Hello, and welcome to Napatech's third quarter 2026 call. Please note that this call is being recorded. After the speaker's prepared remarks, there will be a question-and-answer session. If you would like to ask a question during that time, please press star and then one on your telephone keypad. Thank you. I would now like to hand the call over to Klaus Skovrup, CFO. Please go ahead. Good morning. I am Klaus Skovrup, CFO of Napatech. I am pleased to welcome you all to Napatech's presentation for the third quarter and first nine months of 2026. Joining me today is our CEO, Kartik Srinivasan. Our Q3 report was released earlier this morning on the Oslo Stock Exchange and is also available on the investor relations section of the Napatech website. For your information, a recording of this webcast will be available later today. There will be a question-and-answer session following the presentation. During and after these prepared remarks, you may submit your questions via text on the webcast page, or we can take your questions on the phone. If you would like to ask a question, please follow the instructions on this slide. Please note that this presentation contains forward-looking statements that are subject to risks and uncertainties. Our actual results may differ from those discussed in forward-looking statements. For further information on risk factors, please see company announcement and the slides prepared for this presentation. With that, over to you, Kartik. Thank you, Klaus, and welcome, everyone. I want to begin with the central takeaway from our third-quarter results. Napatech's AI transformation is translating to commercial results. This quarter marks a significant milestone as we commence production shipments in our AI infrastructure business, turning our investments in technology, product development, and customer engagements into revenue. Our strategy is built around the growing importance of efficient data movement in AI infrastructure, particularly as inference architectures become more heterogeneous and compute resources increasingly specialized. We are now seeing that strategy come to life. There are three key messages I want to emphasize today. First, our AI infrastructure strategy has entered commercial production. Our initial shipments establish a foundation for future expansion, supported by a growing range of customer engagements and applications. Second, core infrastructure remains a highly profitable foundation for Napatech. Its established market position, strong margins, and sustained customer demand provide the financial strength to invest in our AI opportunity. Third, we're building for profitable growth at scale. Disciplined pricing, proactive supply chain management, and operational readiness are integral to our strategy as we prepare for higher production volumes. Together, these three pillars position Napatech to capture a growing opportunity in AI infrastructure while building on the strength of our established business. Let me turn to the financial performance that demonstrates this progress. The third quarter delivered strong results across our two growth engines. We achieved the highest Q3 revenue in Napatech's history at $10.3 million, representing 64% year-over-year growth. Core infrastructure contributed $8 million, while AI infrastructure generated $2.3 million, marking our first meaningful production revenue from this business. For the first nine months of 2026, total revenue reached $23.4 million, an increase of 62% year-over-year. Our core infrastructure business contributed $20.9 million in year-to-date revenue, demonstrating the continued strength of our established business. At the consolidated level, gross margin was 55% for the quarter and 63% for the first nine months. Q3 margin was impacted by a one-time temporary lag in repricing of the initial ramp order within AI infrastructure, reflecting higher component costs, particularly memory, and planned introductory pricing for the initial AI infrastructure production shipments. A revised pricing structure takes effect in Q4, improving our gross margin considerably, supporting our unchanged full-year revenue and gross margin guidance. Looking at the quarterly revenue trend, revenue has more than doubled from $4.8 million in the second quarter of 2025 to $10.3 million in the third quarter of 2026. The growing AI infrastructure contribution adds a new dimension to our revenue profile, complementing the strength of our core business. These results demonstrate the progress of our transformation with a strong core infrastructure business and AI infrastructure now contributing production revenue. Let me now take you through each business in more detail. Taking a closer look at our two growth engines. Core infrastructure continues to deliver strong growth and attractive profitability. For the first nine months of 2026, revenue reached $20.9 million, growing approximately 51% year-over-year. We have also secured 13 design wins across markets, including cybersecurity, telco, financial services, and network visibility. These wins demonstrate continued customer adoption and provide opportunities for future revenue growth. Turning to AI infrastructure, we have reached an important milestone. Year-to-date revenue reached $2.6 million, including our first shipments against a previously announced 1,000-unit production order. Beyond this initial production program, we are seeing growing interest from customers across the AI ecosystem, with 15 engagements and technical qualifications, which I will speak to you later about. What makes these two businesses particularly complementary is that they share the same technology foundation. We're taking our proven capabilities in high-performance