Welcome, everybody. Good afternoon. Good morning. Wherever you are around the globe, we wish you a very warm welcome. We're live from Stockholm here in Sweden. It's very sunny outside, which we much appreciate. My name is Sander Arts. I'm the host for this afternoon. I just joined Sivers as their Chief Marketing Officer, and I love value creation, which is what this afternoon is all about. I worked for a variety of companies in the semiconductor space, including Philips, NXP, Atmel, Marvell, Syngent, and SiFive. I do have a background in AI and photonics, as I worked for a few companies, including Grok, Axelera, and Fabrinet, Furiosa, and PhotonDelta. This is today, and Sivers is the center of this afternoon. Before I hand over to Vickram, I'd like to tell you why I joined, which is probably very important.
First of all, I love companies that have a lot of potential. If you go back into my background, you see a few names where a lot of value was created for shareholders. When Vickram asked me to join, I kicked the tires on the business a little bit, and I see tremendous opportunity here in Europe, but also in the United States. Sivers has managed to position itself in two very attractive markets, which we will talk about today: SATCOM and AI data centers. I have been very, very impressed with my colleagues. Before I joined, Vickram asked me to talk to the team. We got incredibly high-caliber people that are transforming this business into a product company. We are getting increasing attention. Testimony is that we are being joined by over 170 people today that are online and here in person. Thank you for coming.
I also talked to customers, and you'll hear from customers today. Before I joined, I said to Vickram, can I talk to a few customers? I did. It excited me to the extent that I decided to join. We got a great program for you this afternoon. For the next three hours, we'll take you through a cross-section of the business. You'll hear from all of my colleagues, including Vickram and Lottie, and the business unit managers. During the breaks, but also as part of the presentations, we'll have a few customer presentations. A big shout-out to Tim, who's joining us at 5:00 A.M. Tim, thank you very much. West Coast time to give us an update from his perspective from Northland Capital Markets. Much appreciate you being part of the program. I'll come back a few times during the breaks.
I'll come back during the Q&A. With this, I'd like to hand it over to Vickram.
Thanks, Sander.
Thank you. Thanks and welcome, everybody, to our Capital Markets Day. It's an exciting day for us. Again, thanks to everybody who's here in person as well as online. I'll start off with a corporate overview, talk about a few things that are relevant to understand about our business, and then I'll pass it on to Lottie to get you more details about the financials. At a glance, for those who are new to our company, we are headquartered in Sweden, Stockholm. We are listed on the Nasdaq Stockholm under the ticker symbol SIVE. We are in three geographical locations: Sweden, the United States, and Scotland. We have two main businesses. We are a total of about 122 people now. Our two main businesses are Sivers Photonics, where we build differentiated lasers and laser arrays, and Sivers Wireless, where we build RF beamforming solutions.
Why invest in Sivers as a potential investor? You have here a unique opportunity to participate in two of the hottest secular trends in our industry. If you've been following the semiconductor industry, we are participating in two of the hottest global secular trends. We are spending a lot of resources on two momentum markets. What do I mean by momentum markets? Markets with a lot of tailwinds that will allow us to deploy our technology and solutions into our customers' offerings to the marketplace. In addition to that, if you look at the market, it's a very sizable market of more than $2 billion of serviceable market by 2028 onwards. For a company our size, we have two highly differentiated technologies, one each supporting these two momentum markets that are being driven by these two secular trends.
As there is more awareness being created with our technology and solutions, we have a growing customer pipeline that we'll talk about too. You put all this together, there is a Sivers potential of tripling or quadrupling this business from where we are over the next four to five years at very healthy gross margins. If you look at this type of revenue growth in the next four to five years, and you look at how semiconductor companies are valued, that is a sizable market valuation at comparable multiples. In addition to that, from the revenue levels where we are right now, we look to growth in two waves. The wireless business is already growing very healthily, and it'll continue to do so.
As we get ready for these radical transformations in the photonics side of our business, a second wave of growth is also going to come into the company. You see a very healthy level of revenues and two waves of growth coming in. The way I want you to kind of think about this is, here's a company that's participating in the two hottest secular trends with relevant technologies, fueled by these momentum markets, and have the ability to contribute to outcomes in Europe and the U.S. Now, that's a pretty unique asset. If you want to play in these secular trends, we also firmly believe at the valuation levels we are today as a company offers a fantastic entry point for new investors as well as existing investors to add to their positions as we continue to execute on this pathway.
The easy way to capture this slide in your heads is it's the tale of the five twos. Two secular trends, two momentum markets, a $2 billion serviceable market, two highly differentiated technologies, and two waves of revenue growth. It's an easy way for you to kind of keep this in your head is the five twos. Let's move forward. The last time when I joined, we were covered by Redeye, who has been covering us for a long time as a retail analyst. I'm super happy to also say that we have now added a couple of institutional analysts that cover us as well, with Carnegie in Sweden, as well as Northland Capital Markets. Tim will talk about this later from his markets perspective.
We're happy to see that analyst coverage is also increasing in the company, which just raises overall awareness for us worldwide. Let's dig in a little bit deeper. We talked about two long-term secular trends. One of them is AI Axeleration. If you've been watching the world of NVIDIA and AI and the hyperscalers, AI is a huge phenomenon. The training models are getting larger. The learning demands are deeper. Now we are even moving from learning to inference. This secular trend, I don't see changing for a couple of decades or more. That's one where it's a long-term secular trend. The other one is millimeter wave adoption or ultra-high frequency RF adoption. That's also happening for a variety of reasons. In some markets, capacity is no longer available at the lower frequencies. In some markets, there's more need for real-time communication.
