Thank you so much for joining the Investor Day of KIOXIA Holdings Corporation today. For the participants who are joining online, if you are using a nickname to log in, please change it to your name and the name of your affiliation. To change your login name, please log out once, go back to the registration page on this Zoom webinar, where you can change your name and the name of your affiliation, and log in again. Thank you for your kind understanding. The session will begin soon. Thank you. Thank you for waiting. We will start the Investor Day of KIOXIA Holdings Corporation. I'm today's facilitator, Fujikawa, from the Corporate Communication Division. With me here are President and Chief Executive Officer, Hiroo Ota; Executive Vice President and Executive Officer, Chief Financial Officer, Yoshihiko Kawamura; Managing Executive Officer, Chief Strategy Officer, Junichiro Yaguchi; Managing Executive Officer, Vice President of SSD Division, Masashi Yokotsuka; Executive Officer, Vice President of Memory Division, Atsushi Inoue. These are five speakers for you today. Let me inform you on the process of today's call. After outlining the disclaimer, the speakers will spend one hour to cover the agenda in front of you on the screen. The material is available on the Investor Relations site of our corporate website. After that, we are going to have a Q&A. The entire session will end at 6:00 P.M., which is subject to change depending on the proceedings. Please be advised that today's session will be recorded for on-demand viewing later. We would like to start with a disclaimer. Forward-looking statements are prepared based upon our expectations and projections in light of the information currently available to us, which involve various risks and uncertainties. Such risks and uncertainties may cause our actual results to be materially different from any future results expressed or implied by these forward-looking statements. We undertake no obligation to update any forward-looking statement included in the material. For more of the disclaimer, please refer to this slide on the screen. Now let us hand over to Mr. Ota. I am Hiroo Ota, President and CEO. Thank you for taking the time to join our Investor Day today. The rapid proliferation of generative AI is bringing a seismic shift to our industry. Today, I would like to outline our strategy for how we, as a provider of the flash memory that underpins the data infrastructure of the AI era, will meet market demands and achieve sustainable growth throughout the transformative period. I'll be explaining the following points. In last June's corporate strategy meeting, we shared our projection that flash memory would evolve from being a simple data repository into a critical component that dictates the performance of the entire AI systems. Over the past year, this projection has materialized as tangible demand. We interpret our current share price as a clear sign of the capital market's recognition and trust in our strategy to effectively capture this AI-driven demand. This growth is underpinned by rising data center investments since 2024. Most notably, as the use of generative AI started to shift from training to inference in the second half of 2025, the vital role of storage has been strongly reaffirmed. Consequently, demand for AI-focused data centers is expanding robustly, and the flash memory market is thriving. We anticipate this favorable environment will persist through 2027. Starting from the second half of 2025, the AI landscape is transitioning from a training phase focused on building large language models to an inference phase where these models are implemented in real-world applications. Furthermore, AI is evolving from a mere tool used as humans into autonomous agents, exemplified by agentic AI and physical AI that can autonomously reference data and achieve specific objectives. We anticipate this trend will trigger an explosion in the volume of inference processing. In this inference phase, calculation speed, data processing capacity, and total cost of ownership, TCO, including power consumption, become paramount. The key is how to execute these massive inference tasks both economically and efficiently. To that end, technologies such as retrieval-augmented generation, RAG, which leverages external knowledge, and KV caching, which reuses previous inference results, are advancing at a rapid pace. However, it will be difficult for high-bandwidth memory alone to meet the massive memory demands generated by these advancements. This is especially true as AI applications expand into the edge domain, where local data processing is a prerequisite. Consequently, the importance of low-power, high-speed storage has never been greater. We believe flash memory is the only storage medium capable of meeting these diverse market demands and economically scaling inference systems from a total cost of ownership perspective. Driven by these trends, discussions involving the entire ecosystem are intensifying around the realization of next-generation AI-native storage. At the heart of these discussions is our key partner, NVIDIA, who are proposing a vision of redefined storage, not as a mere storage device, but as an extended memory tier for GPUs. This vision is embodied in initiatives such as CMS, which utilizes SSDs as a memory tier to record previous calculation results for improved inference efficiency. Another example is NVIDIA's StorageNext, where SSDs are used as extended memory in RAG servers that store external knowledge to enhance response accuracy. Furthermore, we anticipate an accelerating increase in the volume of generated results stored on storage servers. What these developments signify is that alongside GPUs and HBMs, storage has become a core component that determines the performance of AI systems. As specialists in flash memory and SSDs, we will actively propose a wide range of technologies and solutions for these initiatives, leading this AI paradigm shift together with our partners. To capture the opportunities brought about by these structural changes in the market, we plan to make a strategic shift in our business portfolio. First and foremost, we will focus on our data center and enterprise business. Over the medium to long term, we aim to raise this segment's share of our overall revenue to over 60%. While maintaining and enhancing our competitiveness in standardized products, we will expand our offerings of high value-added products, exemplified by super high IOPS SSDs that directly help improve the efficiency of AI systems, thereby driving improved profitability. In our core smartphone and PC segments, we will maintain our revenue scale, positioning these as an unwavering foundation for our business. Building upon this solid base, we will expand our market development efforts targeting the edge AI segment where we foresee significant growth. I am pleased to report that for the second consecutive year, we have achieved record highs in both revenue and operating profit. We expect to maintain this high level of profitability in fiscal 2026, and our Q1 guidance projects an OP margin of 74%. The repayment of our interest-bearing debt is progressing smoothly, and we anticipate achieving a net cash position by the end of the first quarter. In light of this improved financial health, we are actively considering shareholder returns for fiscal 2027. The funds generated will be allocated towards strategic investments for the future alongside shareholder returns. First and foremost, as a provider of the flash memory and SSDs that form the backbone of AI infrastructure, we must resolutely ensure we continue to deliver optimal technologies and products to the market in a timely manner. We will advance the implementation of our technologies through partnerships with our customers. Simultaneously, as the procurement environment becomes increasingly uncertain, we will focus on building a resilient supply chain to ensure the reliable delivery of our products. Our second priority is the commercialization of our R&D achievements. At the research level, we have successfully demonstrated the operation of HCF memory as the next evolution beyond our vertically stacked BiCS architecture, paving the way for the future advancement of 3D flash memory. We will continue to evaluate fundamental cell characteristics to verify its feasibility. Furthermore, at IEDM 2024, we announced the development of OCTRAM technology utilizing vertical transistors made of oxide semiconductors. At last year's IEDM, we presented an evolution of this, a highly stackable oxide semiconductor channel transistor technology in a 3D architecture. We are committed to driving this development forward, exploring commercialization while assessing market viability and mass production readiness. As AI evolves, the required memory solutions must also evolve. We will explore strategic acquisitions to broaden our capabilities into domains that extend beyond flash memory alone. People are the foundation of this growth. We will train highly skilled professionals to support the technological expertise required to succeed in intense global competition and create an environment where diverse talent can thrive. We are accelerating digital transformation and the use of AI across our entire organization. Just as we enable AI infrastructure for our customers, we will lead by example by integrating AI deeply into our own operations to drive efficiency and innovation. Moving forward, we will bring our technological expertise to the forefront. By co-creating