networking and programmable acceleration and applying them to the new AI infrastructure requirements. Core infrastructure delivers profitable growth today. AI infrastructure expands our opportunity for the years ahead. Together, they provide a strong foundation for Napatech's next phase of growth. Let me now take you through our performance by geography. Our growth is also reflected in the performance across our geographic markets, with both the Americas and the rest of the world delivering strong year-over-year revenue increases. Starting with the Americas, revenue reached $14 million for the first nine months of 2026, representing approximately 40% growth over the same period last year. The Americas is a key region for our AI infrastructure strategy, where we are investing in customer relationships, technical engagements, and the capabilities needed to support future production programs. We are seeing encouraging engagement across enterprise customers, AI infrastructure companies, and strategic accounts, positioning the region for long-term expansion. Turning to the rest of the world, revenue reached $9.5 million, an increase of 117% year-over-year. This performance reflects strong execution across OEM, telecommunications, and enterprise customers, supported by our established core infrastructure portfolio. The attractive margins of this business make a meaningful contribution to Napatech's profitability. What is particularly encouraging is that we see opportunities to expand both businesses across our geographic markets. Our established customer relationships, global commercial experience, and expertise in programmable networking provide a strong foundation to introduce new solutions as customer requirements evolve. As AI infrastructure adoption broadens globally, we see opportunities to extend our reach across regions, customer segments, and applications. Our geographic footprint is an asset that supports both profitable growth today and expansion of our AI infrastructure opportunity worldwide. Let me now turn to an important development in AI infrastructure and what it means for Napatech. AI inference is increasingly becoming a team sport. What do I mean by that? We see two major industry shifts that are expanding our market opportunity. The first is workload disaggregation. Rather than performing every task within a single integrated system, AI infrastructure is evolving towards specialized processing, where different parts of a workload can run on different compute resources. This includes functions such as attention, inference processing, and memory management. As these functions become distributed, efficient data movement between them becomes increasingly important, and we are already seeing the benefits in practice. One of our leading AI infrastructure customers recently demonstrated a system combining GPUs and specialized AI accelerators, with Napatech technology providing connectivity on both sides. The demonstration showed approximately 50% lower time per output token compared with GPU-only configuration testing. This is a system-level result that illustrates the potential of combining specialized computing with efficient data movement. Also note that Napatech products are designed for scale-out networking on Ethernet and not for scale-up, where technologies like NVLink Fusion, UALink, and ESUN are the options. The second shift is vendor diversification. Historically, AI computing has been dominated by a single GPU ecosystem. Today, hyperscalers and specialized semiconductor companies are investing heavily in alternative AI compute architectures. We are moving toward a market with a wider variety of CPUs, GPUs, specialized accelerators, and custom AI processors. Importantly, these processors do not all need to operate within the same tightly integrated ecosystem. This is significant for Napatech because our programmable networking technology can support different architectures and different approaches to building AI infrastructure. As customers adopt new generations of processors, the need for efficient data movement across systems and between specialized resources remains an important opportunity. Our opportunity is therefore broader than any single customer, processor, or accelerator architecture. There is another important dimension to this opportunity, the longevity of AI infrastructure deployments. While traditional data center infrastructure has typically operated on three-to-four-year refresh cycles, AI infrastructure deployments can extend to five to seven years. For Napatech, this creates the potential for longer product life cycles, more sustained customer relationships, and greater revenue visibility once we secure a design win. Together, workload disaggregation and vendor diversification expand the market for Napatech's programmable data movement technology. Let me now show you how the industry shifts are translating into customer opportunities. We are translating the expanding market opportunity into customer engagements. We have built an AI infrastructure pipeline of 15 selected opportunities across three distinct application areas. Nine are focused on scale-out networking, four on AI storage acceleration, and two on AI network observability. These opportunities are at different stages of maturity, from initial evaluation to qualification and design engagement. During the same period, we have taken our first AI infrastructure customer into commercial production. Now, why are these customers engaging with Napatech? The answer is programmability. Every customer has unique requirements for how data needs to be moved, processed, or observed. Some need to move data between AI processors with extremely low latency, others need