain are operating with less buffer inventory and expedited order than in prior period. While this creates greater visibility into deployment timing differences during transitional phases, we believe it also reflects a healthier and more operationally disciplined infrastructure environment over the long term. We believe this trend favor company capable of supporting increasingly complex infrastructure environment across multiple deployment layer and evolving architectures. We believe BizLink is positioned well as this infrastructure requirement continue to scale. So how we plan the business. From a portfolio perspective, our business remains balanced between long cycle infrastructure growth area and more stable cash generating businesses. Our HPC and capital equipment business continue to represent the primary long-term growth driver of the company, supported by increasing system complexity, deeper customer engagement and expanding infrastructure requirements. At the same time, our industrial automotive and electrical appliances business continue to provide portfolio diversification across broader end market and customer bases. This balanced portfolio structure support both long-term growth participation and operational resilience across varying industry cycles. We are managing this business conservatively while continuing to focus on engineering differentiation and customer diversification. Overall, our portfolio continues to provide both durability and exposure to structural infrastructure investment cycles. Several important industrial shifts are becoming visible. First, platform transition are becoming less sequential and more overlapping. Earlier, AI infrastructure cycle often follow relatively clean upgrade path. Today, deployment coexist across multiple infrastructure generation architecture. This matters because deployment timing is no longer determined only by silicon availability. It is influenced by rack level validation, power infrastructure readiness, server optimization and interoperability requirements. Second, AI infrastructure architecture themselves are becoming more diverse across interconnect technology. The industry continues to evaluate multiple pathways simultaneously, including both copper and optical approaches depending on deployment requirements, performance target and system architecture. At the same time, rack-level power architecture are also continuing to evolve as power density rises, including different approaches to power distribution and system level infrastructure design. We view this as positive for BizLink as architecture diversify, the importance of connectivity integration and interoperability increases. Customers require partner capable of supporting evolving infrastructure environment rather than depending on a single technology pathway. Third, we are seeing a continued increase in raw material costs, particularly in copper and PVC. This has led to ongoing pricing discussion across portion of the supply chain. We are selectively adjusting pricing where appropriate while continuing to support customer ramp requirements. Finally, we are continuing to invest into capacity and operational readiness across multiple sites, including Batam, Tainan, Penang, Johor and Vietnam. These investments are intended to support long-term structural growth and strengthen our ability to execute complex customer program globally. Operationally, we're continuing to move up the learning curve across several strategic important programs and manufacturing ramps. In HPC infrastructure, customer requirement continued shifting toward higher current density, tighter integration and more complex rack level deployment developments. This includes both power and data infrastructure. As system scale, engineering complexity is increasing faster than unit volume. This is an important distinction. The challenge is no longer simply shipping more units. The challenge involved validation depth, interoperability, thermal reliability, safety and deployment execution across systems. Within capital equipment, we continue to see healthy long-term customer engagement despite a more measured near-term investment cadence in certain areas of semiconductor ecosystem. Demand visibility remained relatively healthy, particularly for more complex box build and subsystem integration projects. Across our global operations, we continue strengthening execution discipline, manufacturing readiness and engineering collaboration. While this investment may create temporary inefficiencies during ramp phases, they position us well for future scaling opportunities. Looking ahead, we believe the broader AI infrastructure opportunity remains substantial, but infrastructure constrained rather than silicon constrained. As rack level complexity rises, deployment execution become a key differentiator. This benefit company with engineering depth, global operational capability, interoperability, experiences, validation capability and system-level infrastructure participation. We believe this link is positioned well in this area. Importantly, we continue approaching the current environment with financial discipline. We are preparing our capability and capacity for long-term infrastructure scaling rather than reacting to short-term deployment fluctuation. While quarterly timing variability may continue, we remain focused on strengthening our long-term position in closed AI infrastructure, semiconductor system automation, health care and next-generation mobility applications. Florian will now provide update on our latest quarterly operational takeaways.