In some markets, the end solutions are getting lighter and smaller. Their payloads also need to get lighter and smaller while preserving high performance. In the land of RF, the higher in frequency you go, you can make things smaller and lighter. In defense, threats are happening at ultra-high frequencies. We need to have methods to deter these threats and mitigate these threats as well. That is two long-term secular trends. Both of them are multi-decade, in my opinion. Another good thing is those two trends are happening in a multitude of markets. However, throughout my career, I've found it easier to deploy our technologies when you have one or two momentum markets, markets with significant tailwinds. We'll talk about that. In our case, those two markets are AI data centers for our photonics business and SATCOM or satellite communications for our wireless business.
You can see there are a couple of other markets here highlighted in orange. We will talk about those because we maintain some outpost engagements there. The 80/20 rule, our focus is on AI data centers for our photonics business and SATCOM for our wireless business. Our value proposition is in both these markets, we can offer energy-efficient laser arrays for the AI data center market and RF beamforming solutions for our SATCOM market. Again, going back to the previous slide, we have the right technology for the right momentum market that is being driven by the right secular trend. We have a pretty strong executive team. We have industry veterans, highly respected business and technology leaders. Our two recent appointments continue to strengthen our team. One of them is Alex McCann, who has joined us as the MD of the Photonics BU.
The other one you met today on site here is Sander Arts, who has joined us as the Chief Marketing Officer. We have an extremely strong executive team because oftentimes it is the team that determines the outcomes. Each of these leaders have very strong, capable talent on their teams that are helping us drive this vision to a reality. We saw revenue momentum in Q4 2024, and that has carried into 2025 as well. We had a very strong Q1. There are a few things to note here. Our top line hit a record in Q1. On the right side, you can see that our product revenues in red continue to grow year on year. That is a very important metric for the company.
We want to drive the absolute level of our product revenues quarter on quarter or year on year as we build this transformation into a product company. Another thing to note about 2025, I said this in my annual report in Q4, we entered 2025 with a very strong backlog to deliver on our 2025 plans. We have continued to build our backlog strength. We are very, very well positioned to go drive outcomes for our 2025. That is something that, again, I and Lottie, we monitor is how's our backlog shaping up for every subsequent year and how do we then go and execute on it. Again, the shift to products continues to accelerate strongly. We have good strength going into the rest of 2025 for what we need to execute towards our plan.
Remember those orange things, the orange markets I had highlighted a couple of slides ago. We call these Sivers Outposts. While our focus markets or our momentum markets are AI data centers and SATCOM, it does not mean we completely ignore everything else, but we are highly selective in a couple of other markets where the potential to do something big or make something big happen exists. We are not going too broad and diluting our efforts. We are focused on very few engagements. You can count them on one hand. They are with market-leading customers who believe in the types of solutions they want to put out in those spaces that rely on Sivers technology. Now, when these things hit and they deploy, of course, we're going to bring that news to the investors. These are what I call outpost engagements, but they're very highly selective.
With that, I want to now hand over to Lottie, who will give us a lot more depth into our financial performance and the financial future. Then I'll come back to talk about markets, technology, and customers. Lottie, please come on over.
Thank you, Vickram. As CFO, I'm very pleased to be here today and provide an update on our progress in pursuing the growth opportunity that Sivers have. I want to start with the more recent progress in 2025 and then look at the trajectory over time. We started 2025 strong. We delivered yet another quarter with record sales. Revenue growth was 40% to we reached SEK 78.5 million. We are delivering in accordance with the new baseline of quarterly sales that we set out in Q4. We continue to focus on improving EBITDA. Although slightly negative, SEK 2.6 million, we're close to break even. It is a 65% improvement year over year. We do have a strong momentum in our wireless business unit, where sales increased 85% to SEK 58 million. Growth largely came from NRE projects, including the U.S. CHIPS Act.
Wireless product sales reached SEK 14 million, which is the second highest quarter to date. Photonics sales reached SEK 20 million, which was a decline of 18%. Photonics NRE projects decreased in the quarter. This was largely due to that the baseline for comparison in 2024 was the highest NRE sales quarter ever in Photonics history. Product sales in Photonics, on the other hand, reached highest to date quarter in first quarter 2025 and reached SEK 10.7 million. For both business units, we had a sequential growth from Q4 to Q1. Following a strong product sales in Q4, we continued to deliver on a high level. Product sales was close to SEK 25 million, which was the second highest to date. We delivered substantial growth of 41% as we continue to advance our transformation towards a product business.
Both business units contributed to the growth, Photonics with SEK 5 million and wireless with SEK 2 million. We are really pleased with this development of increasing product sales as it is demonstrating a focus towards a product-driven revenue model. As Vickram mentioned, in the longer term, we are looking at an 80% share of product sales. We will see layering in product ramps for future growth where wireless is preparing for sampling of broad market SATCOM chips during this year. As Vickram said in his presentation, Photonics ramp-up will come somewhat later. Total product sale in the quarter was 31%, which is in line with the fourth quarter. Driven by the increase in NRE products within wireless, wireless as a share of total sales increased to 74%. It also drove the increase in our presence in North America, which is now 64%.
To summarize Q1, we started the year strong. We grew 40% year over year. This marks the highest quarterly revenue in the company's history, reflecting strong demand in SATCOM and AI data centers. Wireless grew 85%, and we grew 41% in product sales. EBITDA improvement 65%. Our cash flow was SEK 56.1 million and operational cash flow minus SEK 22.9 million. During the quarter, we did strengthen the company's financing through an equity raise. This will, of course, support our continued successful growth. We are pleased with the progress we make, and we are trending in line with our plans. I can say we had a leadership meeting last week. The team across Sivers regions and units are really focused and eager to take the company to the next level on our growth journey.