the next generation of AI capabilities alongside with our customers, we will expand our role as a vital provider of data infrastructure. We are not merely a component supplier. At the core of AI infrastructure, we are driving the next technological paradigm shift. We remain committed to meeting market demands and enhancing our corporate value. We look forward to sharing our future success with you as we continue to grow. Thank you very much for your time and attention today. Next, Mr. Yaguchi, please. I am Junichiro Yaguchi, Chief Strategy Officer. I'm going to cover the outlook for the flash memory market, our resource allocation, and our investment strategy. First, let me outline our bit growth outlook for the NAND flash memory market. CAGR through 2028 is projected to be slightly over 20%. Notably, data center demand is expected to grow significantly by 46%, with increasing AI server capacity driving market expansion. Within this segment, inference AI applications will serve as a primary growth catalyst and are likely to achieve an estimated CAGR of 86%. On the other hand, near-term storage demand for smartphones and PCs is undergoing a minor adjustment phase, with some forecasts projecting a slight decline this year. Although the demand environment varies by application, we are focusing sharply on the expanding AI and data center SSD markets. To meet this increased demand and achieve sustainable growth, we will accelerate our capital expenditure and R&D. Next, let's look at the supply and demand balance and the size of the NAND market. The former is expected to remain tight until the second half of calendar year 2027, driven by strong demand for AI data centers and constraints on NAND supply. Furthermore, NAND prices are on an upward trend. Several research firms estimate that the revenue of this market will be approximately four times larger in 2026 compared to 2025, with this trend projected to continue into 2027. In response to the expansion of the NAND market driven by AI and data center demand, we currently plan our annual capital expenditure to average approximately JPY 470 billion over the next three years. This represents an increase of roughly 60% compared to 2025. This figure includes not only investment in production equipment, but also in clean room interiors and infrastructure. It incorporates medium to long-term investments designed to secure upside flexibility by expanding our available space. Leveraging customer commitments, including LTAs, will establish a supply framework to support a stable demand. Furthermore, we'll continue to maintain higher capital efficiency than the industry average, executing disciplined investments that exceed our internal hurdle rate. We are targeting an average annual reduction in front-end cost per gigabyte in the 10% range. Through the migration of BiCS FLASH, we have factored in bit density improvements of 50% or more per each new generation. During transitions between product generations, we don't base our key performance indicators solely on the layer count. Rather, we plan to achieve competitive cost reductions by pursuing the optimal mix of 2D shrink and layer count. We will cover this point in further detail later by Mr. Inoue. Through efficient capital expenditure and by leveraging our world-leading economies of scale, we have maintained a lower cost per gigabyte than the industry average, and we intend to sustain this competitive advantage in the future. Research and development is a source of our competitive strength. Over the next three years, we plan to increase our R&D investment by 60% over 2025 levels, bringing it to approximately JPY 230 billion annually. Furthermore, we are progressing development tailored to specific memory hierarchy in order to deliver the distinct performance characteristics demanded by AI and data centers. We are allocating R&D investment to the development of 10th and 11th generation BiCS FLASH as next generation products following eighth generation BiCS FLASH. In SSDs, our key focus area, we are targeting investment sufficiently at areas that will help speed up launch of new storage architectures, including our super high IOPS SSD. Additionally, for new memory devices, we will accelerate development at our Frontier Technology R&D Institute to improve the characteristics of OCTRAM devices and facilitate the decision-making around commercialization of these. Although not listed here, we are planning to establish the tentatively named KIOXIA College at our Shin-Koyasu Technology Front campus, our R&D hub. This initiative, which we plan to launch by the end of 2027, is part of our commitment to boosting the development and cultivation of our research and engineers. Thank you for your attention. Next, the presentation will be provided by Mr. Yokotsuka. My name is Yokotsuka, Managing Executive Director and Vice President of the SSD Division. I will discuss our initiatives in the AI field in the context of our SSD business. We view the rapid advancement of generative AI and agentic AI as a critical growth opportunity, and we are boosting our development efforts across both our SSD product portfolio and our software solutions. Here's the agenda of my presentation. First, I will review our key achievements and performance from fiscal year 2025. Next, I will discuss the emerging challenges associated with AI inference systems as they become increasingly commercialized and widely adopted. I will highlight the new applications for SSDs that are being unlocked by these cutting-edge AI inference systems. Finally, I will introduce KIOXIA's comprehensive SSD portfolio, which is designed to fully address and support all of these new applications. Looking back, fiscal 2025 was a year where we achieved significant progress in our generative AI systems businesses. We successfully capitalized on this wave of growth, with our products being adopted by a broad range of customers, primarily in the data center, enterprise, and PC segments. The first of these was our high-capacity SSD, the LC9. Offering an industry-leading capacity of 245 TB, and it is designed to meet the rapidly expanding demand for AI data storage. It has already completed evaluation and validation with key customers and is now being rolled out to the market. The second is our open software solution, KIOXIA AiSAQ. As generative AI moves into practical application, the adoption of RAG is expanding rapidly. Previously, a major challenge was that vector databases, which are critical for response accuracy, were constrained by DRAM capacity and CPU performance limits. The software effectively utilizes SSDs and GPUs to significantly alleviate these constraints, enabling high-speed, cost-effective processing of vector data on a scale of billions. By integrating with Milvus, a leading open-source vector database, we enhance both the practicality and adoption of our technology. Additionally, we have leveraged the software library NVIDIA cuVS, which enables high-speed processing of vector databases through GPU acceleration. This collaboration allows us to construct databases with billions of vectors and achieve indexing speeds up to approximately 20 times faster than using only CPUs. Through these initiatives, we position KIOXIA AiSAQ as a crucial technology that significantly improves the scalability and practicality of generative AI infrastructures. We are advancing integration with various ecosystems. Driven by these achievements, fiscal 2025 was a year of steady progress in our efforts to keep pace with the evolution of AI infrastructure, allowing us to build a solid foundation that will support our future growth. Next, regarding the emerging challenges of growing AI inference systems. This section outlines the system architecture of conventional inference AI and the challenges that are emerging as it evolves. Traditional inference AI systems primarily perform a simple process, executing a single inference in response to a user's prompt and returning an answer. Therefore, a configuration combining inference servers equipped with GPUs, basic RAG, and standard storage was sufficient. The volume of data handled and the frequency of access remain relatively limited, and the demands on storage performance are not particularly high, allowing for the use of HDDs for cold data. However, with the rapid evolution of AI, this premise is fundamentally changing. In the case of agentic AI, the system autonomously repeats inferences multiple times, advancing processes while continuously interacting with external tools and other AI models. Consequently, the system load increases exponentially. A critical bottleneck in this new paradigm is the KV cache. As inference becomes continuous and multi-stage, the context expands rapidly, making it impossible to hold the entire KV cache within the GPU memory alone. This leads to increased cache evictions and reloads, resulting in memory throughput and storage performance becoming the primary bottlenecks, impeding overall inference efficiency. Furthermore, regarding RAG, the expansion of data scale and the surge in access frequency make it difficult to manage with capacity-constrained DRAM or slow storage like HDDs. As you can see, the evolution of AI is significantly elevating the performance requirements of the entire system, necessitating advancements not only in computational power, but also in storage capabilities. Next, regarding the new applications for SSDs unlocked by these cutting-edge AI inference systems. This section details how the latest AI server technologies are