If you would like to ask a question during the Q&A session, please press star one on your telephone keypad. I will now hand over to Lars Boilesen, CEO, to begin.
Good morning. I'm Lars Boilesen, CEO of Napatech. I'm pleased to welcome you all to Napatech's Presentation for the Second Quarter of 2025. Joining me today is our Chief Financial Officer, Heine Thorsgaard. Our second quarter 2025 report was released earlier this morning on Oslo Stock Exchange, and it is also available on Napatech's website. For your information, a recording of this webcast will be available later today. Slide two, please. There will be a question and answer session following this 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 instruction on this slide. Slide three, please. Please note that this presentation contains forward-looking statements that are subject to risk and uncertainties.
Our actual results may differ from those discussed in forward-looking statements. Slide four, please. Today's agenda will cover four areas. First, we will provide a business status for the second quarter of 2024, with focus on the major news items that capture the notable progress by Napatech towards achieving our goals to transform our business. Next, we'll provide a market update that shares information about the opportunity ahead of Napatech. We will then shine a spotlight on some of our most recently announced wins and how those wins will positively impact our business in the future. We will then provide a detailed summary of our second quarter financial results. Finally, we will conclude with a question and answer session open to the attendees of today's call. Slide five, please.
We have made considerable progress in meeting important mileposts to our strategic plan that strengthen our position as an early leader in the new and evolving mass market for our network interface cards, or NICs. This includes both notable improvements in our existing capture business, and we have also made strong progress toward our design win goals that underpin our growth aspiration to deliver 60,000 units- 80,000 units a year when those designs reach peak production. This has progressed according to plan, and if we compare our position to just one year ago, we find ourselves in a much stronger position. Slide six, please. Q2 was a solid quarter for Napatech. We did revenue of $4.8 million, an increase of 43% year- over- year. First half 2025 revenue was 26% above first half 2024. Our Q2 gross margins remained healthy at 67%, up 2% compared to Q2 2024.
First half 2025 gross margins ended at 68.4%. Q2 EBITDA was negative at DKK 19.6 million. This was an improvement of DKK 7.7 million compared to Q2 2024. Working capital slightly reduced from Q1, DKK 97 million to DKK 95 million. Free cash flow in Q2 was negative at DKK 21.2 million. This was an improvement of DKK 23.4 million compared to Q2 2024. Q2 was a step in the right direction for Napatech. I'm also very pleased to report Q3 is looking relatively strong, and I can already report that our revenue for Q3 2025 will be better than both Q2 2025, and it will also be better than Q3 2024. In support of significant steps forward with our current customers and expanding the design win pipeline, Napatech successfully completed a private placement capital raise of NOK 210 million.
The placement attracted interest from both leading existing investors and new high-quality investors. The proceeds will be used to finance the growth opportunity arising from the announced partnership, pipeline, and design wins. We were also fortunate to expand our Board of Directors with new members holding key executive positions from respected companies, with proven success with customers in the markets that Napatech is navigating today. The partnership with Intel Altera continues to go well. Most of our pipeline in strategic design wins continues to come from our partner Intel Altera. We are pleased to see the existing captive business improve, but even more important is the strong progress we are seeing with our strategic design wins.
One year ago, we announced a design win with Intel Altera and a tier one server OEM to build a next-generation solution to their specification for high-volume use cases as well as the general market. I'm glad to report we consistently met the hardware milepost ahead of plan and exceeded their expectations. We are now involved in the next stage of their process, which includes development and integration of key software to meet their deployment requirements. We are excited that our efforts to date have resulted in expanding our cooperation to additional use cases and opportunities, all of which represent potentially high volume to Napatech. More information will be later in the presentation about this opportunity. We are equally excited to share more information about our latest high-volume design win with D-Matrix, a leader in artificial intelligence inferencing for hyperscale cloud, next-wave cloud, and enterprise data centers.