to accelerate data processing for storage applications, and others need greater visibility into their AI infrastructure. What they share is a need to tailor these capabilities to their specific architecture without compromising on performance. This is where Napatech stands apart. Our combination of programmable hardware and software enables customers to optimize solutions for their individual needs while delivering the performance their applications demand. As AI infrastructure becomes more heterogeneous, this flexibility becomes increasingly valuable. Additionally, what makes this particularly exciting is the deployment scale. Individual customer programs could represent approximately 5,000 units annually at scale, with some larger opportunities potentially exceeding 100,000 units per year. Accumulating the units across the programs and based on our lead customers' aspirations of 1 GW by 2029, we can potentially expect to see orders starting 2027 of over 100,000 units for future deliveries. These are illustrative program sizes, not forecasts. Each opportunity has its own development and commercialization timeline. We are also strengthening our go-to-market capabilities by expanding our partner ecosystem. Building on our existing relationships with Altera, AMD, and Intel, we are adding Broadcom and Arm to extend our reach across a broader range of AI infrastructure architectures and opportunities. Think about the progress we have made in just nine months. We have developed a broader customer pipeline, expanded into multiple AI infrastructure applications, added more strategic partners, and entered commercial production. We have achieved this while continuing to grow our profitable core infrastructure business. We have moved from AI ambition to commercial execution, and now we are building the foundation for profitable growth at scale. With that, I will hand it back to Klaus for a deeper look at our financial results and outlook. Thank you, Kartik. As Kartik mentioned, we delivered strong revenue growth and improved operating performance in Q3, now being close to break-even. Our revenue in Q3 was $10.3 million. In Danish kroner, revenue increased by 67% year-over-year to DKK 67 million. This growth reflects continued demand in core infrastructure and the first commercial shipments in AI infrastructure. Our gross margin was 55%. As Kartik mentioned, the Q3 margin was impacted by a one-time temporary lag in repricing of the initial ramp order within AI infrastructure, reflecting higher component costs, particularly memory, and planned introductory pricing for initial AI infrastructure products and shipments. A revised pricing structure takes effect in Q4, improving our gross margin considerably and supporting our unchanged full-year revenue and gross margin guidance. For the first nine months, revenue increased 59% to DKK 152 million. The gross margin was 63% compared with 69% in the same period last year. Staff costs and external costs were DKK 39 million in Q3 compared with DKK 35 million a year earlier. For the first nine months, these costs declined around 2% to DKK 124 million. EBITDA was close to break-even in Q3 compared with negative DKK 6 million last year. For the first nine months, EBITDA improved to negative DKK 25 million compared to negative DKK 55 million last year. We expect to make a positive EBITDA in Q4. Free cash flow in Q3 was negative DKK 17 million, mainly reflecting increased working capital. For the first nine months, free cash flow was negative DKK 58 million compared with negative DKK 52 million last year. Net working capital increased from DKK 99 million at the end of Q2 to DKK 160 million at the end of Q3. Receivables increased by DKK 18 million, reflecting backloaded invoicing again in this quarter and continued inventory buildup to support future sales. This was partly offset by an increase in payables and provisions. Cash and cash equivalents were DKK 45 million at the end of Q3. Including undrawn committed credit facilities of DKK 22 million, total available liquidity was at DKK 67 million. Despite the improvement in operating performance, higher working capital absorbed cash during the first nine months, and the negative year-to-date working capital adjustment was at DKK 33 million. We continue to manage the business with a strong focus on cost discipline, cash preservation, and ensuring that working capital supports expected production ramps and customer demand. We do expect that Q4 will be cash positive. Our guidance for 2026 is unchanged. Our revenue guidance is still DKK 200 million-DKK 240 million, corresponding to approximately $32 million-$38 million. We continue to see demand move towards higher speed, higher value products, which carry higher average selling prices. That also means a higher profit per card. We expect to end at the lower end of the unit range. We maintain our gross margin guidance of 60%-70%. We maintain the guided range of DKK 170 million-DKK 180 million for staff expenses and other external costs, but expect to end at the lower end due to lower spending in the first nine months. Staff costs transferred to capitalized development costs are maintained to be DKK 5 million-DKK 8 million. At the midpoint of our guided ranges, EBITDA would be negative at around DKK 25 million for the full year. As always, all outlook remains subject to normal risks, including currency movements, market uncertainty, trade barriers, and supply chain volatility, which we continue to monitor closely. As we wrap up today's presentation, we would like to invite you to visit Napatech at one of these upcoming events. Our full-year event plan is shown online at the link provided. If you happen to be in one of these great cities