Thank you, Charles. Across our operations, the biggest change we are seeing is that infrastructure complexity is now increasing faster than deployment volumes. This is true across HPC infrastructure, semiconductor systems and within automation and advanced industrial applications as well. The GPU may attract the headlines, but the infrastructure surrounding it determines how quickly and reliably systems can scale. That includes power delivery, thermal management, interconnect routing, serviceability and validation reliability. These requirements are becoming more demanding quarter after quarter. We view BizLink as an infrastructure execution and integration company rather than purely a component supplier. Across multiple businesses, customers are engaging with us earlier in their development cycles. And in many cases, we are participating during qualification, validation, routing optimization, manufacturability reviews and deployment preparation phases. This is especially visible in HPC infrastructure and semiconductor systems, where customers' requirements now extend beyond stand-alone components towards system-level execution capabilities. Our global footprint also continues to become more strategically important. Customers prioritize geographic flexibility, manufacturing redundancy and operational resilience as part of their supplier selection process. One of the clearest shifts is that AI infrastructure is becoming rack-centric. Historically, much of the industry's focus was on server level optimization. Today, power distribution, cooling and interconnect architectures are being designed at the rack level. As rack power continues to rise, the industry is evaluating multiple evolving architectures simultaneously, including centralized power distribution, large rack bus bars, liquid cooled power infrastructure and modular power delivery systems. At these power levels, the challenge is no longer simply delivering compute. The challenge is to deliver the infrastructure around the compute. In many cases, infrastructure readiness is now becoming just as important as compute availability itself. This is an important industry transition. We are also seeing increasing architectural fragmentation across optics and copper interconnect technologies. Multiple optical approaches continue to coexist and standards are evolving. Rather than betting on a single pathway, our focus remains on supporting the connectivity and integration layers across evolving architectures. This includes high-speed copper, optical interconnect infrastructure, structured connectivity, routing optimization and power delivery integration. Operationally, our teams continue focusing on execution readiness and capability expansion. Within HPC infrastructure, we continue scaling our participation across both power and data connectivity layers. These include continued customer engagement related to higher current power delivery systems and next-generation interconnect solutions. Within semiconductor systems, customer engagement remains active across subsystem integration, EDS and FDS-related opportunities and complex box build environments. We continue to see customers seeking greater outsourcing capability and global manufacturing support. In Penang, our primary challenge remains operational scaling and labor capability development rather than demand generation. Similar operational learning curves continue across other expansion sites as well. Within factory automation and robotics, activity levels continue gradually improving. We continue to see increasing customer interest in application requiring higher durability, higher data transmission capability and more complex motion environments. Importantly, many of these applications require engineering capabilities similar to those seen in HPC infrastructure, including signal integrity, EMI management and higher density connectivity environments. Looking forward, we believe the industry remains in the early stages of a much broader infrastructure scaling cycle. The key difference today is that infrastructure scaling is becoming cumulative rather than purely generational. Platform transitions overlap rather than replace one another. This extends deployment cycles while also increasing infrastructure complexity. We believe this supports longer duration opportunities across power infrastructure, interconnect infrastructure, subsystem integration and structured connectivity. At the same time, execution requirements are becoming more demanding. As architectures evolve, validation depth and interoperability become the gating factors for deployment. We believe our engineering culture, operational footprint and cross-domain experience position us well for this environment. Felix will now provide updates on our latest quarterly corporate takeaways.