Taking the longer perspective on our performance, our Photonics business has a stable revenue while getting in position for ramping up production. We do have an opportunity for margin expansion and increased leverage of fixed costs. During the quarter, Photonics has partnered with Wind Semiconductors to enhance production capacity. This is an important milestone that has been reached to secure a flexible business model with an asset-light production. Pivoting focus to SATCOM in the wireless business, plus selective engagement in other markets, allows strong growth and improved EBITDA. We have delivered a sharp improvement in EBITDA margin in wireless, reaching positive EBITDA for 2024 and for Q1 2025. In wireless, we're preparing operations for efficient scale-up.
Since we met last time at the Capital Markets Day, we have implemented an ERP system across both Sweden and the U.S. to allow the company to work across regions in a more efficient way. We are strengthening management processes both in general in operations, but also in terms of commercial terms in our contract engagements. We are working on improving wireless working capital. We are certain we are on the right path to higher margins as volumes pick up and customers move to mass production. A key financial theme continues to be capital-efficient, profitable growth. This includes focused development in target strategic verticals, meaning AI data centers and SATCOM. Our development is done in conjunction with our customers' priorities and shortest time to commercial ramp-up. As we said, capitalized R&D includes the development of a broad market chip for SATCOM within wireless.
In general, we can say that capitalized R&D is about 10-15% of sales. Another cornerstone in our capital-efficient, profitable growth strategy is the asset-light manufacturing. As we mentioned in Q1, Photonics have partnered with WIN Semiconductors. Wireless, already from before, worked with GlobalFoundries. We have now fabless production in both business units. This is, again, an important cornerstone for scalable growth. Sivers has been on a transformational journey the past few years. In 2022, we had strategic restructurings and pursuing growth opportunities, including the acquisition of MixComm that was done in Q1 2022. This also allowed us to expand our organizational footprint into the U.S. market. 2023 was a record-breaking year. We had 78% growth, and we improved operations in the company as a whole.
2024 continued to have sales on a strong level, and we made strategic advancements. We focused further on AI acceleration and millimeter wave adoption. We signed some major contracts and further improving operational efficiency. As we said before, 2025 has started out strong. Over the years, we have demonstrated our ability to turn increased sales to profit by improving operational efficiency. We have strengthened our financials to support growth. As I mentioned, we completed an equity raise in Q1 of SEK 108 million. Additionally, we added Boardman Bay as a U.S. investor, a deep tech investor that is here for the long term and takes a long-term perspective of Sivers as we do. In Q2, we also refinanced our debt successfully to better terms than we had before. We continue to improve our financial KPIs.
Although 2024, to some extent, was a year of consolidation and preparing for the next step, 2025 has started really strong. We have talked about a strong order backlog that will support us for the next couple of years, our strategic focus in growth markets, and our operational goals to improve efficiency and expand margins. When it comes to cash flow, we're also focusing on reducing working capital. This is negotiating better terms in our customer contracts, both new and existing, and introducing more frequent milestones for billing. We want to better balance the customer payments and supplier payments in general. An example of this is the recent contracts that we have signed with material upfront payments. To summarize, over the past few years, Sivers has demonstrated a strong trajectory towards growth and operational efficiency.
We've had a CAGR of 40% for the past three years, and we had 40% growth in Q1. We have a successful shift towards a product-driven business model. This consistent improvement in margins that we chase reflects a successful execution of business model transformation. In fact, EBITDA was positive second half of 2023 and second half of 2024, and we were close to break even in Q1 2025. With our financial strategy, improving capital efficiency by focused R&D and fabless production, and strengthening our financial position with the equity raised and refinancing of our debt, we are well positioned for further capital-efficient, profitable growth. Thank you.
Thanks, Lottie. Tim, we have you on tap. Tim Savageaux is a Senior Research Analyst from Northland Capital Markets who initiated coverage on us recently. We have invited Tim to provide a markets perspective.
Tim covers both Photonics, AI, as well as wireless SATCOM. Let's hand this over to Tim. Tim, welcome.
Thanks a lot, Vickram. I appreciate it. Good morning, everybody. The sun is actually just starting to come up here in California. I want to talk a little bit about my background in Northland and then why I thought Sivers in particular was a really nice addition to our coverage universe where we initiated coverage in late March of this year. As Vickram said, given the considerable overlap on both the optical side and the wireless and SATCOM side of my current coverage universe. Just a little bit about Northland. I'm in California, but Northland's based in Minneapolis. We've been around for going on 20 years. We've got about a little over a dozen research analysts working across principally technology, but also elements of consumer financial technologies, what have you. A real heavy focus on tech and a considerable investment banking effort as well.
We were actually bought by First National Bank of Omaha a couple of years ago, a very large private bank with about $30 billion in assets. Some good financial stability there. I have been with Northland for nearly 10 years, based both in San Francisco and down in Southern California where I am now. That is part of a 25-year-plus career focused on elements of the communications technology, photonics, and optical markets that stretches back way back to the mid to late 1990s in San Francisco with firms like Robertson Stevens, in New York with JPMorgan prior to that. As I mentioned, in recently picking up coverage, and hopefully I will demonstrate here over the next few minutes, I did feel like Sivers was a very natural addition to our coverage universe. Can I get the next slide, please? Great.