addressing these challenges. Currently, various solutions for inference AI are being proposed, notably by industry leaders like NVIDIA. The common denominator among all these approaches is the critically important role played by SSDs. The first point is addressing the lack of capacity of GPU memory, particularly HBM. In agentic AI, KV caches multiply rapidly and exceed the internal capacity of the GPU. To counter this, Context Memory Storage, CMS technology expands the cache onto high-speed SSDs to maintain inference efficiency. Furthermore, NVIDIA StorageNext advanced architecture and bypasses the CPU, enables GPUs to directly access SSDs, achieving an effective expansion of memory capacity. The second point is the evolution of RAG servers. While high-speed search is essential for handling large-scale up-to-date data, DRAM capacity has traditionally been a limiting factor. Today, the utilization of high-performance, high-capacity SSDs is successfully balancing scalability with cost efficiency. The third point is the transformation of storage servers. As data volumes surge alongside the widespread adoption of generative AI, high-capacity SSDs are becoming increasingly vital as the foundational infrastructure for delivering high deployment density and high throughput. Building on these developments, next-generation AI systems will achieve comprehensive system optimization by combining the externalization of KV cache via CMS framework, high-speed processing through StorageNext, and large-scale RAG that is independent of DRAM capacity utilizing KIOXIA AiSAQ. Ultimately, these new inference AI architectures are designed to resolve the structural challenges brought about by the evolution of AI with SSDs positioned firmly at their core. From here, I would like to introduce our product portfolio designed to cover these new applications for SSDs. As I have outlined, the advancement of agentic AI is driving an exponential increase in both data access volumes and processing density within inference AI systems. Consequently, data transfer performance between the GPU storage, namely bandwidth, has become crucial. To utilize storage not merely as an auxiliary space, but as an integral part of the effective memory hierarchy, storage solutions with high throughput and low latency are absolutely essential. To meet these exacting requirements, we offer the CM Series high-bandwidth SSDs equipped with TLC flash memory. The CM Series is fully compatible with CMS framework and is specifically architected to optimize KV cache applications. Incorporating our most advanced NAND flash memory allows it to achieve even greater capacities than previous generations, and to efficiently retain and utilize the massive amounts of data generated during the prolonged multi-stage inferences characteristic of agentic AI. Furthermore, the adoption of our in-house development SoC solution delivers superior power efficiency, directly contributing to the optimization of overall power consumption in data centers. In addition, its support for the latest PCIe generation and compliance with cutting-edge standards such as NVMe and OCP ensure seamless integration into advanced AI infrastructures. Crucially, the CM Series is also compatible with liquid cooling systems, which are increasingly being adopted in AI environments. This enables the drives to deliver stable, high performance, even in high-density deployments involving equipment with significant heat generation, such as AI servers. In summary, the CM Series is a core component that underpins the high-bandwidth data transfer essential for next-generation inference AI systems. On the other hand, the proliferation of agentic AI, characterized by a dramatic increase in the number of iterative inferences, generates extremely granular and high-frequent data access. In this domain, further improvements in latency and IO performance are critical. To address this, we are undertaking a performance innovation through a novel architecture that transcends conventional SSD technology. Meeting these exacting needs is the GP Series, a high-performance SSD powered by XL-FLASH. The GP Series is engineered specifically for direct integration with GPUs, fully supporting NVIDIA's StorageNext initiative, a framework designed to seamlessly extend GPU-accessible memory. By combining our proprietary XL-FLASH with a custom SoC capable of handling data access at an incredibly fine 512-byte granularity, we have achieved a major breakthrough. This combination delivers both exceptionally low latency and the high IOPS performance required by next-generation AI systems, which requires IOPS of 100 million or more. Crucially, this performance enables the SSD to function as an effective extension and complement to HBM. By mitigating memory capacity constraints, we are paving the way for an entirely new AI system architecture. Ultimately, the GP Series is the core product driving the low latency, high IOPS data access essential for the future of inference AI systems. Furthermore, the advancement of agentic AI presents another challenge that cannot be ignored, the explosive growth in data volumes. As inference processes become more multi-staged and prolonged, and the volume of generated data itself increases exponentially, there is a growing requirement for the overall system to efficiently store far larger amounts of data. Addressing the specific needs is the LC Series, our high-capacity SSD equipped with QLC flash memory. The LC Series utilizes the 2-TB QLC of our eighth-generation BiCS FLASH. Leveraging high advanced packaging technology, we successfully stacked 32 memory chips within a single compact package. This enables an industry-leading capacity of 245 TB within the standard E3.L form factor. We have now commenced mass production shipments of this product, and it is already being deployed in ultra-high capacity servers. This remarkable increase in data density allows for significant storage expansion within limited physical space, directly helping to improve efficiency across the entire data center and to optimize the total cost of ownership. Moreover, it serves as an ideal foundation infrastructure to cost-effectively store the expanding massive data sets and cold data typical of inference AI systems. In short, the LC Series provides the foundational storage capacity required to support the continued scaling of AI infrastructure. As Mr. Yaguchi explained earlier, we are already witnessing a tangible expansion in demand within the flash memory market, driven primarily by inference AI systems. In particular, inference AI systems designed for agentic AI architectures handle an exponentially growing volume of data, including KV caches, RAG vector data, and inference histories. To address this rapidly evolving and expanding market, we offer a comprehensive portfolio. The CM Series for high-bandwidth data transfer, the GP Series for ultra-low latency and high IOPS performance, and the LC Series for ultra-high capacity data storage. Through this strategic alignment, we are actively driving the enhancement of performance and scalability in inference AI systems. With the continuous advancements of AI systems, storage has evolved from a simple data repository into a critical component that dictates the performance of the entire system. Moving forward, in addition to developing cutting-edge NAND flash memory and SSD technologies, we are strengthening our total solutions capabilities, including the seamless integration of storage and memory. Through these initiatives, we will reinforce our competitive edge in the next-generation AI infrastructure market and drive our mid to long-term growth. That would be all from me. Thank you very much for your attention. Next, Mr. Inoue, please. I am Atsushi Inoue, head of our Memory Business Unit. I'd like to outline our BiCS FLASH development strategy for the AI era, focusing specifically on the latest developments of our 10th-generation BiCS FLASH. Here is the agenda for today's presentation. First, I will discuss our NAND development strategy in response to the evolution of AI inference systems. Next, I will highlight the technological leadership of our cutting-edge 10th-generation BiCS FLASH. Finally, I will provide an update on the progress of our development of that 10th generation. I'll start by focusing on our memory development strategy. As noted in the earlier session on the SSD business strategy, the evolution of generative AI is diversifying customer requirements for both SSDs and NAND flash memory. To address the rapidly expanding inference market, we are advancing our product development with a focus on three core areas. First is addressing the need for greater efficiency in inference processing. For the Context Memory Storage shown on the left, KV cache, which stores the previous calculation results, must be swapped with HBM at high speeds. This requires high bandwidth read and write performance. In this domain, TLC NAND is the optimal solution. Second is meeting the demand for improved response accuracy. Processes such as vector database generation on RAG servers, shown in the center right of the slide, demand low latency and exceptional random read performance. Here, SLC or MLC-based XL-FLASH is the optimal solution. Positioned as an HBM extension, this technology will be deployed in our super high-performance GP Series SSDs, which deliver exceptional IOPS. Third is managing the explosive volume of data generated by AI. We are driving the development of high-capacity QLC NAND tailored for high-capacity