All right. Thank you, Florian. We believe the AI infrastructure industry is entering a new phase. The first phase focused primarily on compute acceleration and deployment needs speed. The next phase focuses on infrastructure efficiency, scalability and interoperability. This is a natural evolution. As AI systems become larger and more distributed, infrastructure coordination becomes important. Expanding AI infrastructure resembles expanding a city's power grid and transportation network rather than simply adding more compute boxes. That transition creates both complexity and opportunity. BizLink role within this ecosystem continues to evolve. Historically, many viewed connectivity primarily as supporting components. Today, connectivity acts as a coordination layer across compute, power, cooling and deployment infrastructure. This is especially important because the industry is no longer scaling through a single architecture pathway. Multiple compute, optical and power architectures continue to coexist and evolve simultaneously. We believe this increases the value of interoperability. Our objective is not to depend on a single architecture winner. Our objective is to help customers scale reliably across evolving infrastructure environments. One of the most important industry shifts is that infrastructure deployment itself is becoming an engineering challenge. As rack power rises and architectures become more complex, customers require deployment reliability, operational resilience, faster validation cycles, tighter system integration, scalable manufacturing capabilities. At the same time, AI infrastructure deployment is becoming global. Customers are balancing performance, supply chain resilience, geopolitical flexibility, operational redundancy. This increases the importance of companies capable of supporting customers across regions and across multiple infrastructure layers. We are also seeing increasing convergence between AI infrastructure and other advanced application domains, including automation, robotics, autonomous systems, advanced industrial systems. These areas remain early, but many are beginning to require similar characteristics, high power density, low latency, high reliability, compact integration and continuous operational durability. Our focus remains execution-driven and long-term oriented. We continue strengthening our engineering capability, operational footprint and customer engagement depth across key growth areas. Within optics, we continue expanding our infrastructure and integration capabilities while remaining disciplined regarding where we participate within the ecosystem. Within power infrastructure, we continue preparing for higher density architecture and more complex deployment environments. Within automation and advanced industrial applications, we continue positioning ourselves around higher-value engineering-intensive solutions rather than purely volume-driven opportunities. Importantly, we continue approaching growth with financial discipline and operational pragmatism. As we have discussed previously, our objective is not simply to grow faster. Our objective is to build durable long-term capabilities that can scale across multiple infrastructure cycles and customer generation. Looking ahead, we believe AI infrastructure growth is becoming both cumulative and content-driven as deployment architectures become more complex, infrastructure content persistence rises, interoperability requirements increase, engineering depth become more important and execution capability become more valuable. We believe these trends align well with business' long-term direction. Importantly, we do not believe the industry will converge quickly toward a single architecture pathway. Multiple approaches are likely to coexist for years across updates, power distribution and deployment topologies. This creates complexity for the ecosystem, but also opportunities for companies capable of bridging multiple environments effectively. Our role is to help customers deploy complex infrastructure environments more reliably, efficiently and at greater scale. We remain focused on execution, customer intimacy, operational resilience, engineering capability and disciplined long-term growth. We believe this foundation positions our business well for the next phase of infrastructure scaling. Now let me turn the call over to Mike.

Thank you, Felix, Florian and Charles. This concludes our prepared statements section. Now let us begin the Q&A section. Please type in your questions and then we will answer as many of them as possible in the time remaining. I want to remind everyone that there will be no forward-looking quantitative comments. Now for the first question, questions from Masalalis, Morgan Stanley, QJ, of course, BofA Doris, this all has to do with our margins in the first quarter. Now for this one, generally, sort of the summary for the question. If margins were weaker than expected, what specifically drove the compression? How should we think about margin recovery for this one? I'd like to hand it over to Charles.

Okay. Thank you, Mike. Maybe let me take a step back and address this holistically because there are several moving parts, and they will lie back to one core dynamic. What we're seeing is primarily a transition between power generation with a more normalized and increasingly synchronized supply chain environment rather than a structural change in demand or probability. On the product side, our legacy solution are declining as expected, while next-generation programs are still in early stage of ramp. In that case, volume are not yet fully scaled. Utilization is still building and the mix has not yet fully shifted toward higher-value content. So that naturally create temporary pressure on margins and operating leverage during the transition period. One important difference versus prior upgrade cycle is that these transitions are now becoming increasingly compressed and overlapping. In earlier generations, products typically experienced a long ramp and monetization period before next major transition began. Today, next-generation platform are being introduced earlier while existing platforms are still scaling. As a result, multiple generation can coexist simultaneously, creating a more complex operating environment from both a mix and operational standpoint. During this overlap period, we might see a combination of legacy pricing normalization, early-stage ramp inefficiency for newer programs and elevated operational complexity as multiple product generations are supported in parallel. At the same time, the supply chain sales has become more disciplined with significantly less buffer inventory and fewer expedited compared to prior periods. Demand is now much more close aligned with actual deployment schedule, which means timing difference are more visible in the near term. Importantly, we do not view this as a structural demand issue. In many cases, it reflects customer accelerating infrastructure planning cycle and preparing earlier for next-generation deployment as system complexity and performance requirements continue to increase. So as new program continue to scale, we would expect the natural offset to come through over time, including improving utilization, better operating leverage and a more complete shift toward higher-value content. So from our perspective, the current margin profile is best understood as part of the transition and overlap phases within broader structural infrastructure growth and rather than a change in underlying long-term trajectory of the business. So yes, I hope that answers the question, and back to you, Mike.