This has been more recently kind of a two-year journey in terms of investor focus on optical connectivity in the data centers driven by AI infrastructure deployments. We cover a lot of territory when we focused on that, right? Vickram mentioned NVIDIA, but in sort of the optical neck of the woods, it was really Marvell who actually bought a company I covered for years called Inphi. Really, two years ago, almost today, actually, they reported results and really got the investors interested in what was happening in AI land, kind of quantified their AI revenue, talked about it expecting to double this year, then double again. That number started at $200 million annual run rate. I think it is over $1.5 billion now and expected to continue doubling.
Those are mostly chips going into 800 gig and above optical modules and also some pluggables to interconnect data centers. Again, coming more from the Inphi side, that's really driven that growth. More recently, they've had a lot of growth in compute Axelerators, which is less direct overlap. That's really kind of where it all started, as you can see. Other aspects of my group, the Lumentums, Coherent, Fabrinets, and Cienas of the world, started to move up that very day. I've got a picture here of our report from early June, kind of heralding that initial focus on AI optical. Next slide, please.
What has happened in the meantime is what we estimate to be $30 billion in equity value creation driven by this AI optical opportunity initially inside the data center and then more recently moving outside the data center and even starting to positively impact the traditional service provider market as they add bandwidth to deal with all this. Here you have a range of my coverage companies. What we have seen, I also include kind of a negative factor because while this is happening, certainly over the last 18 months, and this has improved recently, you have had a lot of pressure on the traditional telecom side of the house. If anything, service provider was a headwind, and the strength seen in these stocks over that time period was despite that.
I try to factor that in with some kind of add-on, if you will, to the AI upside and come up with that $30 billion. That's a pretty healthy multiple on incremental revenue. Depending on where you look, some of the more directly exposed IC companies focused on AI optical are trading well into double digits from a revenue multiple perspective. This is my attempt to look at the incremental AI revenue and where that's getting valued, and it's all quite healthy. As we'll see later, names that I consider very comparable here, though much larger, like MACOM, are trading up around 10 times revenue.
Now, what's been the case is the real focus of all this value creation has been on pluggable modules and the components that go inside them by and large, as well as some of the modules that interconnect data centers across distance, although we think that's starting to change. Next slide, please. The catalyst for that change, which has been the case for a couple of years running now, right? I think two years ago, NVIDIA's CEO was talking about how much he liked copper, much to the chagrin of a lot of my companies and investors. Last year, it was this year, sorry, it was co-packaged optics and really onboard optics and the switch.
I think you saw this at the OFC show as well, where we're starting to see a shift in investor focus towards Silicon Photonics and what's happening with some high-profile startups and onboard optics like Ayar Labs, which is part of the Sivers growth story as well. I think that continues to change as we continue to wrestle with what the implications for pluggable optics are of some of what's going on right now. I'm sure we'll be talking a lot about that tomorrow. Actually, Coherent is holding an analyst day in New York, which is one of the larger pluggable transceiver suppliers in the world. I think these subjects will be continued to be focused on.
My own view is there's some real opportunities emerging that actually look kind of familiar based on what I've seen in the past in the Silicon Photonics and optical I/O arena for light sources in particular that look pretty interesting. Next slide, please. When I say we may have seen this before in some degree, it looks a little bit like the type of, and these were for VCSEL arrays, unit volumes we saw with the introduction of Face ID on Apple iPhones going back almost 10 years now. There's a couple of interesting things about this, right? Thematically, we're talking about what are typically historically pretty limited unit volumes. For telecom lasers in particular, high-speed long distance, we're talking about hundreds of thousands of units, not tens or hundreds of millions. That's certainly what we saw in 3D sensing.
Same thing for VCSELs, which are typically used in short-reach data comm applications, both pre-AI and with AI. As I note here, Lumentum generated $2 billion in revenue, high-margin revenue from this opportunity in this five-year period after the rollout of Face ID. Interestingly, the stock market pretty much discounted this all in advance, right? As the stock market has wanted to do. From mid-2016, when I think people started to become aware that something was happening here, up until the launch of the first iPhone with Face ID, we saw $25 billion in equity market value generated over that time period. That's, like I say, proper sort of discounting mechanism.
Also, I think it speaks to the type of opportunity for laser arrays that we could be looking at as some of these optical opportunities start to really ramp up and roll out in the 2027 timeframe. Again, expect that to be reflected in stocks prior to that. At least, that's what we expect. That is why we decided to get involved with Sivers when we did, or at least part of the reason. Next slide, please. Interestingly enough, I just initiated coverage of an Israeli SATCOM IC company called Satisfy late in 2024. Looking at the Sivers portfolio was really kind of, like I say, a perfect fit on both sides of the house. Subsequent to that, and that was, I think, below $1 for Satisfy where we picked that up.
Their largest customer, MDA Space based in Canada, bought the company just recently here for $270 million. As I mentioned, about five-six times our 2026 revenue estimate at the time. I think that's a solid data point for value creation in the space, as well as MACOM, which is kind of a coverage company of mine for a long time. I've been talking a lot more about SATCOM growth, putting a pretty big contract for LEO broadband. MACOM also shares Ayar Labs as a LightSource customer. Again, despite the much larger size, I think that'll end up being a pretty solid comparable in terms of looking at how Sivers evolves over time. With that, I don't know whether we're heading into a break or where I'm handing it, but certainly my pleasure to be here this morning.
Thanks, Vickram and the team, for inviting me.
Hey, thanks a lot, Tim. Thanks for providing us an outside perspective. Hope that was useful for the audience both here and online. Right now, I believe we get into a break. Is that right? Yeah. We will have a 10-minute break, stretch your legs, get some coffee, whatever. Then we will go into my section on markets, technology, and customers. Thank you. Breaks. There will be a couple of interesting videos playing. I am sure we all will learn something from it. Space right now is a turning point. That whole concept where you saw Elon Musk's visionary look at how we could vertically integrate with his investment in SpaceX and the entire particular sector and create a LEO network that can occupy and therefore deliver to the world has been fascinating.