Our Intel Altera programmable NIC enables AI infrastructure that D-Matrix requires to deliver efficient, ultra-low latency disruptive inferencing across many AI servers. Their customers get the benefit of deploying an industry-standard Ethernet-based scale-out solution to take advantage of D-Matrix's blazing-fast inferencing speeds, all at massive scale. This delivery is also going according to plan. We have also recently disclosed another winning solution across an array of applications and services by leading companies in their respective vertical markets, including a cybersecurity solution with Broadcom Symantec, a network and server infrastructure solution with Eideticom, and an artificial intelligence solution with Myrdl.ai, and Xelera for ultra-low latency transition in financial service markets. These are all examples of Napatech expanding its business to large or high-growth segments that require our programmable NICs.
We are also pleased that one of our UPF 5G packet core design wins has secured a design win from a North American 5G operator that is expected to lead our first production deployment in late 2025. Slide seven, please. In this section, I would like to share an update on our market and how it is positively being impacted by artificial intelligence. Artificial intelligence is not just the buzz in the present; it is the future. It is the majority of the future of all our segments we serve from network infrastructure, storage, cybersecurity, and more. AI begins inside one processor in one server, but it only achieves its goal while working among an array of processors put together across a neural network, connected at very high speed over secure links with ultra-low latency that are highly dynamic and able to evolve quickly. These are networks that think.
You will come to see in the next section why it is we believe Napatech's networking cards, co-created with market maker Intel Altera, are the optimal solution to power these new networks. Slide eight, please. Over a long 25-year evolution, standard servers with basic networking cards provided best effort access to services like websites, emails, and remote storage of files. These things were measured in speeds of I GB-10 GB , and the world was relatively static. Now, what was a 25-year evolution has given way to a near-daily revolution with the network required to deliver time-sensitive services. New artificial intelligence applications, along with cloud services and 5G mobile access, have caused these networks to change daily to support billions of users and devices in real time at blistering speeds without delay. Napatech sits in the eye of this storm of change that is happening right now.
Basic networking cards cannot fulfill the AI vision, and a new solution is needed to build modern data centers focusing on AI, cloud, and mobile services. The solution is to replace basic server network cards with an advanced programmable network interface card . This innovation has many names, ranging from smart NIC, intelligent NIC, accelerator card, offload card, data processing unit, infrastructure processing unit, and more. Despite many names, they have one thing in common. They are designed to replace basic NIC and fix the processing, networking, and scalability issues. Advanced programmable NICs bring three important improvements to data centers and servers. First, they fix the server CPU processing problem by offloading and accelerating the networking, security, storage, and other burdensome tasks from the CPU to the advanced NIC.
This significantly increases the performance of the server and those tasks, while at the same time returning the CPU resources to the core function of AI, cloud, and mobile services. Second, the advanced NIC is fully programmable, enabling the data center to reconfigure itself on a daily basis to respond to changes in technologies, new applications, new services, unforeseen problems, and new cybersecurity threats. Third, the increased networking performance improves CPU capacity, creates and increases improvement in server performance, dramatically reducing power, space, and cooling requirements, and enabling the data center to expand as needed to meet the new demands. Slide nine, please. Advanced programmable network interface cards are widely used today by the largest data center operators, known as hyperscale cloud operators. They were the first to deploy AI, cloud, and other advanced services and ran into these issues.
Lacking a merchant solution, they deployed massive software and hardware engineering resources to build their own advanced NICs and deployed more than 12 million units last year. It is important to note that Intel Altera helped to make this market and is deployed in nearly all of the largest data centers globally. Today, as large enterprises, telco operators, neoclouds, and tier two regional cloud providers accelerate the adoption of AI workloads and cloud-native infrastructure, they are facing the same networking challenges as the top hyperscale cloud operators. Unfortunately, these next-wave users must solve the same problems without having hyperscaler-level hardware and software engineering resources. This creates an urgent demand for merchant-based programmable high-performance advanced network cards to bridge the gap. As you can see in the graph, all these new players are expected to grow significantly in the coming years.
As the primary provider of an engineering solution to hyperscale operators via component sales, Intel Altera has partnered with Napatech to deliver a merchant solution for the mass market. Based on Intel Altera's proven winning hardware and software design that is widely deployed in the largest early adopter hyperscaler networks, Napatech is uniquely positioned to deliver the turnkey, production-grade, and complete solution to the IT buyers in the mass market. Slide 10, please. As demand for more AI, cloud, and mobile services continues to grow, an increasing number of servers are in demand, with more than 20 million a year being deployed by 2029. These servers require connectivity from network interface cards, and it's clear that advanced network cards are the future for data centers of all sizes.