during the coming period, we would love to meet you in person. With that, we are now ready for the Q&A. Operator, we are now ready to take the first question. Thank you. We are now opening the floor for question and answer session. If you would like to ask a question, please press star followed by one on your telephone keypad. That's star followed by one on your telephone keypad. Your first question comes from the line of Christoffer Wang Bjørnsen of DNB. Your line is now open. Good morning. Can you hear me? Yes, we can. Yep, we can hear you, Christopher. How are you? Great, thank you. Just starting off on the, I think you made a comment, if I heard it correctly, that you are looking to broaden your kind of capability set by adding silicon from partners like Broadcom and Arm. If I heard you correctly, can you maybe expand a bit on that? Does that mean that you are adding potentially a part of your portfolio which is not necessarily reliant on FPGAs and is more special purpose networking ASICs from the likes of Broadcom? Just found that very interesting. If you could kind of expand a bit, it would be helpful. Yeah. Fantastic. Thanks again for your question, Christopher. Our expansion with our partnership with Broadcom and Arm is not necessarily just purely based on silicon, procuring silicon from them, but it is also a go-to-market expansion where they will actually take us with them to some of these joint discussions and joint customer engagements. The engagement right now as a partnership does include procuring silicon from them as well to provide some sort of joint solution where we can leverage their expertise in their silicon, their expertise in their software, along with Napatech's exclusive expertise in our FPGA programming. You can imagine that we are jointly trying to solve customer problems where there is not a solution today, leveraging both their unique expertise and our unique expertise towards solving joint problems. That is clear. Thank you. Then on this helpful slide that you introduced on the customer engagement, can you give an update on where you are with the tier 1 server OEM and how you think about the kind of the timing or lead time to potential conclusion? We will kind of potentially learn more about what is meant by AI storage acceleration. Yeah. Any updates since last quarter? Yeah. As I mentioned in the last quarter, obviously things are going slower than what all of us collectively would like, but directionally, things are still as positive as they were. That is where I am encouraged by the progress that we have made and we continue to make. As you can see in the pipeline, obviously I am not at the liberty of sharing any names given the relationships and the contracts that we currently have with a lot of these customers in terms of NDAs. But what I can tell you that the tier 1 server OEM partnership is moving and making progress. As always, the when, Christoffer, is the difficult question for us to answer, but the progress is measurable. The milestones that we are tracking towards progress, towards a pilot production, they are all in line. All right. Then finally for me, I think, if I heard you correctly, you stated that in next year you could see aggregate orders moving beyond 100,000 units, though for future delivery. Based on that customer engagement matrix, could you maybe give some more color on what kind of applications? Is it like scale-out? Is it AI storage acceleration, observability? Which end segments would you have to be successful in to kind of drive that kind of demand in terms of orders for next year? Is it just simply your current in-production customer moving faster than initially expected, or is it that you are also seeing another runner-up potentially contributing meaningfully to orders next year? Yeah. Good question. Frankly speaking, the single customer that we have taken into production, they themselves can provide us that level of orders, Christoffer, in the 2027 timeframe. I am sure you guys are tracking the power ratings that they are aspiring to build. That being said, that is not how we are planning it. We are fully expecting the collection of these different customers that we are taking through this different stages of our pipeline amounting to that level of orders that we will get into 2027, right? So that is the collection of what we have currently shown here in production, design engagement, and qualification, as well as the other 12 that we are showing in initial evaluation that will continue to proceed. So a combination of that is what I was talking about that has that potential. I guess I will jump in the back of the queue. Thank you. All right. Thank you, Christopher. Your next question comes from the line of Øystein Lodgaard of ABG. Your line is now open. Good morning. I thought I would start with the new table that you presented with your AI design win engagements. Very interesting slide. Can you maybe say how has this progressed during the year? How many of these engagements did you have when the year started? And maybe if you could explain to us how do you define qualification and define engagement? All right, perfect. Good morning, Øystein, and thanks for your question here. I will start with some of the easier ones in the list of questions that you had. When I started, obviously, we had the one engagement with our AI infrastructure, which I would have said was in design engagement phase, and now we have proceeded, progressed that into production. The other one was in the tier 1 server OEM that we were talking about. That was in qualification phase, and now we have proceeded that, progressed that into design engagement. That is the way I would encourage you to look at it. The rest of it is all