SSDs. These serve as a primary storage medium for the exponential surge of data generated by RAG server database queries and agentic AI. The common requirement across all three of these approaches is performance per watt. In other words, power efficiency. In the data infrastructure of the AI era, achieving low power consumption to boost power efficiency is of paramount importance. As we develop BiCS FLASH, our 3D flash memory, our primary focus is to deliver high capacity, high performance, and high power efficiency or low power consumption. This builds upon the foundational value of cost competitiveness that the market has historically demanded. I will outline our path to achieving this high capacity and high performance guided by our BiCS FLASH roadmap. Our eighth-generation BiCS FLASH serves as a starting point for our future product development roadmap. A defining feature of the eighth generation is its adoption of CBA technology. This optimizes wafer processing by separately manufacturing the memory cell wafer and the CMOS circuit wafer before bonding them together. By enabling industry-leading performance, capacity, and power efficiency, this technology has earned exceptionally high praise from our customers. We are currently scaling up mass production for numerous projects, and by the end of fiscal year 2026, we plan to transition approximately 80% of KIOXIA's annual gigabyte outputs to this eighth generation Additionally, with the eighth generation, we became the first in the industry to establish the technology that allows 32 dies to be stacked within a single package. This breakthrough has enabled us to achieve a massive eight terabyte capacity in a single small form factor BGA package. This high-density packaging is key to delivering our 245-TB LC9 SSD for the generative AI market. Building upon this eighth-generation foundation and to address the diverse needs of the AI era, last year, we introduced a dual-axis strategy for our product development. The first axis of development, shown extending to the upper right of the graph, is a pursuit of high capacity and high density. For our 10th-generation BiCS FLASH and the subsequent 11th generation, we will continue to combine increased vertical stacking with lateral scaling. This will deliver high bit density and massive capacity products tailored primarily for the enterprise and data center SSD markets. The second axis of the development is the pursuit of high performance. For our ninth-generation BiCS FLASH and the subsequent generation Y, we will combine cell technology featuring an optimized layer count with the latest CMOS technology to achieve exceptional performance. This will primarily target the needs of the AI-enabled PC and mobile device markets. Today, within this dual-axis strategy, I will focus specifically on our 10th-generation BiCS FLASH. This latest development is primarily targeted at the next-generation AI inference market. Over the next two slides, I will explain the technological leadership of our 10th-generation BiCS FLASH. First, this section addresses the relationship between layer count, cost per gigabyte, reliability, and power efficiency. As illustrated in the diagram on the left. Conventional 3D flash memory has traditionally relied on increasing the vertical layer count to enhance the bit density and achieve cost reductions. However, excessive stacking requires substantial capital expenditures and additional manufacturing steps, which ultimately increases wafer costs. For the 10th generation, as with the eighth generation, cost reduction is not pursued solely through increased vertical stacking. By incorporating lateral scaling technology in a well-balanced manner, the gigabyte cost has been optimized with a lower layer count than the industry average. Furthermore, excessive vertical stacking introduces two performance-related concerns. The first is the degradation of power efficiency, and the second is the decrease in reliability. As shown in the structure on the left, with extreme stacking of 400 layers or more, the number of memory layers activated during data read and write operations is higher than a 332-layer architecture, which tends to increase power consumption. The second concern involves the reliability of the memory cells. To stack 400 or more layers, the individual memory layers must be made thinner. This reduces the electrical charge that each cell can retain, thereby degrading memory cell reliability. For the 10th generation, selecting the optimized 332-layer architecture has successfully improved power efficiency and secured memory cell reliability. The table on the bottom right compares our 332-layer 10th generation product with a case where a product is developed with 400 layers or more. By limiting the layer count to 332, estimates show that the gigabyte cost is approximately 10% lower. Additionally, it achieves a 10% advantage in power efficiency and a 35% advantage in memory cell reliability. In conclusion, rather than simply pursuing a higher layer count, the 332-layer 10th generation BiCS FLASH has been optimized from multiple perspectives, including gigabyte cost, memory cell reliability, and power efficiency. Consequently, this architecture has produced a highly competitive product in the market. The second technological leadership is the CMOS directly Bonded to Array, or CBA, technology. By inheriting the CBA technology, which was introduced early in the eighth generation, and applying it to the 10th generation, we're able to maintain our industry-wide advantage in interface performance, a key performance indicator following the success of the eighth generation. As shown in the diagram on the left, CBA technology was introduced beginning with the eighth generation BiCS FLASH, representing a market launch approximately four years ahead of the industry average. Consequently, an interface performance of 3.6 Gbps was achieved two years ahead of the rest of the industry. Similarly, an interface performance of 4.8 Gbps is expected to be achieved approximately one year ahead of the industry by implementing it in the ninth and 10th generations. Our deployment of high-speed interface products from an early stage effectively differentiates our product portfolio. As illustrated in the diagram on the right, by 2029, the proportion of our products supporting 4.8 Gbps within our product mix is projected to be approximately 20% higher than the industry average. This will enable us to precisely address the requirements of growth markets, such as high-performance SSDs supporting PCIe Gen 6 and Gen 7, that demand high bandwidth for future generative AI markets as well as high-speed mobile applications. Finally, I'd like to update you on the current development status of our 10th generation BiCS FLASH. Development of the 10th generation BiCS FLASH 1 TB TLC product is progressing smoothly. We are currently conducting reliability testing, the final phase of development, and are targeting to commence sample shipments this summer. As shown in the table on the left, we've confirmed steady performance improvements across key metrics compared to our eighth generation products. Specifically, bit density has improved by 59%, interface speed by 33%, data read throughput by over 15%, and write throughput by over 30%. Furthermore, regarding power efficiency, which we projected to improve by 10% at last year's corporate strategy meeting, we've achieved substantial gains surpassing a 40% improvement in read operations and a 30% improvement in write operations. The 10th generation 1 TB TLC BiCS FLASH will be deployed in high-performance, high-bandwidth SSD products, including KIOXIA's next generation CM series SSDs. With this highly competitive NAND product that will support the ongoing evolution of AI, KIOXIA is positioned to precisely meet the demands of the market. That concludes my section. Thank you. The next speaker is going to be our CFO, Mr. Kawamura. Hello, my name is Kawamura. I am the Executive Vice President and Executive Officer. In the end, I would like to talk about our financial strategy to support our growth. Please take a look at the agenda. I would like to talk about the five themes. First is further enhancement of our profit-generating capabilities. I would like to talk about the quality of profit. The second is the structural business transformation. I would like to talk about our growth curve moving upwards significantly. Next, I would like to talk about achieving best-in-class capital efficiency. The fourth point is maximizing cash generation and strengthening financial position. We have been able to significantly improve the balance sheet. The balance sheet has been significantly reinforced. Last but not least, I would like to talk about the capital allocation. Related to that, I would like to talk about our thoughts around shareholder return. Please move on to the next page. I would like to talk about how we plan to elevate our quality of profit. Mr. Ota, Yokotsuka, and Mr. Inoue have already talked about this. In summary, the demand for data center will be captured. We want to make sure that we transition to a structurally stable revenue model. To further reinforce our activities around this, we would like to invest on CapEx, R&D, and human capital investments to further elevate the quality of our earnings. High growth, high margin means we are going to be able to generate profits through our technology. We are able to capture new customers and new markets. The earnings from that we define as high-quality earnings, and this ties to