The hyperscale cloud networks have all moved on from basic NICs, and consequently, they consume more than 12 million advanced NICs last year. They will continue to do so, and their demand will grow to be nearly $11 billion in the next five years. Equally important, as the next wave of data centers races to deploy similar AI, cloud, security, mobile, and other services, this mass market opens up for merchant providers like Napatech and Intel Altera. A simple reason is that you cannot call up a hyperscaler and buy their programmable network card, which they have developed in-house for only internal use. This greenfield growth is anticipated to approach $6 billion by 2029 and follow the same growth that fueled the hyperscaler market. The transition window is opening now for advanced NICs for merchant suppliers to the next wave data centers that make up the mass market.
They are expected to be the fastest growing segment of network interface cards over the next five years. Slide 11, please. There is no shortage of application services within the mass market data centers, driving their demand for advanced NICs. They include high-volume designs for cloud and artificial intelligence. It also includes business-critical applications like cybersecurity and financial services, and it includes advanced network infrastructure like 5G mobile infrastructure. Slide 12, please. These megatrends create the opportunity for Napatech. As shown in this inside-the-box view of a server, Napatech's advanced NICs play a crucial role in modern data center server designs. Our products provide the access from the outside world to the CPU and the GPU FPGA resources that deliver the AI, cloud, mobile, security, financial, and other critical services.
Our advanced NICs ensure these servers can deliver those services at the optimal mix of price, performance, space, and power for the data center operator. Slide 13, please. A unique aspect behind Napatech's business model is an innovative go-to-market model that lowers our operating expenses and increases the scalability of our business. Napatech and Intel Altera have partnered to deliver advanced NICs to the mass market. This partnership brings four valuable points to Napatech. The first is increased volumes. While Napatech has been building NICs for nearly 20 years, our solutions have historically targeted critical but smaller niche applications with lower growth. This resulted in a stable capture business that drove 5,000 cards approximately per year. The Intel Altera hardware and software design for hyperscalers is now also addressing the mass market.
The application and service is addressable by the enhanced design now enabling Napatech to target a new set of customers and use cases that are at a transformational scale, oftentimes requiring 1,000 units, 5,000 units, 10,000 units, 25,000 units, or even in some cases up to 50,000 plus cards per data center per year. Next, we can scale our reach beyond what we could have done on our own, limited by size, geographic, and cost. As a part of our partnership, Intel Altera does a significant portion of the sales, marketing, market creation, and business development, bringing Napatech into big opportunities who have already embraced the solution and are looking to Napatech to fulfill that demand.
Further, as a consequence of Intel Altera's desire to strategically develop the mass market based on their hyperscale success, they have granted Napatech access to the latest CPU and FPGA technologies, creating an incredible first-mover advantage ahead of other companies looking to enter this rapidly emerging market. Finally, the partnership requires frequent engineering collaboration throughout the joint product developments. We gain high-confidence technical guidance from Intel Altera's proven hardware and software solutions that significantly de-risk our product roadmap in the engineering designs. Slide 14, please. While this unique partnership with Intel Altera has been two years in the making, the products and their solutions have largely only been in development over the past years. In this next section, I would like to share with you both updates and news related to just a few examples of the success and potential we see from our new solution. Slide 15, please.
As we look at the totality of our business, two bright spots emerge. We continue to expand the total pipeline in both the number of opportunities and the total value of those opportunities. Both are far beyond any Napatech historical metrics. They continue to expand each month, and the primary in our new products and solutions created in the partnership with Altera. Also, as we mentioned earlier, we have seen an improvement and acceleration in our capture core business. This is driven by large customers burning their built-up inventories and returning to historical consumption, particularly fintech and cybersecurity. In those segments, we have seen a growth in the last three months, which is record high. The improving base business plus constantly expanding pipeline of opportunities and early success in new design wins provide great optimism towards realizing our growth ambitions. Slide 16, please.