Hi, my name's Paul McCarter.
I'm the Chief Executive of ALL.SPACE. ALL.SPACE is a company that develops a ubiquitous technology for terminal connections to any orbit of satellite in space, but does so so it can connect to more than one satellite simultaneously with a full connection. Dual-use technology is the ability to use either military technology that's evolved from the military in a commercial market or a commercial technology that gets used in a defense market. Clearly, from the name and definition of that, it's very interesting to see how technologies can change either the cost of being able to use something or the infrastructure that's required and the operating mode that it will get used in. It's fascinating to see since the beginning of about 2000 when the internet really took a hold of making global markets accessible for different investments.
Previous to 2000, you saw a lot of defense technology where it was more advanced in some cases than the global markets in the commercial world being transferred and then being used in different places. Of late, because of this huge opportunity in the business-to-consumer marketplace, you're seeing investment creating moves in technology where commercial capability is seen and then adapted and used for military use. One of the challenges with that, however, is a lot of the capability you find in the commercial market isn't built to some of the stringent requirements and specifications you need if you're going to use something in a military market. If you look at things that need to have the ability to kind of vibration, extreme heat and cold temperatures, or water ingress, things that get used in the defense market need to be a lot more ruggedized.
When you have war fighters using them on any location on the planet, they have to be able to handle that environment and be used. There are advantages to using commercial capabilities in that it might be cheaper and it might be more globally available. When you're using defense products, they have to be able to deal with what the war fighter is going to throw at them and still keep working. That is an interesting one for ALL.SPACE because the reason we exist is to serve that particular need of serving the huge proliferate and investment in space from a space asset perspective, from a ubiquitous terminal using a particular piece of technology.
Now, the patents that we have that sit across the transformational optics layer of our phased array capability enable us to do something that nobody else can currently do in the marketplace, which is connect to two satellites in any orbit simultaneously for full bandwidth connection. What that means for us is it's actually applicable to all markets. It could be used in aerospace. It could be used in transportation. It could be used in the business-to-consumer market. The pressing need and the urgent requirement that we've got from a defense market perspective means that we're aiming for that first, building that as our revenue beachhead, and then we'll reinvest into diversifying into the other product areas.
Yes, it's a dual-use capability, and we're choosing how we can therefore grow the company in the fastest way to ensure we have the money to invest into the different market sectors. Yes, it's vitally important for us to be adaptable. If you look, for instance, and we're talking about the U.S. Department of Defense, are 50% roughly of the global C5 ISTAR market. Therefore, the U.S. DOD is often a market leader when you look at electronics communications products because they have more money to invest in any other single country. When you look at their strategy and their roadmap, which looks at how their satellite communications equipment needs to be used, they recognize a number of things.
The three key things they point out are resilience, i.e., you have to be able to connect to any orbit, GEO, MEO, or LEO; diversity, the ability to connect to any one of those orbits, more than one full simultaneous link at a time so that if one link is taken away, be it in space or on the ground, you're still connected to another communication link; or modularity, which is the third one, which enables them to, in the field, with a small communications platform, either on a land, sea, or air terminal capability, be able to swap in and out things like modems where you don't need to change the entire platform, but just the modem so you can connect to any operator's network, any operator's asset, and use their modulation scheme so you can create seamless communication capability. It's vitally important for us.
We've created a communication platform that isn't just an antenna. It's got four modular bays that enables you to pull out one modem, plug another modem in, and work simultaneously with any orbit or any network operator. It also has the ability to use artificial intelligence based on an edge compute module, which we can plug in and create any software stack and any software that the customer might want to use to interact not only with the environment it's in, but also sensing and looking at different uses of the data from that communications platform as it's connected to all different types of networks, not only satellites, LTE, 5G, Wi-Fi, or any other communication protocol that you wish to put upon it. We can interface and deal with all of those communication networks from the total system platform.
Dual-use technology, clearly as we advance and we need to not only move from the defense space, but also going into aerospace and then going into any other market, what we need to do is to invest and change the technology. Whereas defense is absolutely vital that you have the ability to meet the stringent regulations for war fighters, in the commercial market, they're much more price-sensitive.
Depending if you're going to do the business-to-business market or the business-to-consumer, what we need to do is to change the focus of how the technology is implemented from stringent regulation to more price-sensitive, and therefore you can create capability that will be able to put it in the hands of either the consumer or a business that wants to use it in their value chain and make sure it's efficient from a capital perspective for them to be able to use that. It might well be, as you look at lease models, a through-life cost model. That dual-use capability therefore has a big impact on how you invest your capital and how you put it in the hands of a user that want and can afford to buy it.
Actually, I mean, as a young child from the age of 12, I built a communications antenna and radio in the shed in my backyard. I was so fascinated by radio communications that I made it my choice of career. I joined the army when I was 16, and they paid me through college to do electronic engineering and then through university to do software engineering. I became a chartered engineer. Communications has been my life and career ever since I was 12 years old. Be it satellite communications, land-based terrestrial networks, fiber optics, I have been operating in that space from a research and development capability, but also taking new and novel technologies to the world for the last 30 years. It was not a case of did I choose it.
I think it was more a case of the fact that I was so fascinated by it, it became the thing I always wanted to do. Space right now is a turning point. That whole concept where you saw Elon Musk's visionary look at how we could vertically integrate with his investment in SpaceX and the entire particular sector and create a LEO network that can occupy and therefore deliver to the world has been fascinating. It has disrupted the market, the space technology market. Now you have Amazon Kuiper coming in 2025 to 2026, and then Telesat Lightspeed coming shortly after. That concept of having all of those LEO networks means that the entire value chain from GEO, MEO, and LEO is going to change.