high profit. To be more specific, I would like to talk about the three points. Starting from the left, market, technological leadership, and cost competitiveness. In the red font on the bottom for the market, there has been a change in our portfolio and the technology leadership. We would like to further advance our R&D activities. For cost competitiveness, now we have migration. We basically want to shift our resource to the new generation. Starting from the left, our data center products are expanding. The conventional smartphones and computers were the traditional part of our business, but we are moving to a more focused sector, which is industrial and enterprise customers. Through long-term agreements with our customers, we have been able to stabilize our income. The investment decisions have become more accurate due to this, and the portfolio mix has changed due to this. The revenue contribution rate for data centers is expected to exceed 60% by FY 2028. If you can take a look at the middle of the slide, technology leadership. We are focusing on technology leadership for R&D activities. Against FY 2025, we would like to achieve over 60%. We would like to focus our investment on advanced technology. We would like to deliver high-value added products, a super high IOPS SSD for AI systems, and we also want to expand high-capacity QLC and high-performance products, which are 4.8 Gbps. Moving to cost competitiveness. We want to accelerate our migration to next-generation products. The ratio of BiCS FLASH generation eight will exceed 80% by the end of FY 2026. This is the target that we have set for FY 2026. On the next slide, I would like to talk about the structural business transformation. Starting from the left in pink, we have our trajectory of revenue from FY 2019- 2025. The line chart would show our non-GAAP operating margin. If you can take a look at this table, you can see that the pink bar has become larger when we compare left to right, and we also see a big curve in the line graph. In the right side, you can see the blue bar chart as well as the arrows. This blue bar would be multiplying the Q1 guidance of FY 2026 by four. The revenue size has significantly increased. If you can look at the dotted line on the right, the memory business will continue, and so there will be a cyclicality, but the volatility has been reduced. The growth curve has already been elevated towards the top of the chart, and there will be cyclicality, but you can see that the arrow is moving towards the upper right. We can assume that we may be entering into a super cycle. Even so, there are four drivers that we can think of. One is the market. We are shifting from consumer-driven cyclicality to AI infrastructure-led stable growth. We will be able to shift to a more stabled structure. For the business model, we are shifting from a single-year LTA to multi-year LTA, which will enable revenue visibility and stability. For technology leadership, the NAND technology that we have pioneered will be maintained. CBA, CMOS Bonded Array will enable us to have a technological advantage, and we also will be launching high-value added SSD products so that we can maintain to be the market leader. Cost and capital efficiency will also be reinforced while we maintain industry-leading cost and capital efficiency through technology advantages and operational scale. Please move to the next slide. We are working to achieve best-in-class capital efficiency. We have a very strict investment criteria, and we will structurally shift to a high ROIC model. Return on invested capital. The capital cost average will exceed with return. Please take a look at the bottom of the slide. The ROIC for FY 2024 was 18%. This was raised to 31% in FY 2025. When we look at the first quarter of FY 2026, we are currently expecting to be over 60%. If you can take a look at the bottom, we have the weighted average cost of capital, WACC. It's about 10% or maybe in the teens. We will add our investment hurdle rate on top of this, ROIC and WACC. In between the two, we call it the spread, and it's really about how we can elevate this level, and you can see our efforts. If you can take a look at the right side, we are targeting to be the best in class ROIC. We will concentrate on R&D and human capital investments in high growth, high margin sectors to further elevate a higher ROE. For investment discipline, the hurdle rate will be controlled sufficiently, and we will only fund projects that exceed the hurdle rate. Through our efforts around this, we want to achieve a significant improvement in both EPS and free cash flow per share. Moving to the next slide. This is the slide that talks about maximizing cash generation and strengthening financial position. In the caption, you can see that we are working to achieve both the financial soundness while we invest for future growth. These are contradictory to each other, but we want to achieve both so that we can further elevate corporate value. If you can take a look at the left, we have our assets, and the right would show our liabilities and equity. When we take a look at the assets on the left, you can see there are two bars. The left would be FY 2025, and the right is the illustrative FY 2026 trajectory based on Q1 run rate. For FY 2025, the total asset was JPY 3.690 trillion, and this is going to double by the end of the fiscal year. We haven't been able to quantify the actual figures yet, we believe that the total asset is going to be almost double of what it is and what is actually going to increase. If you can take a look at the comparison, cash will increase. Liquidity for future CapEx and R&D will increase. This will further make our financial position sound. BiCS8 and BiCS10. We will be investing on these two technologies, and this would double our asset. Moving to the right, liabilities and equity. For FY 2025, we were at JPY 3.69 trillion in liabilities and equity. If you could take a look at the right. You can see that there is a substantial increase, especially in the blue bar. This would be our equity ratio. FY 2025 was 38%, and this is going to go to the late 50%, close to 60%. Our financial position will be very sound, we can make proactive investments. Lastly, I would like to talk about our capital allocation policy and our thoughts around shareholder return. We're still in the middle of the fiscal year, how much cash flow will be generated in the end cannot be quantified, I would like to share our thoughts. Starting with our thought. We want to allocate based on cash flow. We have the blue box in the middle. This is the excess cumulative free cash flow. There are cash generated, as you can see in the caveat on the bottom, there is investment already made in four areas. One is CapEx. The second would be R&D and human capital investment. The fourth would be a reserve for operational needs. All of these costs have already been dispersed. With that excluded, we have the excess cumulative free cash flow, and we will return based on this. Taking a look at the pink part, our key policy for shareholder return, it will be dividend based, and we plan to start a progressive dividend. The timing, as I mentioned earlier, we haven't been able to fix this year's figures yet, and so we can't give you any specific figures or percentage as well as the timing. According to our current plan, based on the FY 2026 actual, we will start paying dividends from FY 2027 according to what we're examining today. Depending on the cash flow situation, we may examine the opportunity of a dividend in the second half of FY 2026. If, for example, the shareholder return were to be 50% of the net cash flow, about 50% of that would be shareholder return. That's the basic dividend, and if there is excess income, then there will be special dividend. According to the situation of the share price, a share buyback will be examined. The 50% will be shareholder return, and for the remaining 50%, it would be growth investments such as M&As. In some cases, the M&A may not happen, and so out of the net free cash flow, 100% of that can be returned to shareholders, depending on the scenario. Those are all the options that we will be examining as we decide on our shareholder return. That's it with my explanation. Thank you very much for your attention. That concludes our presentation. Now we would like to move on to the Q&A session. If you have a question, please tap the raise hand icon on the Zoom screen. We will appoint you in order. When you are appointed, please unmute yourself to speak up. If you would like to ask questions in writing, please use the Q&A feature in Zoom. Feel free to submit them in either Japanese or English. First, we will take questions from investors and analysts. After that, those from the media outlets. Please allow us to limit the number of questions to one per person. If we have time left before closing, we will take additional questions. We are planning to wrap up the session at 6:00 P.M. Please kindly note that because of time constraint, we might not be able to take all the questions. Today's session is focusing on our mid to longer term growth strategy, so we would like to ask you to refrain from asking questions about the short term or the current business situations. For the information about the recent business situations, please find our presentation material of fiscal year 2025 earnings call on the investor relations section of the corporate website. To the investors and analysts, questions please. From CLSA, Mr. or Miss Yoshida, what's your