Our work with the tier one server manufacturer continues according to plan. We have met on-time delivery for all hardware deliverables according to their and the Altera specification. The design is anchored in an innovative hardware platform that includes a 400 GB smart NIC. It includes the ability to add an optional CPU, turning it into a data processing unit. We offer the DPU in two configurations optimized for servers of different sizes. These hardware programs develop at an incredible pace, and we met each critical milepost ahead of schedule, exceeding expectations of all parties. As the hardware projects matured earlier this year, the team began the parallel work on the many key software requirements around network storage and security offload and acceleration. This will allow development of the customer solution for the use cases to begin on the hardware throughout 2025 and beyond.
We're happy to report exceptional progress so far, resulting in stunning benchmarks for each of those demonstrated today. During these successful stages, we were pleased to expand the opportunities from the initial project to include several opportunistic spanning use cases, including AI and cloud storage, data content reduction, cloud tenant services, 5G mobile infrastructure, and more. Slide 17, please. We remain excited about the potential from our design win with D-Matrix, the creator of Corsair, the world's most efficient artificial intelligence computing platform used for inferencing in data centers. Inferencing is the cornerstone of artificial intelligence. It transformed the knowledge from trained models to create predictions, decisions, and insights that are practical and beneficial in real-world applications. There are several things that underpin our excitement about this design. It is in a very hot space of artificial intelligence.
It is a strong validation of everything we are doing in our partnership with Intel Altera. It is a NIC solution for AI infrastructure in the AI data center backend network, and as such, it is an extremely high-volume deployment. It is also the same product shown on the previous slides, built in conjunction with Intel Altera and the Tier 1 server manufacturer. This particular design included a master supply agreement with five years of product availability, volume commitments, and an NAE fee for work related to the software to power the solution. We're expecting production deployment in 2026. Our current biggest design wins are all related to AI. We are actively pursuing the increasing list of companies looking for AI infrastructure solutions, and this is a great example of our investment strategy for building a winning solution and leveraging that investment across additional customers with similar requirements.
Slide 18, please. I would like to highlight a few of our recent design wins and the progress those companies are making in their markets. First, Broadcom Symantec and Napatech created a joint solution in Q1 to deliver their data loss prevention cybersecurity application to enterprise customers with exceptional performance. We are pleased to report that in a short amount of time, we have captured our first end-user win at a large global banking and finance institution valued at north of $250,000. This first customer represents potentially just one of many among the global Fortune 5,000 enterprises that could benefit from the same solution. Second, one of our software design wins with the company A5G, who provides a 5G packet core solution, is using our hardware and software to accelerate their application. They have just been awarded their first design win.
We anticipate the solution to go live on towers at the end of this year or early next year. The customer is a North America 5G provider. Next, a leading hardware manufacturer, Kontron, has created a solution for an edge server powered by Napatech's hardware and software solution to enable 5G UPF offload and acceleration. They are targeting telco operators who are looking to strengthen their 5G backbone networks. Most operators are these days running their IP core on a 4G architecture. Soon, many operators will have to upgrade to a new 5G IP core since the number of 5G subs will increase beyond the number of subscribers that can be supported by their existing 4G packet core. Finally, we have earlier disclosed the design win with Myrdl.ai and their inferencing product, Volo.
In Q2, they announced and disclosed their product launch to the general market, highlighting microsecond machine learning and inferencing with ultra-high performance and low latency powered by Napatech. These solutions powered by Napatech, among others, are great examples of work we have done in the past, laying a foundation for future growth, beginning to show signs of success. Slide 19, please. As I wrap up the business section of today's reporting, I would like to invite you to come to see Napatech at one of these great events through the remainder of 2025. If you happen to be in one of these great cities during one of these events, we would love you to stop by in person and talk to one of our experts on hand. Slide 20, please.
The last few slides of our update today provide more details on the financial results outlook for our fourth quarter and full financial year of 2024. I would like to ask our CFO, Heine Thorsgaard, to provide these updates. Please, Heine.
Thank you, Lars. Slide 21, please. Revenue in Q2 amounted to $4.8 million, up 43% compared to Q2 last year, and up 41% compared to Q1 this year. In DKK, revenue amounted to DKK 31.6 million, up 36% compared to Q2 last year. For the first half of the year, revenue in DKK was up 24% compared to last year and amounted to DKK $55.4 million. In USD, revenue in the first half of the year amounted to $8.2 million compared to $6.4 million in the first half of 2024. Gross margins in Q2 ended at 67% compared to 65% in Q2 last year. The gross margins for the first half of 2025 were 68.4% compared to 68.2% in the first half of 2024. Our staff costs and other external costs in Q2 amounted to $42.6 million compared to $43.3 million in Q2 last year.