new, right? If you look at the 13 other entries that have made their way into this table, those are the pursuits and engagements that we have unfolded this year. Very, very exciting year for us in that regard. If you look at the different stages in here, the qualification stage, it kind of results in identifying use cases and almost identifying a statement of work that both the company, the customer, and the vendors have agreed upon, the vendor being Napatech here. The design engagement stage is when we are into the execution mode of the statement of work. Of course, production is we have gone through, executed on the statement of work, the end customers have qualified it and they are happy with performance, and then we are into production. Very interesting. Can you say something about I know things have taken a long, longer time with the tier 1 server OEM than you expected, and that is still a bit uncertain on the timing. Generally, how long time would you expect for it to take to go for a customer from initial evaluation to qualification, and then from qualification and into production? Yeah, so in this regard, Øystein, in the data center, especially in the AI data center business, there is no such thing as a typical progression. Each customer can take their own kind of time, as you guys have been patient enough to see. The one that we did have in production started about three years ago. The one that is not in production yet also started about three years ago. So you can see that this is an involved design win process, but not all of them are built equally. As an example, you can already see that one of them is in qualification for the scale-out networking application, and that, in my mind, is already a pretty reasonably aggressive progress for us. Having said that, our goal, Napatech's goal, is to find the uniqueness in each of those initial evaluation customers that we have. Where do we add unique value, and what can we develop into the qualification and design engagement stage as soon as we can? Because time to solution, time to revenue remains our focus, but each of these customers, each of these engagements, Øystein, is pretty involved. Hmm. Interesting. And here on this AI network observability segment here, that is a bit new. Can you say what kind of volumes are we talking about with those types of opportunities? Yeah, so I will give you a kind of quick briefing on what that is, and then we can quickly address the volume. Our entire 20 years of Napatech as a company, we have existed in providing this feature called network packet capture. And that network packet capture today, we work in the core infrastructure business as the way we have segmented it. Some of our customers have liked the idea of having this kind of network packet capture visibility, observability, and telemetry information even in the AI infrastructure. Especially as the AI infrastructure evolves into a heterogeneous deployment, each of the vendors in that heterogeneous deployment, whether you are a GPU, whether you are an accelerator, come with your own set of telemetry tools. But when you are a big data center, you need a consistent tool for visibility and telemetry, and that is where our DNA, Napatech's unique DNA for packet capture is proving to be very valuable. So you can imagine that each of the data centers that deploy AI infrastructure will be looking to deploy a solution like this once we can get this up and running. Thank you very much. Thank you very much. Last question from me on the gross margin. You say that you have repriced your products and that it should be up materially in Q4. Can you say, is that from the beginning of Q4? We should think Q4 will be normal in terms of gross margins, or has it happened kind of somewhere in the middle of the quarters, you only get part of the effect in Q4, then we are back to kind of more normal levels, whatever that is, in Q1? Right. Yes, important question. I am sure this is on everybody's mind, and I am glad that you asked it. I do want to start by saying that the Q3 is an isolated event, right? The temporary lag in price that we were adjusting to, plus the initial orders that we had that we delivered to our customers, those are the two reasons that the Q3 gross margin, like we said, dipped a bit. Pricing arrangements across the customer base, especially on the AI infrastructure, is already in place, so you will see the recovery in Q4. Thank you very much. Your next question comes from the line of Christoffer Wang Bjørnsen of DNB. Your line is now open. Yes, thanks for letting me back on. Just with regards to that 100,000 plus orders, can you maybe talk a bit about how kind of staggered or distributed do you think that will be in terms of how you would need to deliver that? Just thinking like from a working capital point of view and so on, it would be a pretty big step up compared to what seems to be your current capacity, both in terms of balance sheet, but also in terms of your current supply chain. Just like how you think about that whole bridging that. Yeah. Good question. What you do in the supply chain to set yourself up for that. Yeah. Thank you. Good question, Christopher. That kind of goes back to the third pillar that I opened the presentation with. We are preparing ourselves for that level of scale, for that level of profitability, and the way we're going to look at this is, of course, ensure that the working capital supports the demand and supports the way we provide these customers a seamless kind of delivery. Towards that, we are also getting creative with our own customers. The terms of engagement with our customers have already been in discussion about some level of prepayments that we can get in place with