question? Please unmute yourself to speak up. Thank you. I am Yoshida. Thank you for your interesting presentation today. Just one question. Page 28. My question is about page 28. For the inference server NAND exabyte estimate here with CM series, GP series, and LC series. What about the TAM by product series? What's your forecast of the TAM by series? Also your competitiveness against peers. For example, GP Series, I know that you are collaborating with NVIDIA ahead of competitors. What about the same series, which I believe the competitors are following? How do you see the competitive edge against especially the global peers? Any comments about China competitors as well. Thank you. For inference server or the AI's NAND, by series CM Series, GP Series, and LC Series we presented today. Your question is a market growth estimate by series. Also the competitiveness of KIOXIA's products by series. Mr. Yokotsuka will take a question. Thank you for your question. First of all, I would like to share that inference AI entirely will continuously grow. Not just growing, but also the growth speed is significant. In that context, which area will be outpacing the others? First, LC Series to support the storage capacity using our latest QLC chip. For example, 245 terabyte to support large capacity storage. As data volume grows, we believe that our business size will grow accordingly. As we presented for RAG, a drive product for RAG system and TLC-based CM Series, which will be used as storage. That will show the growth trend. You are asking us about competitiveness. Our drive is serving our enterprise major customers as well as hyperscalers. We've already entered the mass production or the qualification phase. In that sense, that shows that a high technological potential of our SSD. The GP Series, as you pointed out in your question, which will meet the low latency requirement by developing the unique chip, which is ready for mass production. The next generation chip is now planned to accommodate such new emerging demands through the development of NAND flash and controller SSDs. Developing both, we believe that we'll be able to differentiate ourselves against the peers. Not just for customers, but also we will support our ecosystem partners, and that we have strong engagements from those partners. That we'll be able to leverage such strengths through our product development to expand our business. Got it. Thank you. Next. Oh, let me add some comments. In your question, so that you mentioned about China players. Regarding the Chinese players, our understanding is that they are staying mainly in the consumer market within China. Going forward, SSD, especially for data center, they might enter that space in the future. Right now, they are staying within China because of certain characteristics of those players. The China data center is not that big yet, and that they are not developing or controllers in-house yet. It is expected that we can maintain our competitiveness against them. Thank you. Next from JP Morgan, Ms. Shikanai, please unmute yourself to give us question. I'm Shikanai from JP Morgan. page 17, about the production strategy. On the right bottom, you show that you will expand your capacity and almost double to 2028. What about the big growth of CAGR and also the market growth estimate? LTA. I have a question about the LTA and then what's your estimate of the demand based upon LTA. Thank you. Your question is about the production expansion. Let me take your question. Thank you for your question. As Mr. Yaguchi presented earlier, this shows the 22% as expected market growth. Our NAND output should be in line with that expected growth of the market. We have locations in Yokkaichi and Kitakami as a manufacturing base. Those two fabs have still space available. That will make sufficient investment, that space, so that we can accommodate that 22% over CAGR. That's what we are discussing internally, and we are planning to make sufficient investment that way. Year 2029 and beyond, of course, we will carefully monitor the development of the market to make investment decision. Also that we will maintain the discipline in our investment activities. Regarding LTA, at this moment, with multiple, quite a few number of customers specifically, have been proposing the multi-year LTA, mainly from data center and enterprise segments. From those fields, they are demanding, showing that strong potential growth of the inference to the AI server area, which we presented today. Right now we are in the middle of discussion with those customers and with some of them, we've concluded the contracts. There are other areas that we are still negotiating or discussing with them, so that we will refrain from sharing with you any detail. Confirming LTA with the customers will allow us to have higher visibility of investment requirement in year 29, leading up to the lowering the risk. Also that will help our company grow. We would like to make sure that we will accommodate such needs for the long-term agreement. Thank you. Next, from BofA, Mr. Hirakawa, what's your question? Please unmute yourself to speak up. Thank you. I am Hirakawa. You previously mentioned LTA. You say that you refrain from sharing any more detail. I have a question. You said that the NAND has entered the super cycle. You talked about the market growth potential and the technology leadership of KIOXIA, which makes sense to me and quite convincing in your presentation. In order to ensure them, you have to transform the business model to make that happen. The LTA will be the key for business transformation. You say that the LTA will allow you to have the visibility of CapEx. How could we believe what you said, for example, that the LTA coverage over your entire business or that if market changes suddenly, how you will activate or exercise the terms of LTA to accommodate such risks? Can you share a little bit more about the LTA status so that we can be more convinced about what you said in the presentation? Thank you. Your question is about the LTA such as the binding or the visibility or, and also the certainty of LTA. Let me add some more comments about LTA. First of all, thank you for your question. About our LTA situation with an individual customer, because of confidentiality today, we are afraid we have to refrain from talking. But in year 2028, not just year 2028, but even 2029 and beyond, many customers are keen to conclude the LTA, including hyperscalers and enterprise customers. Right now, we are continuously discussing that with them. About the LTA detail of individual customers that we can't say more, but as you commented as part of a question, the LTA will ensure our CapEx, including the price and also the term of the contract, which will allow us to make the wise investment decision. Internally, we are discussing each individual of LTA right now. That will be it for myself. Thank you. From Nomura, Mr. or Ms., Virginia. Thank you. I am Virginia from Nomura Securities. My question is about page 16. Slide 16, especially the left-hand side. Up until 2027, the market is tight. Toward year-end, it seems that the supply shortage will get milder toward year-end. The right-hand side, this is the yen amount or the dollar amount Exabyte basis as well. Year 2026 and 2027, then the Context Memory or the super high IOPS will be launched. That how much of those new technologies you have factored in to create this slide? Your question is about year 2026 and 2027, KV cache Context Memory, and such new technology demands. How much we have factored in such the new technology demands to create this slide? I will ask Mr. Yaguchi to take that question. Thank you for your question. Looking at supply-demand balance, the source is TechInsights it says in this slide. Because we have multiple data sources and always we are monitoring them. Also through our marketing division, which is getting close enough to customers to get customer voices to continuously research the future supply-demand situation. In year 2027, whether the supply-demand balance going to be 100% in a balanced manner or not, then we have to wait and see to be certain. Looking at the entire market supply volume and demand, second half of year 2027 and onward, it seems to be quite good in shape. On the right-hand side, you see in the bar chart regarding ASP, which is consistent with the supply-demand balance. There are nothing really worries us, unlike the year 2023 or so that when we experienced a downturn. That's how we see at this moment. That would be our thoughts around the supply and demand balance. Out of this, how much are for AI, like Context Memory? As Mr. Yokotsuka mentioned, whether it's Context Memory or whether it's for low latency products like GPU. There was a question around the breakdown of this. As it was explained earlier, the big volume, we believe, the QLC will be the driver, but GPU and low latency or high bandwidth will also be drivers. We do need to keep an eye on the market growth and how that will evolve as we figure out what the actual breakdown is. That would be our outlook as of today. Very well understood. Thank you very much. That's it from me. We have a question on chat. We have a question from Mr. Zach Zhao from Renaissance Capital. This is a question for Mr. Ota, our President. From your position as CEO, how do you evaluate the technological leadership with super high IOPS SSD? For the next 12- 24 months, how will the adoption of your products contribute to your revenue? I would like to answer to your question. Thank you