Thank you, Charles. Let me see for the next couple of questions, still kind of talking about HPC, talking about -- how should investors think about business positioning within AI infrastructure as architectures continue evolving, thanks to Doris, Kenny, Derrick and of course, Helen. Since it's a bigger picture sort of question, I'd like to hand it over to Roger.

Okay. Our HPC strategy is centered around enabling the physical infrastructure required to scale increasingly dense AI deployment. Early phase of the industry focused primarily on compute acceleration sales. Today, the challenge is increasingly shifting towards how these systems are power interconnected, navigated and deployed reliably at scale. As AI infrastructure became more rack centric, the importance of power delivery, thermal coordination, high-speed connectivity and interoperability continues to increase. In many cases, infrastructure readiness is becoming just as important as compute availability itself. Our positioning reflect this evolution. We participate across 3 key infrastructure layers, copper interconnect, optical connectivity and power delivery, which allows us to support customers across multiple deployment architectures and evolving technology pathways. Importantly, our strategy is not dependent on any single architecture outcome. Multiple power, optical and interconnect approach are likely to coexist for years, depending on deployment requirement, performance targets and operational consideration. As a result, we believe the value creation increasingly shift towards company capable of supporting interoperability validation, integration and execution across increasingly complex infrastructure environment. And we believe it is positioned well in this area given our engineering depth, manufacturing footprint and long-standing customer engagement across multiple infrastructure layer. Thank you.

Thank you for that, Roger. Again, that's a big picture view. I think we get asked pretty much every quarter. Now for the next set of questions, again, a lot of people are still quite curious about the optics side business, especially given the deal that we did earlier in the quarter, XFS. How should we think about the optical business opportunity and our ramp trajectory, thanks to Ally, Amber, Helen, Nicky, North and Kevin. For this one, I'd like to hand it over to Felix.

All right. Yes, talking about XFS or Xin Fu Sheng, okay. So we consolidated the business starting early January. So the first quarter this year largely reflects a baseline run rate that is broadly consistent with 2025. So what we are seeing more recently, particularly into April, is a meaningful pickup in activities, which suggests that we are beginning to move beyond that initial baseline and into an early scaling phase. That said, we would still characterize this as an early stage of development rather than a fully established ramp, and we would prefer to see a few more quarters of consistency before drawing firm conclusions on the trajectory. Stepping back, the strategic rationale of XFS continues to strengthen as AI infrastructure scales. Bandwidth requirements, fiber density and interconnect complexity are all increasingly meaningful across next-generation systems. At the same time, the industry is increasingly recognizing that optical connectivity infrastructure itself is becoming an important scaling layer within AI deployments. While multiple architectures are still evolving across optics and interconnect technologies, we believe the long-term direction is clear. Higher performance systems will require significantly greater optical density, more sophisticated fiber management and increasingly scalable optical integration capabilities. As fiber density rises, routing, managing and integrating those fibers become operationally and mechanically more complex. This creates a growing need for high-density fiber assemblies, precision optical integration and scalable manufacturing capability. Importantly, scaling this ecosystem is not only about optical engines themselves. As optical density rises, the industry also needs supporting infrastructure across fiber supply, connector ecosystems, routing architectures, assembly capacity and qualification process to scale together. We believe this ecosystem securitization requirements may increasingly become an important consideration as next-generation AI infrastructure deployments continue expanding. XFS strengthens our capabilities in exactly this area, particularly in fiber assembly and optical integrations, which positions us to support the next phase of optical interconnect infrastructure development. So while near-term revenue contribution will still depend on the pace and timing of customer deployments, strategically, we view XFS as an important long-term capability expansion with our broader HPC infrastructure portfolio. More broadly, this also strengthens our overall HPC positioning, where we now participate across 3 key infrastructure layers, copper interconnect, optical connectivities and power deliveries, allowing us to support increasingly complex AI infrastructure environments as system architectures continue to evolve. Now back to you, Mike.