them. So we are planning to fund this on our own, and that's the kind of discussions that we're having with our customers. There is absolutely no interest or kind of path right now that we are looking at in doing anything else in terms of raising capital. Our entire work, our cash flow here, and our ability to deliver these kind of huge orders is in my self-sustained, self-reliant manner. Thank you. Just turning to the core infra segment, which is kind of being this important part of the business right now. It seems to be doing really well, and just interested in hearing your reflections on what kind of upside potential there is in that business. Like, how do you look at that business compared to when you came into Napatech and the historical numbers there? It just seems like it is not only AI which is kind of expanding. You are seeing it in high-frequency trading is booming. You are seeing cybersecurity obviously being on top of everyone's mind right now. Just like, how are you thinking about the opportunity within core infra, which I think many times can be a bit neglected or underappreciated, where you have a strong legacy and track record, I guess? Yeah. Christopher, thanks for being an avid supporter of our core infrastructure business, along with all the excitement about the infrastructure growth. I do believe that the AI as a technology is obviously all-pervasive. Every group, every enterprise, every department is deploying it, and so is Napatech, as an example. That is impacting the core infrastructure business for us, because there are overhauls, there are further purchases being made on the core infrastructure side, which has given us a good boost on our products there. That is something that I fully expect we will continue seeing as the rest of the enterprise kind of evolves and adopts AI as a technology. The second thing that we are doing within Napatech to take advantage of this is we are building a roadmap towards a bigger solution portfolio in that space. Today, we are working with one product called Capture that serves the core infrastructure business, but we are evolving, growing our portfolio of products, not just to grow in the AI infrastructure space, but also to leverage the expansion in core infrastructure. All right. Super. Thank you. Thank you so much. I'd now like to hand the call back to the management to address the webcast questions. Thank you very much. We do have a number of questions coming in here. Some of them are overlapping, but I think what we will do is we'll just take them chronologically and then try to go through them as we go. First question is coming in from Niels Foldal. He's asking, "There has been a lot of speculation regarding Raptor. Could you give some color on Raptor from Napatech's perspective going forward? Hi. Thank you, Niels, for your question. Let me answer by saying two things in here. One is, Napatech is fundamentally a scale-out networking specialist. In the world of Raptor, there are two networks that will be involved. One is called scale-up, which will be serviced by what you've heard about their partnership with NVIDIA on NVLink Fusion. Napatech doesn't play in the scale-up domain at all. We are a specialist in the scale-out network. That's what we're working in the current generation as well. We continue to work in that scale-out network, doing differentiated value-added solutions with that customer right now. In fact, we're almost exclusively working with them on the scale-out networking and doing some sort of joint development on technologies like RDMA. If we are looking at the Raptor solution from our tier 1 AI customer, we are exclusively working with them today on RDMA-based solutions that will unfold in that Raptor timeframe. In fact, that will start unfolding even before that. We are working on the first pass of our RDMA scale-out networking solution even in the Corsair era. I think a similar question from Per Lau. "Jimnes here, can you confirm that Napatech is part of the Raptor program to d-Matrix? Yeah, so the confirmation portion of this is obviously something that I will do when we get something signed and we are in contract, right? Till then, I am going to play it completely by the rules. What I can assure you is that we are working on developing a value-added feature, much like what we did in the first generation Corsair era before we went into production. It is the same kind of work that is going on for the next generation as well. We are jointly developing a solution that will hit the market in that timeframe, and not only are we jointly working with it, we are exclusively working with them on that kind of solution. A further question here also, "Can you tell something about development in tier 1 OEM?" I think you already covered that. Yeah, I did. Yeah. Then another one here, "Earlier there has been giving a forecast in 2027 of 32,000 shipped units. You still uphold this forecast from second quarter of 2025? Yeah, so we are in the process of putting together our 2027 budget, which will roll into what we will come in and provide in terms of forecast for 2027. I would request your patience till we come and do that in Feb. Then there is a question here from Neil Steer, "You mentioned the new pricing structure introduced in Q4 to recover gross margins. Has the demand elasticity been impacted at all, or is the market/demand simply not impacted by price increase at this point in time? Yeah, it's a very, very good question, Neil. Right now, the availability of products and the pricing of these products is an industry-wide kind of concern. What Napatech has done really, really well is in securing the parts. What we're also doing now is not just securing the parts, making sure that the