for the question. NAND that is used in SSD would be XL-FLASH, and it is SLC based, so it's a specific NAND chip. These chips, at this moment, we are the only player that would have this technology, and so we want to fully utilize this technological advantage and make sure that we are specced in to NVIDIA's StorageNext. We are in the midst of development. Within this year, we would like to be able to offer this to customers as early as possible so that they can evaluate this, and we will wait for their decision to be made. For KIOXIA, against the special demand of AI, we want to leverage the specific technology so that we can open up the market further. This SHI SSD is just one example that we have raised, but we believe that we can make even better proposals. NVIDIA and hyperscalers, we're hoping we can open discussion with them going forward. You also asked about our outlook for the next 12- 24 months. We would like to finish our development as early as possible and go into the stage of being certified by our customers. Once we get to that phase, we would have more clarity on the future projection. This is an offer that is using our BiCS technology. The so-called capacity of NAND, a part of that can be used to manufacture this. We can keep an eye on the market and be flexible in making decisions on how much we want to manufacture, depending on the market demand. As long as we can capture the accuracy of the market demand, we should be able to meet and capture the opportunity. That's it from me. Mr. Wang, thank you very much for your question. Next question is from Mr. Nakamura from Goldman Sachs. Please unmute yourself before you ask your question. This is Nakamura from Goldman Sachs. Thank you very much for the presentation today. In FY 2024, you showed us your financial model and the OP margin of 20% through the cycle was mentioned, and for enterprise SSD, you talked about market share of 15%. As we enter into this new cycle, what are your projections and outlook? On page 39, you talked about the profit margin curve in the new phase. I know this is just an illustration, but I think that the new norm is about 70%-80%. Also regarding enterprise SSD, you mentioned about the strong demand, but I think the current share at the moment is about 10%. How much are you able to raise that, is my other question. The financial model that we showed in FY 2024, the share that we are currently targeting, especially for enterprise SSD and the profit margin target, if there is any, based on the FY 2024 financial model, I think was the question. Thank you very much for your question. I would like to answer to the question first. In FY 2024, as you have mentioned, we did announce the figures that was just mentioned. As you can see in our presentation today, inference in AI will probably drive the NAND market going forward. That has become reality today. The business model that we had in mind at that time has substantially changed. The situation has changed substantially. As Mr. Kawamura explained earlier, the AI inference demand, we also talked about long-term agreement, which is probably related to this topic. We are seeing a very strong demand that is generated from that. At the moment, we have a very high profit margin. Our expectation is that we can maintain this going forward, but obviously, we do need to compete with our peers. We also need to strike the right balance in investment. Therefore, whether this 70% is going to last forever or not, we cannot give you any specific figures as of today. Compared to the level of FY 2024, the overall base has elevated, and as long as we are in the memory business, cyclicality is always going to remain. Even so, we want to have stability and try to shrink the volatility. To your question about enterprise SSD share, at the moment, as you have mentioned, it's about 10%, or maybe a little less than that. At least that's how the survey companies are disclosing. We have very powerful products centered around BiCS8. Obviously, as we have mentioned from before, we want to aim for a share of 15%. We mentioned this in FY 2024 as well. We also have the captive hyperscalers that have their own SSDs. Even with them, we have a very strong position with our NAND chips and NAND device. If we add all of these together, we believe we have a reasonable share. If we further elevate our SSD share to 15%, we believe that compared to the overall NAND share against the D&D market, I think we can aim for a higher share. That's it from me. Thank you very much. Next question, Yasui-san from UBS. You can unmute yourself and ask your question. Thank you very much for taking my question. This is Yasui from UBS. Can you hear me? Yes, we can hear you. Page 15 of your presentation material, you talked about a CAGR of 46% for data centers. I know that this is a third-party data from TechInsights. This AI inference demand, how should we forecast this? This is a personal question that I have because I'm having difficulty figuring this out internally. Maybe the number of agents or the model contextual becoming bigger and the coding AI agent number of users. What are the indices that you are monitoring to get to this figure? What are the key indices that I should also be looking into to understand the overall demand? That's my question. Your question is about our methodologies to estimate the demands of the AI, especially the inference AI business. Let me take your question first. Thank you for your question. About that, we don't have any clear indicator which tells us the inference demands. Unfortunately, we don't have one. Today, Mr. Yokotsuka presented, for example, for CM Series. Obviously it's for CMS, so it's for the demands of CMS. In that sense, we can see that which series is being purchased by the customers because that's the indicator showing which market segment is growing. This last fiscal year to this fiscal year, demand is getting stronger, and also the demand itself by nature has changed. I think that you can feel the same way. Obviously, it's driven by the AI inference supported by NVIDIA, and that's also a part of that strong demands. We kept talking about AI opportunities. Training AI area will remain slightly increasing, and that remains unchanged. That's not driving the capacity demands. Also the traditional servers, of course, are existing, but the CAGR is not that strong. Obviously, compared to those two segments, the inference server areas is driving the entire market. That's how we see the market growth opportunities based upon that. Unfortunately, we can't answer your question precisely and clearly, but that's how we see the market potential right now. Thank you. Next from Citi. Mr. or Mrs. Fujiwara, please unmute yourself to give us your question. From Citi group, I'm Fujiwara. The same question from the previous ones about the Ota, your concept behind it, and then your performance will be stable once you get into the super cycle. In coming three years under such circumstances, how much of your entire business volume will be covered by LTA? When you determine the ASP, what will be your hurdle rates to determine ASP? Your question is regarding the LTA. Any target in terms of the LTA coverage of the volume and/or the ASP? Let me take your question first. Thank you for your question. About the LTA, we kept being asked, and it's quite difficult to take those questions because of confidentiality. Probably in two years we will make sure, we will increase the coverage through the contracts with the customers. We can't specifically tell you how much percent of our entire volume to be supported or covered by LTA, but significant percentage of the entire business should be covered by LTA in the year 2028 and beyond. It's uncertain what kind of products customers would want, so it's hard to tell. Right now, we are seriously discussing with the customers about that. Roughly speaking, it's not a fixed number right now, but this moment, around 50%, I would say, LTA coverage. As time goes, hopefully that portion will increase to make our business more stable. Hope that answered your question. Thank you. Next, from Phillip, Mr. Izumi. Please unmute yourself to speak up. Thank you. Page 11. Growth investment for the higher enterprise value, organic and inorganic investment. Inorganic investment, we have never saw from your presentation. Wherever possible, can you tell us a little bit more about your image of this inorganic approach? You said that if that M&A slips, you said that it's going to be 100% shareholders' return. Is it linked anyhow to that inorganic approach you presented? Inorganic investment, your question is a clearer image about that, what it means. Also that in Mr. Kawamura's presentation about the shareholders' return, the size of it. That was your question. Mr. Kawamura will take your question. Thank you for your question. About the shareholders' return, what I presented is just a concept and no figures decided, and the timing is yet to be decided. We will carefully discuss internally. Regarding inorganic approach with investment. Organic means that there is growth within our day-to-day operations. Inorganic means that M&A type of activities. It doesn't mean that we have any clear cases of M&A or opportunities at this moment. Nothing has been decided yet. Right now in the NAND industry, we are enjoying almost the highest profitabilities. Therefore, likely that we will stay in this lucrative business. Only way to go is the upward or the downstream. For example, upward forward integration means that there is stronger