Thank you, Felix, and thank you to those who submitted a similar set of questions. Now we've been talking a lot about infrastructure complexity. So we do feel that this is something that we do need to stress. And so for the next set of questions, again, thank you for -- to Kenny, Derrick and Billy. Why are power delivery interoperability and validation becoming more important. So for this one, given it's talking about operations, I'd like to hand it over to Florian.

All right. Thanks, Mike. So as we already laid out, I think the biggest industry changes is that AI infrastructure is increasingly becoming more rack-centric rather than server-centric. And as the rack power and system density rises, the challenge is no longer simply deploying more compute. The challenge is more reliably operating increasingly complex infrastructure environments around the computing system. And at these power levels we see currently, the customers must simultaneously manage different things, power distribution, thermal coordination, thermal management, signal integration, redundancy, serviceability and also, as I said already, interoperability. And this all has to be managed across dense rack environments. And you can imagine a failure at infrastructure layer can affect a large portion of the deployed compute power simultaneously. And this, on the other side, elevates validation and also reliability requirements materially. And at the same time, multiple architectures continue to evolve simultaneously across different technologies like optics, copper interconnects or power distribution, and customers, therefore, in our view, increasingly prioritize interoperability and execution capability rather than dependence on a single technology pathway. As a result, the qualification cycles are becoming much deeper and the deployment readiness is becoming more operationally intensive over the process. So in many cases, infrastructure readiness is now becoming just as important as compute availability itself. And we believe these trends favor companies, especially with engineering depth and validation capabilities and also operational execution experience across multiple infrastructure layers. And we believe that we are well positioned in that respect. So I hope this answers this question. Back to you, Mike.

Thank you, Florian. And also thank you to those that presented those questions. Now again, this kind of ties back to our margins as well. So we would like to address this in the Q&A. But this also has to do with the whole evolution of the AI infrastructure cycle. And for this particular question, where are we currently in the AI infrastructure investment cycle? Thanks to Derrick, Billy, Helen and of course, Vicky. So for this one, I would like to hand it over to Felix.

All right. Yes, we believe the industry is transitioning from a phase focused primarily on accelerating compute deployment toward a phase increasingly centered around scalable infrastructure operations. So earlier deployment cycles were more sequential where one platform generation largely replaced the previous generation before the next major transition began. Today, platform cycles are becoming increasingly overlapping and accumulated. So customers are scaling existing deployments while simultaneously preparing for next-generation architectures. So this creates a more complex operating environment, but also extends the overall infrastructure investment cycle. Importantly, infrastructure scaling today involves much more than semiconductor availability alone. This deployment timing increasingly depends on rack levelization, power infrastructure readiness, thermal optimation, interoperability and operational deployment capabilities. As a result, we believe the industry remains in the early stage of much broader infrastructure scaling cycle and one that is becoming increasingly infrastructure intensive and operationally complex over time. Yes. So to sum up, I think we are, again, in a very early stage of the AI total development investment cycle. Back to you, Mike.

Thank you, Felix. Then of course, to those that asked about that particular topic. Now I do see there is a question posted in the chat. And this also has something to do with the operating leverage and long-term profitability, thanks to Eden. And so this one is, how should investors think about operating leverage given higher engineering intensity and optics investments, of course, in addition to Eden and the analysts that presented this particular question, Kenny, Terry and Derrick. I'd like to hand this one over to Charles.

Okay. Thank you, Mike. I think that we can look at this way. We continue to believe that the long-term operating leverage opportunity remain intact. Although the path may not be perfectly linear quarter-to-quarter due to ongoing platform transitions and earlier stage program ramp as we just mentioned. So as AI infrastructure scales, our value creation increasingly shift towards higher content and more engineering-intensive solution across power delivery, high-speed connectivity, subsystem integration and interoperability layers. But importantly, many of those newer programs leverage engineering platform, manufacturing processes and qualification framework that we already developed over multiple customer generations. So during early-stage ramps, we typically experience temporary inefficiency related to utilization, engineering support, operating readiness and qualification of activity. However, as volume scale and architecture mature, we will expect operating leverage to improve although greater platform reviews, better utilization and a more complete shift toward higher-value content. So structurally, we continue to believe that the business is moving toward higher operating efficiency over time, although near-term variability can occur during transitional period as one that we're currently navigating. So yes. And also one thing to look at is that if we look at the operating profit level in this quarter, actually, it's still -- I think it still show that we have strong OpEx discipline, although we're investing in those capability. So I hope that answered the question and also back to you, Mike.