It also brings very closely into focus what ALL.SPACE is trying to do in that what people really want, be it in a business-to-consumer market or a government market, is a ubiquitous terminal that can talk to all of the assets in space without having to change anything or just use software data architectures and update that. You start to move into where we are in the world of mobile communications, where it becomes more about the data and what you can do with that data and the intelligence that sits upon it. That is a fascinating area that is going to expand and change massively over the next 10 years.
All right, welcome back. I hope you had a quick break. You talked a little bit out here and online you got your bio breaks, et cetera. Now we are back.
Now I want to start talking about markets, technology, and customers. I'm trying to keep it at the level where a vast majority of the audience can get the key messages. That's the intent as we step through the next set of slides. Just to recap, we're going after two long-term secular trends, and there are two momentum markets, AI data centers and SATCOM. Let's dig into each of these markets. Why AI data centers? What you see here is the top hyperscalers, Microsoft, Meta, Google, Amazon, have all committed to significantly increasing their CapEx for AI data centers in 2025 compared to 2024. These are billions of dollars you see here. They spent $225 billion last year, and they're going to be spending $315 billion in 2025. What are they seeing?
What they are seeing is this tremendous explosion of AI learning models. The use cases are starting to get articulated. They want to be ahead of the curve and get ready. They see this tsunami coming. In addition to these hyperscalers, there are also nation-states that want to have their own AI gigafactories. I'm sure many of you have read about Europe wanting to catch up to the AI race. They want the European AI gigafactories as well. There are many other nations that also, from a security perspective, want to have their own AI data center footprint. This thing is steamrolling and gathering momentum as we go. Okay, that's number one. It's a fantastic market with tailwinds from deep pockets. This is not companies that do not have the money to spend to build this infrastructure.
This is real money going into this market that offers a tremendous tailwind. As they do this, now what I show you is a picture of a data center. If you actually look closely, this is the backside of the racks, and you see a mangle of wires. There is a lot of wires in there if you kind of look through. What is happening is as these hyperscalers are starting to put down more and more of these data centers, there is still a heavy mix of copper and optical inside these data centers. Actually, there is a lot more copper inside the racks. When the racks have to talk to each other, they have to go up to the switches and talk through optical. A little graphic on the left pops up. At the very bottom are those server racks that you see in the picture.
They all each have computing chips that are doing a lot of computing, but they really can't talk to each other across racks. If you need to kind of rely on connecting them, you got to go up to the top of the rack switches. Those switches are all talking using optical between them. The optical fabric is limited to the switches, whether they are top of the rack switches or aggregation switches. Everything today inside the server racks is mostly copper. All these hyperscalers and chipset makers have come to the realization that that's not good enough to where they want to go. They want to enable these server racks to talk to each other faster while dissipating less power. They want more optical interconnects because light doesn't heat up like copper does.
Copper, when you run electrons through copper, copper heats up, and electrons can't travel as fast as light. To make these massive architectures work, we need more interconnections between racks that move towards optical. If you do that, two things happen. Number one, it's 20 times faster learning. All these models can run faster because you're effectively using all the compute resources inside a data center better than the old architecture. Number two, because it's optical and not copper, it's 10 times lower energy per bit that's transferred. You got a power savings benefit, and you got a faster learning benefit. These are big ticket items for the hyperscalers. As those hyperscalers want more and more optical interconnects, so do the NVIDIAs and the AMDs of the world because they're producing these really fast compute chipsets.
For 60%-70% of the time, these chipsets are sitting idle because the interconnects are not fast enough. They're also demanding optical interconnects, which is resulting in photonics vendors, like we heard from Tim, the Ayar Labs of the world, who are an existing partner customer of ours, and others who are their peers, like Celestial and Lightmatter. They're all investing in photonics chipsets that convert electrical signals from these computing chipsets into optical so that across the data center, most of the transmissions become optical instead of electrical. To do that, they need a light source. They need a laser. Okay? That's where Sivers Semiconductors comes in. We're not alone here. We have other competition like Lumentum and MACOM, but we have strong technical advantages we'll talk about on what we are leading with our technology advantage.
Again, to sum up, billions and billions of dollars on the next generation AI data centers. Everybody needs more optical interconnects. To do that, we need the likes of photonics vendors working in conjunction with laser vendors to make this happen, right? This is going to happen to connect all these compute resources that are on the server racks, but also in top of the rack switches where the architecture of the pluggable optics is also going to change. All this boiled down for investors, what does that mean? No matter how much photonics and optics come to the play, the lasers are going to stay external. We supply lasers. Even the pluggables, the way they'll adapt their architecture will result in just more laser and light sources inside a data center.
That means the TAM increases, which is a good thing for us because we are in the business of supplying lasers and laser arrays. Even at a very reasonable attach rate or a conversion rate of copper to optical, this is easily a billion-dollar market by 2028 plus for laser arrays. Okay? And that's our target. Now, there are two terminologies you should keep in mind: scale up and scale out. Scale up is very simple, connecting all the compute chipsets to each other directly, many to many. A GPU in one rack talks to a GPU four racks across. That is called scale up. You are enabling any GPU to talk to any other GPU inside a data center. Scale out is the switches have the pluggable optics, and there they are trying to suck in the optics into the switches and leaving outside only the light sources.