engagement from customers through their appropriate investment. The backward means that some investment to ensure the stable supply. Right now, the widening scope, we are reviewing the global opportunities. Inorganic approach, and also what happens if that inorganic approach slips. That's how we link it to the shareholders' return. Thank you. Next. In the chat, we got Mr. S.K. Kim from Daiwa. Looking at the global major players which are focusing on the DRAM and HBM. In that context, KIOXIA has a clean room space in Kitakami, and also KIOXIA has technology leadership. Looking at the current CapEx, then he believes that KIOXIA can step on the accelerator to invest more in the CapEx. Mr. Yaguchi will take your question. Thank you for your question. Looking at the competitors' investment trends, as you said, Mr. Kim, as we expect ample cash flow going forward, should we step on the accelerator to make more investment? No, that's not how it should play out. Instead, we should continuously maintain the discipline attitude to strike the best balance. If the KIOXIA alone makes too much investment, and that will affect the industry. We have to ensure appropriate level of investment, which is represented by the JPY 470 billion per year to support 22% CAGR of expected market growth. We might play around some volume allocation, overall that our investment should be in line with the market growth. That's our philosophy behind it. Hope that answered your question. Next. Again, from the chat, from MST Financial Services, David Gibson. About the BiCS 8, BiCS10, and his question is about the mass production plan of the roadmap of BiCS10. Mr. Inoue will take your question. Thank you for your question. About BiCS10, as I explained today, this summer, we are planning the sample shipment this summer, and the customers that will go on the qualification activities. Then the one year after that. Of course, it depends upon the market situation, but we will move on to the mass production phase, hopefully. Thank you. Next, from Mizuho, Mr. or Mrs. Yamamoto. Please unmute yourself to speak up. I'm Yamamoto on Mizuho. Can you hear me? Yes. Thank you. NAND market is expected to grow, and then KIOXIA is positioned quite well. Makes sense. Having said that, in your presentation, new building construction, I think that it's time for you to think of the new building to be constructed. I wonder why you haven't mentioned at all about your plan to construct new buildings in the fab facilities. Your question is about the future plan of building the new facilities or the clean room. Let me take your question. This moment, KIOXIA has the Y7 building in Yokkaichi and the K2 in Kitakami. Those buildings are already in place. Half of that space is occupied our day-to-day operation right now of those two buildings. Before building new facilities and Y7 of Yokkaichi and K2 of Kitakami, still we have a remaining half of the entire space we can use to catch up with 22% CAGR of the expected market growth. Beyond that, of course, we will need new fabs. In that case, in Kitakami, probably K3, the third building. To accommodate the 22% CAGR, we will carefully see the right timing to catch up with that market growth to build that third new building in appropriate timing. In FY 2029, 22%, if you want to grow the output 22%, I believe that you will need a K3, or at least at FY 2030, you will need that new building. What about the lead time to meet that timeline? When do you start? Do you have to start the building construction? Let me add some comments. Let's assume FY 2029, but when in 2029, we don't know yet, but sometime in FY 2029 or early in 2029. Of course, we will be able to cover 22% CAGR with the existing open space. After that, FY 2029, beyond 2030, especially beyond 2030, of course, we have to make sure we have the third new building in place. We just started that discussion right now. Got it. Thank you. Now we would like to start taking questions from the media outlets. First, from TV Tokyo WBS, Mr. Mizutani. Please unmute yourself. Mr. Mizutani? What we are hearing is some background noise through your microphone. It seems somebody else speaking. Ms. Tankai, could you move to the quieter place to speak once again? Next, Suga- san from Nikkei, please. This is Suga from Nikkei. Yes, we can hear you. In the latter part of Kawamura-san's presentation, he talked about 50% of free cash flow is a possibility. I know nothing is decided, but this is just an idea that you have in mind. Yes, regarding shareholder return. 50% is that an image that we should have in mind? Yes, this is Kawamura speaking. As I mentioned earlier, whether we are going to return that 50% or not, it really depends on how much we allocate to a growth investment. When we have that discussion, we just said 50% as an example, just to have everybody get an image. It's not a decision made that it's 50%. It really depends on the M&A opportunity and also depending on the timing of if we do a buyback, it will fluctuate. I hope that helps you picture an image. It's not that we are going to provide 50% as a dividend or buyback. In this model, we want to look at shareholder return and growth investment and try to strike the right balance. This 50% is just a virtual example. Understood. Thank you very much. Sorry, the other question that I have is regarding LTA, which has been raised a couple of times already. Is LTA something that you should increase more, or is there any disadvantage in increasing LTA from a company perspective? The LTA coverage, the thoughts that we have as a company, I think was the question. Let me answer to that question. LTAs are concluded with each customer and each negotiation is different. Depending on the content of the contract, there is a possibility that we don't conclude an LTA and instead go with the conventional agreement. It's not a simple answer that it has to be concluded with all customers. As I mentioned already, for customers where we believe we should have close communication, we are looking into LTA as an option. We would like to look at the demand of the customers and the volume that is expected out of each customer and decide whether LTA should be concluded with them or not. That's it from me. Thank you very much. TV Tokyo, Ms. Tankai, are you ready to ask your question? This is Tankai from TV Tokyo. Thank you very much for taking my question. I apologize for the noise earlier. I have a question to Mr. Ota. On page 17, you talked about your CapEx plan. If you can elaborate on the details, please. Also the capacity planning is also raised here. For FY 2028, it says about double of FY 2025. Can you also explain the rationale behind this? Regarding our CapEx plan, you want to ask for some color around this and how the capacity is expected to increase going forward. Thank you very much for your question. As Mr. Yaguchi explained earlier, this year we're thinking of JPY 470 billion of CapEx. Inside this CapEx, BiCS8 further capacity increase and BiCS10 production preparation is included in the JPY 470 billion. In addition to that, we also touched upon the future beyond that timing. As we get closer to FY 2028, we also need to prep ourselves. As we have been mentioning from before, Yokkaichi's Y7 and K3, there are some investments required for the preparation, and so we want to make sure we invest in these facilities so that we are prepared for that new fiscal year. For FY 2025 it was only JPY 250 billion, but we've raised that to JPY 470 billion so that against the CAGR of 22%, we can continuously supply. We want to invest to prepare for that. For FY 2027 as well, we have a very strong tailwind that is blowing, we believe it will continue. According to our current plan, we believe the investment amount is probably going to be the same level as FY 2026. That is it from me. We would like to take the last question because we are running out of time. From Toyo Keizai, Yokoyama-san, you can ask your question. My name is Yokoyama from Toyo Keizai. Am I audible? Yes. I would like to confirm about the CapEx that was mentioned. According to your explanation, capacity increase of Yokkaichi and Kitakami, I think has already been communicated. How about the other areas like outside of Japan, like U.S.? Is that also a possibility? Production sites for the future. Our thoughts around that, I think was questioned. This is going to be answered by Mr. Yaguchi. Let me answer to your question. When we look at five years from now, as our president mentioned, regarding new sites, the expandability and also from an economic perspective, Kitakami is a high potential. Our thoughts would be centered around that. For overseas, the possibility always is looked into, but the construction cost of the building as well as the operation cost, water, gas, electricity cost, how does that look like? Also from a technological perspective, if we bring the technology overseas and try to ramp up during that preparation period, there may be a loss of business opportunity. We also need to look at this from a business continuity perspective. Therefore, for the meantime, we believe that domestic investment is more efficient. That's it. Understood. Thank you very much. Since it is time, we would like to end the Q&A session. For those watching the live stream, please click the exit button on the bottom right and answer to the survey. We would like to ask for your feedback for further improvements going forward. With this, we would like to end KIOXIA's Investor Day. Thank you very much for taking time despite your busy schedule.