Thank you, Charles. Thank you to Eden as well and for the analysts that posted those questions. For the next one, this is something that we do want to address given that this narrative was present for pretty much early this year. This has to do with copper versus optics. We do believe in a coexistence rather than a replacement. So this one should be the question sort of how should investors think about copper versus optical over the longer term? Is optics replacing copper. For this one, thanks again to Kenny, Ally, Kevin, Vicky and Kemper. I'd like to hand it over to Florian.

Thanks, Mike. So this is probably one of the most discussed and questions in these times. So as already said, we do not view this as a binary transition where one technology is just replaced by another. When we now see the AI infrastructure scaling, we see that different interconnect technologies are increasingly being optimized for different application and requirements in terms of distances, in terms of power envelopes, in terms of latency requirements and so on, also operational considerations. And this is also depending on the deployment architecture. So there's a coexistence of both technologies. For example, copper-based electrical interconnects, they are still highly efficient for shorter reach. They have high-density environments where power efficiency, latency, serviceability and also integration simplicity are critical requirements. Optical technologies become more important if it comes to bandwidth requirements or transmission distances when they get longer than optical technologies will become more popular. So what we see is not so much a substitution, but an increasing architectural diversification and also an interconnect complexity. That's what we already said in our remarks. So in practice, the AI systems are becoming more sort of a hybrid environment where different technologies coexist across different portions and different areas of the infrastructure stack. So importantly, as the optical density also increases, the challenge extends beyond also optical engines themselves, routing and fiber management, connector ecosystems, assembly scalability, but also qualification processes and operational serviceability, these are all now becoming increasingly important. And this is one reason why the optical infrastructure layer itself is strategically becoming more and more important as deployments are scaling. And our approach, as already pointed out, is, therefore, not centered around predicting one single architecture winner. Our focus remains more supporting evolving infrastructure environments across all technologies, copper interconnects, but also optical connectivity, structured routing and also power delivery integration. So to put it in a nutshell, we see at least for the foreseeable future that both technologies will still exist and coexist in the infrastructure. Thanks, and back to you, Mike.

Thank you, Florian, of course, and then also thank you to the analysts that posted those questions. Now I see actually quite a few questions around CapEx. So let me see how should we think about CapEx and manufacturing expansion over the next few years, given that a lot of the content that we've been sharing with everyone is talking about more of a longer-term capability building for BizLink. Thanks again to Kevin, Norris, Kenny. And I'd like to hand this over to Charles.

Okay. Thank you, Mike. Yes, actually, that is a very important question given our growth trajectory. So our overall approach remains disciplined and visibility driven rather than speculating. So the investments we're making today are primarily intended to strengthen operational readiness, engineering capability and development flexibility across multiple long-term infrastructure growth areas. So this includes expansion activity across sites such as Batam, Tainan, Penang, Johor and Vietnam, where we are preparing for increasing system complexity, higher infrastructure density and broader customer deployment required. And actually more importantly is that we manage our footprint as an integrated global operating system rather than isolated factories. Different sites support different roles across engineering, new product introduction, high mix execution and scale manufacturing. So in sum, we also continue pacing investment alongside customer program visibility and operational maturity rather than purely against short-term demand fluctuation. So as a result, we believe that our current expansion strategy remains financially disciplined while still supporting long-term infrastructure scaling opportunity. So yes, I hope that answers your question. So back to you, Mike.

Thank you, Charles, and of course, to the analysts that share those questions. For -- the next question, again, it seems like every quarter where we review the quarter, we're asked about our RA. So for this one, again, thanks to the analyst that put the question, how we should think about our long-term M&A strategy. Again, this is a bigger picture. So I'll hand it over to Roger to highlight this one.