Compared to the previous quarter, our staff costs and other external costs declined by 12.9% in Q2. For the first half of 2025, staff costs and other external costs are added to $91.5 million compared to $83.9 million last year. Staff costs transferred to capitalized development costs in Q2 amounted to $1.8 million compared to $0.9 million in Q2 last year and to $4.9 million for the first half of 2025 compared to $1.7 million in the first half of 2024. EBITDA in Q2 amounted to - $19.6 million compared to - $27.4 million in Q2 last year. EBITDA for the first half of 2025 amounted to -$ 48.7 million compared to -$ 51.7 million in the first half of 2024. Slide 22, please. Net cash flows from operating activities in Q2 amounted to -$ 18.6 million compared to -$ 41.9 million last year.
For the first half of 2025, net cash flows from operating activities amounted to -$ 44.3 million compared to -$ 47.4 million last year. Net cash used in investing activities in Q2 amounted to $ 2.5 million compared to $2.7 million in Q2 of 2024. For the half year, net cash used in investing activities amounted to$ 6.3 million compared to $3.7 million last year. Free cash flows in the first half of 2025 amounted to negative $50.6 million compared to -$ 51.0 million in the first half of 2024. Cash and cash equivalents at the end of Q2 2025 amounted to $133.4 million compared to $103.2 million at the end of Q2 2024. Slide 23, please. During 2025, we've planned and implemented several improvement initiatives aimed at reducing our working capital in the second half of 2025 and through 2026.
Napatech's main initiatives to improve working capital are focused on reducing inventory levels and optimizing the planning of future production, on negotiating better payment terms with key suppliers and current customers, and on optimizing payment terms and sourcing models for new high-volume agreements. In the coming years, we anticipate that most of our new business will stem from large volume orders rather than from production for stock. This shift will lead to an even more optimized logistics setup with a significantly lower need for working capital. Back to you, Lars.
Given the accelerated growth in our capture-based business, we are on track to reach our guiding for 2025. We expect to be within the range and are positive that unit ships will be on target or more than 5,700 cards in 2025. In addition, we expect that OpEx will be lower in the coming quarters due to cost reduction measures implemented in Q2 2025. We expect the company is fully funded and cash flow will continue to be improved in the coming quarters. Next slide, please. Based on our design win pipeline and initial wins, we gain additional confidence in our financial outlook. Increasing unit sales are driven by increasing demand across mass market customers. High gross margins are enabled by the strong software component in Napatech's products, which also promotes customer stickiness. Our OpEx is reduced through right-sizing the organization in Q2 2025 while still enabling growth.
Additionally, working capital is expected to be significantly improved due to new models taking orders before enduring expenses. Based on the work done in Q2 this year and the good start of Q3, we feel more confident that the unit numbers shipped going forward are within reach. We are ready to go to the Q&A session. Operator, please initiate.
Thank you. If you would like to ask a question, please press star one on your telephone keypad. If you would like to withdraw your question, please press star two. Once again, it's star one on your telephone keypad for any questions. As we currently have no questions on the conference call, I'll just hand over for any webcast questions.
This is Lars Boilesen, and I see three questions. Should I just go ahead and answer them?
Please go ahead.
There's a question from Lars Knudsen. The question is, Myrdl.ai have launched Volo. Have you received first orders or what is the path to orders and delivery? I think the status there is that Myrdl.ai have launched the product and they are now out basically working with clients. We have these products on stock. This is the, they have ported to our 200 GB DPU, and they have basically secured that everything works with that product. What they're using our product for is that they need to run their software on advanced hardware and they have selected our products. If they get orders, they will place an order to us and we will ship. They have just launched, so we have not received orders yet. Next question is Olav Mellingseter. The report describes D-Matrix AI inferencing design as a high-volume design with unit commitment and NAE fees.