That's called scale out. Both these things are going to happen in AI data centers: scale up and scale out. I'll give you a couple of examples. This is becoming very real. At the OFC show in San Francisco in 2025, Ayar Labs unveiled their first solution for scale up, up to 8 terabits per second. And that's using Sivers' technology and Ayar Labs' technology to show that GPUs can connect to each other directly and not have to wait to communicate through the switch matrix. That's a huge bottleneck reliever, number one. Number two, NVIDIA has a big show every year. They call it their GTC show. Their CEO, Jensen, basically came and said, "Hey, listen, if you need me to build the AI gigafactories of the future, I need to suck in optics into my switches." I call it co-packaged optics.
The market has to go there, which means I'm going to suck in optics into the switches, and I need external laser light sources. That is him wanting to enable scale out. Both scale up and scale out, very meaningful benchmarks of progress have happened early 2025. As this vision plays out, the way investors have to look at a company like Sivers is when you look at racks inside an AI data center, the server racks, there are going to be dedicated racks. I show you an example here that's just full of laser light sources. They call them laser array trays. You are going to start seeing more and more laser array trays inside the servers, and each tray will house a huge number of light sources. If you look inside, that is the potential for Sivers laser arrays to be on those laser trays.
Once this becomes standard form factor, we are talking about the volumes that Tim was talking about. That is the opportunity here with the Sivers laser arrays. All that is great, but why Sivers lasers? I showed Lumentum, MACOM. I mean, why should people not just use them, right? That is where each company has its advantages and the way it approaches the market. With Sivers, it is the performance, number one, which is we have the highest precision, densest wavelength of laser grids we can provide in our arrays. That is very important for these architectures. We can make these arrays in different sizes: 8-element arrays, 16-element arrays, 32-element arrays. Depending on the customer, they want the power output from these lasers at different levels too. Our technology roadmap shows the highest precision, the largest array formats, and multiple power levels.
That is our technology leadership that we are continuing to maintain. The second biggest thing is what I call light source economics. If you remember what Tim said is, as these scale up and scale out architectures happen, most of the co-packaged optics get sucked into the GPU or the switch, leaving only the laser sources, but in much higher volumes. If you look at these light sources, one of the biggest costs is manually aligning every laser into the light source to make sure very little light gets lost by the time it comes out the other end into the fiber optic cable. Let's say you have an 8-element array. Today, with my competition, you have to put down one laser at a time. You have to make sure it couples carefully into the fiber optic cable at the other end.
You have to do that eight times. If you have a 16-element array, you got to do that 16 times, 32, 32 times. The assembly economics becomes ridiculous if you do not change the way you build the lasers. If you do them as arrays, one piece of semiconductor which already has all the eight elements or 16 or 32, now you can only align the last elements on both sides. Everything in between is auto aligned. It is a huge assembly economic saving. We completely disrupt the light source economics. That is a huge selling point in addition to the technology advantage. One of the things with a smaller company like Sivers, with our customers and their customers, is, will you be there when I need millions of units from you? We have addressed that.
Production at scale, as Lottie mentioned before, we have our WIN Semiconductors partnership. We also have our O-Net partnership. O-Net is in the business of building those light modules. Directly from the lion's mouth, we know that by producing laser arrays, we completely change their economics of building the light sources. We have reliable production at scale partners to complement our performance advantage as well as the light source economics. That is the exciting part as our teams continue to execute on that. You will hear from the Ayar Labs CEO later in one of the videos on how he sees this partnership too. It has been a long journey. This is the thing that you have to realize: it takes a journey. 2022, 2023 was technology development. 2024, we showed the product application fit.
25, 26, the focus is on getting the product qualified and manufacturing ready. Everybody's asking for readiness going into 27 on for volume production. Now, is it exactly 27? Does it move around? You can never exactly say the time, but everybody wants to be ready by this time frame. That's where the focus in the business is. That's about the AI data center market. Hopefully, it clarifies what's happening in the markets and how we have the right technology to intersect this opportunity. If you then move to the other momentum market, which is SATCOM, again, it's got tremendous market momentum, deep pockets from customers like SpaceX and Amazon and others. This is real money, again, going into modernizing space. That's a huge tailwind for the market.
If you've been following the SATCOM industry, more and more what they call small satellites are going up in the sky. In the past, it used to be these huge, but very few satellites called geostationary satellites. Now the world is all mostly LEO satellites, low Earth orbit satellites. That unit volumes are going up. There's no point in putting satellites up in the sky if they're not going to talk to many things on the ground, in the air, in water, right? On water. Terminals is a huge growing market. Our focus is on defense terminals and commercial and enterprise terminals where there is the ability to demand value-based pricing and get better margins. That's also where innovation comes very strongly as well. That's our current focus.
The other thing to know about LEO satellites and these terminals, they're all starting to get smaller and lighter. Once again, this is the drive for millimeter wave technology because as you go higher in frequency, as I told everybody, in RF land, things get smaller and lighter as long as you can deliver the performance they need. Okay? What are the key drivers? Even five, ten years ago, satellite services were mainly for video on demand. You had a TV broadcast coming into the house through satellites. Now it's all about ubiquitous broadband. Everybody wants broadband coverage everywhere. Even cellular networks are seeing the amount of money that's going into satellite networks that they want to use satellite networks even to enhance cellular coverage. Okay? Two big elements. One is ubiquitous broadband, and the other is cellular leverage.
Also, if you've seen some of these older terminals, there used to be these big dish-type terminals that were mechanically steered to catch a satellite. The whole market's moving to electronically steered flat panel arrays. That's where beamforming technology becomes relevant because in the electronic domain, you can move the beam rather than in the mechanical domain. That means lighter weight solutions, more efficient solutions. The U.S. defense has basically said the space infrastructure for U.S. defense is outdated, and they have to modernize space. We're talking about tens, if not hundreds, of billions of dollars going into modernizing U.S. space. You can see now in Europe, the same thing is happening with AyarIS2. There is an acknowledgment that space has to be modernized in Europe too. Again, there's a huge tailwind here to modernize space.