Okay. Our M&A -- long-term M&A strategy remains the same, unchanged, okay? So our M&A philosophy remains capability-driven and strategically selective. Historically, we have focused on acquisition that strengthened the engineering impacts, customer engagement capability, manufacturing footprint or participation across structurally attractive infrastructure markets. And importantly, our objective is not simply adding revenue scale. We prioritize opportunities that enhance long-term position with an evolving infrastructure ecosystem, where complexity, interoperability and execution capabilities are becoming increasingly important. We also continue maintaining financial discipline regarding valuation, integration capability and the long-term strategic fit. So as the industry continues evolving across AI infrastructure, semiconductor system, automation and next-generation connectivity, we believe there may continue to be selective opportunity to strengthen our broader capability portfolio over time. I hope this answers your question. Back to you, Mike.

Thank you, Roger and then, of course, to those who continue to be very curious about our long-term M&A strategy. And there's some questions about order visibility. And so I would like to think this is going to be talking about our HPC business. This one is going to be how should we think about visibility compared to prior hardware cycles. For this one, I'd like to hand it over to Felix.

All right. Yes. We believe AI infrastructure differs from many prior hardware cycles because deployment planning increasingly occurs much earlier and involves significantly deeper architecture coordination across the ecosystem. So power architecture, thermal coordination, interoperability, rack level validation and deployment readiness often need to be defined well before system enter large-scale production. So as a result, supplier engagement increasingly begins earlier and extends across multiple deployment generations rather than being tied purely to short-term shipment cycle. While quarterly deployment timing can still vary depending on customer schedules and infrastructure readiness, the underlying infrastructure build-out itself appears increasingly multiyear and cumulative in nature. For this reason, we viewed the industry less as a traditional cyclical hardware market and more as a longer duration infrastructure scaling cycles with overlapping deployment phases and continuing efficiency upgrades over time, right? Back to you, Mike.

Thank you, Felix. Given the time, we'd like to take a look at the questions that were posted as well. And let me see the questions in the chat in the public. Let me see what -- so there's a question about, could you share the technical differences between power and busbar in traditional 48, 54-volt systems and 800 feet HVD systems and specifically, how much would the technical gap increase the ASP? Thank you. Now we -- there were some questions that were posted earlier about HVDC. And these will go back -- thanks to Doris, of course, Derrick, Kenny and Billy. And this one also addresses that particular question. Why is the industry moving toward HVDC? And what does this mean for BizLink? This is going to be the final question for the call. I'd like to hand it over to Florian for this last one.

All right. Thanks. So the transition towards a higher voltage power architecture is primarily driven by the physics and scaling efficiency rather than a technology or a certain technology preference of customers. So if the infrastructure is scaling up, direct power density is significantly increasing. And at these power levels, the delivery of large amounts of power in an efficient way using traditional low-voltage approaches becomes increasingly difficult due to current intensity, but also thermal load and efficiency limitations. So higher voltage architectures are significantly reducing the current levels. And this is on the turn improving efficiency and lowers the thermal stress of the system and therefore, enables more scalable power distribution across these dense rack deployments. And importantly, the different customers are still evaluating different approaches, so including various HVDC and rack level distribution architectures, different level of current. And we are therefore, expect multiple pathways to coexist for the future, at least for the near-time future, depending on the deployment philosophy and the system design of the customers. From our perspective, this transition increases the importance of higher current connectivity and power distribution integration, validation capability and also rack level infrastructure engineering. And we believe that these trends align very well with our long-term positioning in power infrastructure. And at the same time, just a last remark, I would also like to emphasize that -- for sure, HVDC is not replacing data connectivity, but these technologies address for sure, different infrastructure requirement. But in fact, improving power stability and thermal efficiency can enhance also operating conditions for high-speed electrical interconnects inside a very dense rack environment. So hopefully, this is addressing in the short terms -- short time, the question. Back to you, Mike.

Thank you, Florian, and also thank you to the investors that posted that question. So thank you, Roger, Felix, Florian and Charles. This concludes our Q&A section. A replay of the conference call today will be available on our IR website within 24 hours from now. If you have any further questions, please feel free to reach out to BizLink Investor Relations team. We thank you very much for joining today's call. You may now disconnect.