pricing structure is into Napatech's kind of playbook in terms of margin profile. The pricing elasticity is not something that is being tested right now. People are more interested in just getting the parts so that they can deploy the solution, and that's kind of why we are confident that in Q4, we will be able to recover the margins. Then there's a question here from Pauline Jimnes. "How many of the 15 AI customer engagements are new this quarter? Yeah, so as far as the AI customer engagements, I would say up to maybe 50% of those were in the previous quarter, and the rest of them were this quarter. Our top-of-line prospect is, in a lot of ways, driven by our relationships with these customers and the events that we attend. So, I would say out of those 15 AI customer engagements, we can probably say about six or seven is this quarter. How many of the 13 design wins, which was in core infrastructure, is new this quarter? Yeah, out of those 13 design wins, I think when we came in last quarter, we said we had about 12, so I would say there is one additional one for core this quarter. Yeah, great. There is a question here from Anders Knudsen, "When you talk about the total addressable market expansion workload disaggregation, can you elaborate a bit more on that, please, so we better can understand the expansion that you do see? Yeah, so this is a very, very exciting place for where we are right now, Anders, in terms of the workload disaggregation. Both of those different themes that I spoke about, both workload disaggregation as well as vendor diversification, they translate into TAM expansion for us. Your question specifically about workload disaggregation, essentially what I want you to imagine is, back about six or seven months ago, the workload of AI inference was thought out to be a monolithic workload, meaning you take one workload, one piece of compute architecture, whether it is a GPU or whether it is an accelerator, that is kind of where the entire workflow is managed. Now what has happened is it has become pipelined. Section one of that workload happens in compute one, section two of that workload happens in compute two, section three happens in compute three, and so on. What happens in that case, Anders, is that those different compute engines now need to be connected together, wherein earlier it was monolithic, now it is disaggregated, and the room for Napatech to provide that data movement between those compute elements is the TAM expansion. Good. Then a question here from Svein Erik Sivertsen, "Gimlet Labs announced that it will deploy Corsair in its inference cloud. Could this have an impact for Napatech also? 100%, right? Because wherever you see Corsair being developed or deployed, I would request that you can imagine Napatech is in there, because we are the exclusive scale-out partner for them. Question from Pauline Jimnes, "Do you need CapEx to grow in AI infrastructure, or can the ramp be funded from working capital and your current liquidity?" Yeah, and I can answer that one. We definitely believe that we can fund it through our current liquidity and working capital here. Again here, the only reason for our liquidity going down here in Q3 was that we built additional working capital, and that will, of course, get in again these increase in trade receivables. Another question from Svein Erik Sivertsen. There's actually a couple here. "In the Q2 presentation, it was said that ramp-up was going a little bit slow, but the engagement on the technical level remains extremely strong between Napatech and the tier 1 server company. Can you add some more on how this has developed?" I think you may have covered that already. Yeah, I can just say that it's making good progress. You can see that we are proceeding that through the pipeline, and we're excited about how that is moving. Yeah, another question here on the Raptor program. Klaus confirmed that Napatech will be part of the Raptor program. Does this mean that Napatech will be involved in the cooperation between d-Matrix and NVIDIA? Yeah, I can take that one. The Raptor program is a joint development that we will proceed to do, and as soon as that gets some confirmation about that, we will come back here and provide you that confirmation. Again, NVIDIA and d-Matrix are partnering specifically in the scale-up network. Right now, again, I will reiterate that Napatech's joint development and value proposition is in the scale-out. They are two different networks, and we are playing and uniquely adding value in scale-out. And maybe just to comment, I think we had here confirmation from the d-Matrix management that they expect us to be part of it. We also had their CFO, head of strategy, visiting us in Denmark a couple of months back, where he participated meeting some investors. There, I think his comment was that we see Napatech as a multi-generation partner. Question coming in here from Tore Moland. AI margin, Q3 gross margin was 55%. The report calls it a one-time repricing lag. The slide calls planned introductory pricing, which is what AI margin does the new Q4 pricing give? Maybe we should just stop there. Okay, I can take that one, at least for the first part. There are two different components in here. One is what we are calling a one-time repricing lag, and that is mostly because of the memory components that are so volatile, both in availability and pricing. In that case, what we did for the first part of the shipments that we did for the AI infrastructure company is we honored the pricing and prioritized the delivery of the product, and that also has an impact on how we priced, how we had agreements for pricing, which is what I am calling the planned introductory pricing component of