Can you quantify the expected revenue contribution from D-Matrix in 2026 and beyond and how it factors into the 60,000 unit- 80,000 unit peak production goal? As we said in the presentation, there's no doubt that D-Matrix, they have a very interesting backend inferencing design. What we see now is that there's a lot of focus on inferencing in AI. There's a lot of big clients who are running different language models, training models, where some of these models have become very mature. In that process, it makes a lot of sense to take these more mature language models and put them on some advanced hardware in the backend. This is where they have selected Napatech for their mesh backend network. When we did the capital increase, we arranged a meeting with our three analysts from Spaarbekken, from DNB, and from ABG.
They had a meeting with D-Matrix, and I was just participating there, listening in. What D-Matrix said on that call was that for each data center they deliver to, the volume for D-Matrix will be from 10,000- 100,000. You have to divide that with four. Basically, each data center for Napatech, if D-Matrix gets such a win, will be from 2,500 units- 25,000 units. Obviously, it's very hard for us to predict. The product is not ready. We're going to deliver this at the end of the year. What we have done is basically we have cut off 80% of what they told us, and we have around 2,000 units- 5,000 units into 2026. It's very hard to predict. We'll just have to wait and see. It's looking very promising based on what D-Matrix tells us.
A third question from Øystein Løgård: Can you give an indication of revenue potential of the first 5G packet core deployment? Will the initial deployment be spread over several quarters or be mostly delivering Q4 or Q1? This is a relatively smaller operator in North America. They are basically going from having 1 million subscribers to 10 million. Such a deployment will require a three-digit number of cards from us initially in Q4 or Q1. These were the questions I have on hand now.
Thank you. Confirming we have no questions on the conference call, handing back to you once more for any closing comments.
Here's actually one more question here coming in. Olav Mellingseter. When it comes to potential bottlenecks, what are the key risks or challenges in scaling the D-Matrix partnerships, such as supply chain dependencies on Intel Altera FPGAs, the competition in-house, hyperscaler solutions, and how is Napatech mitigating them? What is really great about the D-Matrix project is that they are basically, when we delivered to the tier one server manufacturer, a 400 GB product, we were waiting for the CPU from Intel Altera. We actually delivered a version without a CPU in September last year. This is the product D-Matrix are using. Basically, what we're doing there is a bit more testing. The product is already developed in-house last year. We feel very comfortable that we will deliver on time, and we are on track, as I said in the presentation, on the first initial milestone.
Everything's going according to plan. We don't expect any bottlenecks yet. There's another question from Karl Ulvener. Do you know the volume D-Matrix is currently shipping in the second half of 2025? We do not have access to that. What we are shipping to them is basically where they today have a box with their AI chips inside. They are shipping that in volume today to hyperscalers and other clients. What they're building together with us is a network of these AI boxes, right? They're going to ship a mesh backend network for inferencing, and this is the product we are helping them to deliver based on our 400 GB DPU. There's another question. Just reading here, just a moment. In your volume guidance for 2026, can you say how much of that is expected to be driven by traditional smart NIC business?
How much is expected to come for D-Matrix, and how much of other new IPU design wins? We have a guidance for next year for 9,000 units 10,000 units, and we expect around 5,000- 6,000 from existing business, and 80% of the rest is from D-Matrix. Obviously, we hoped we will get more than that, but that's basically our guidance. We basically try to take a small percentage of what our clients tell us to put into the guidance next year. Question number seven. Oyster and Lowgard. In your first plot delivery to D-Matrix in Q4, is the success. When do you expect to receive the first commercial production order? We're very focused on delivering ahead of their expectations. We've done that so far, and we are going to have meetings in September, October on what is their forecast for 2026.
We need to get that forecast because it takes anything from 2-4 months to deliver to D-Matrix, depending on what kind of volume they're looking for. These discussions will have to take place before we deliver the final product in November this year. I think this ends the number of questions I see on my screen, operator.
Thank you. Confirming we have no questions on the conference call. As we have no further questions, I'll hand back to you once more for any closing remarks.
Yeah, so I think this was a solid quarter for Napatech. It's great to see our base business coming back. We also see that this trend has continued into Q3, and we feel good about our biggest design wins, where we're delivering now. We definitely delivered ahead of schedule on both the tier one server manufacturer, but also D-Matrix. Thank you very much.
Thank you. This concludes today's Napatech First Half 2025 Conference Call. You may now disconnect your line.