it. Both of those components kind of play into each other. One is a temporary repricing lag, which is owing to the volatility of availability as well as the repricing. The second one is just honoring the introductory pricing that we had with this customer. Hmm. Then there is a follow-up question whether this applies in the rest of the 1,000 unit order or only to the two. Also, I think you can ask also on the 2,000 unit follow-on. Yeah, so when I said that you will see that we already have a new structure in place in Q4, what I want you all to read up is starting Q4, that is kind of how we will proceed. Again, the volatility in pricing makes us be on this throughout the kind of pricing arrangements. So in the follow-on order of the 2,000 that we have announced already, there will be maybe further pricing arrangements or changes so that we stick to our margin profile. Good. Then there is a question here around the inventory, which is up DKK 28 million to DKK 85 million. In Q2, we said that we had enough stock for the high end of the guidance. How much of it is stock brought ahead of AI orders that are not signed yet, and what happens if those ramps are delayed? I mean, the inventory we have is of course to support the future revenue we will be having, but we try to plan, so we do the production builds and also the inventory build alongside here. So we do not have inventory builds for a lot of quarters ahead. We try to plan it so it goes along one quarter ahead or similar like that. But of course, we have already placed orders now due to the long lead times going a long way into 2027 also based on our current forecast and also dialogues we have with our customers. Then there is one here, "How many of the 1,000 units shipped in Q3 was included in the $2.3 million, and how many are left for the Q4?" So I think to that one, it was slightly less of the half of the 1,000 units we shipped in Q3, and then we will be shipping the last part during Q4 here. And then a question here from Pauline Jimnes again. Okay, something about Raptor. Yeah, that is okay. "Lisa VMware, could you provide an update on Napatech status with the d-Matrix Raptor?" I think we did that. Yep. Does your long-term volume ambitions assume Napatech content in Raptor? I can answer that one. So the long-term volume ambitions, obviously when we come back, Lisa, in terms of our plan and what we are providing later, we can give you some more guidance on how that is involved with the Raptor program. But right now, our development towards Raptor is 100% focused on being part of the Raptor era. Yeah. And then a question from Svein Erik Sivertsen, "Can you say something about what you expect of sale in 2027? Yeah. Again, my request to you is that we will come back and give you that level of information when we close out 2026 and start 2027 in the Feb session. Yeah. Then a question here from Håvard Nilsson, "Can explain the ASP dynamics given the memory pricing? Yes. The memory pricing, Håvard, as you have rightly called out, is probably the most volatile portion of the entire playbook for us here, not just the pricing, but also the availability. So, like we discussed when we came here in Q2, our ASP adjustments to a large extent towards our customers is based on the increase in the memory pricing. But unfortunately, the memory price increase does not correspond to the same level of margin percentage increase because that will just be in a zone where it won't make sense. I think Klaus gave us a very good example last quarter when we came here. But the ASP definitely increases across all the components that have gone up in terms of the BOM for us, the biggest portion of that being memory. Yeah, a follow-up question here also that Håvard says, "Focus rather on the gross profit per card." I think that's a good point because, of course, margins will be diluted because we will not be able to pass it fully on with the full margin on the price increases we do see. That said, we are doing price increases on our customers also, and we recently done another one here going into Q4, which is the second we are doing this year for some customers, and even the third was for other customers. So we are definitely trying to pass as much as this on and also earn money on it still. Of course, trying to get as high gross profit per card as possible. And then we think we got the last question here from Benjamin Aune, "If you are able to comment, will Napatech be part of the d-Matrix Raptor collaboration with NVIDIA, and will Napatech be part of the Squadron?" I think we did cover the Raptor, but the Squadron is- Yeah, the Squadron is, just so that we define it, Squadron is a rack-level solution that d-Matrix announced. Yes, so I can definitely speak, like I said earlier, I can speak about the partnership with Corsair because that is obviously inked, and that is something that even d-Matrix is openly talking about. Raptor is a bit out there, and like I said, it is still in development phase. But wherever we are involved with d-Matrix on Squadron and Corsair, you can assume that d-Matrix and Napatech will be there together because Corsair's exclusive and only scale-out solution is Napatech. Our development towards Raptor is ongoing, and we fully hope that the scale-out solution announcement for Raptor will also be there with Napatech. I think that was the last question we got on the webcast here. So I will just go back to the operator. Any more questions on the line? We do not have any pending questions in the conference line. Thank you very much. Then I think we can call it. Thank you very much for participating, everyone. Enjoy the rest of the day. Thank you. Thank you for attending today's call. You may now disconnect. Goodbye.