In the past, the defense agencies were happy with eight, ten-year-long projects, lots of money spent, and then you get technology. They're seeing that things are happening on the commercial side much faster. The defense teams also want faster solutions. Now the time to market, time to ramps are Axelerating on the defense side of the business too. Again, another tailwind to the momentum in SATCOM. That by itself, just the SATCOM piece with what Sivers can address today, can easily be a $500,000 million market or more by 2028, 2029. Let's now talk about momentum builders for Sivers here before we talk about why Sivers. Pretty soon in the sky, there are going to be low Earth orbit satellites, middle Earth orbit satellites, geostationary satellites. There are going to be terminals all around.
If there are satellites up in the sky, customers want to find a way to use all of them and mix and match. The market wants multi-orbit, which means a terminal on the ground can talk to a low Earth orbit satellite if needed. Or if it cannot find it, it should latch onto a middle Earth satellite or a geostationary satellite, right? It is multi-orbit. They want multi-band. There are many frequencies in millimeter wave that have been historically used, Ka-band, Ku-band, et cetera. They want solutions that can use any band. Because again, if I want a link and that satellite can only take a Ku-band and my terminal does not have Ku, now I am kind of stuck. They want solutions that are multi-band as well.
The other thing is they want multi-beam, which is if I'm connected to this satellite and I'm going to lose it, I need to connect to this satellite, they want me to make this connection before they break this connection. They want multi-beam capability. This is becoming the norm now, multi-orbit, multi-band, multi-beam. That plays very well into our roadmap because we have multi-beam, multi-band, and the capability to address multi-orbit solutions. You'll hear more from our top customer in space, ALL.SPACE. They're making very strong deployment progress. They're already in mass production with us, but they are winning a lot of programs with the U.S. Navy and the Army, and their pipeline is Axelerating. That is tremendous tailwinds for our revenue builds in the coming years as well as they win more programs. That is again a momentum builder for us.
The third one is probably super key, which is there is a consortium called DIFI that's come up, which says, "Hey, look, in the olden days, we used to make these terminals on the ground or on ships, et cetera, that had all the antennas, RF, digital, everything in one box. Now that we have cloud resources, let's move all the processing into the cloud and make the terminals mostly about antennas and RF, and then take the signal digitally and go into the cloud." What that means is the terminal has now basically become a high-performance RF terminal, which has the antenna arrays and the RF and a digitizer. Now Sivers has an opportunity to dominate this box because we have all three technologies. We have the ability to do antenna arrays, which we already announced some products at MWC. We do our RF beamformers.
With our customer Intelsat, now we are doing digitizers. This is value migration in the terminals for Sivers with the technologies we can bring to these RF-heavy terminals now. Everybody's aware of the geopolitics here where the U.S. wants its own self-sufficiency in space infrastructure. The same thing Europe wants, right? These are all things that are pushing more activities in SATCOM. What does this mean for Sivers? As I mentioned, because the terminals are becoming RF-centric and the need for multi-orbit, multi-beam, multi-band electronically steered arrays, our RF beamformer value proposition grows stronger every day. In addition to that, now we can expand our serviceable market because we do arrays, antenna arrays. We do beamformers. We also do digitizers. We can be one of the highest electronic content inside a ground terminal.
That is something that we are seeing because our opportunity pipeline and the customer interest is also expanding strongly ever since we made the Intelsat announcement as well. Why Sivers wireless? It is the RF performance and expertise we have. Our RF beamformers are three to five times more energy efficient than anybody else out there. Think of these mobile units. They do not have infinite battery packs. Think of these small satellites up in the sky. They can only carry so much load, and they have only so much battery capacity. If I can make my beamformers more efficient, those satellites can stay functional longer. The ground terminals do not run out of battery too fast. It is an important value proposition for us. Additionally, all those capabilities we want, multi-band, multi-beam, multi-orbit, we have and are delivering on our roadmap.
Now that the terminal has become an RF-centric design, most customers are telling us, "Why don't you also build the antenna arrays along with the RF beamformers? You have the capability." Many of our competition do not have that capability to build the antenna arrays as well. Okay? As you can see, that is the RF performance side with our partnership with Intelsat. Now we bring digitizers. We have added digital capabilities. Just like the photonics laser side, on this side, we have GlobalFoundries as our foundry partner for production at scale that is reliable. On both sides of our business, we now have a very successful model, which is high technology value proposition with strong production partners that is hitting the value proposition for our markets. Okay? Again, we talked about this. Both these GlobalFoundries win are top of their game.
WIN Semiconductors is one of the world's largest gallium arsenide foundries. They supply the Apples, the Broadcoms, the Qualcomms of the world. You have GlobalFoundries that supplies to pretty much all the blue chip customers as well. As I mentioned before, we have a partnership with O-Net Technologies who help us take our laser arrays and help them make very economical at scale light sources. If you remember, we said the assembly costs were the highest for light sources. The other thing to note about the supply chain is GlobalFoundries or WIN, they have the deep pockets to address any need for local EU production. If geopolitics then dictate the need for local foundries in different geographies, GlobalFoundries already has production sites in Dresden. WIN Semiconductors is similar to TSMC, et cetera.
They all have plans to leverage sites in Europe if geopolitics takes us there. We have sound partners who have plans to address geopolitics as well as it needs to be if it comes through. The last thing I will say is also important because we have existing investors on the call. We have potential new investors. A big change that we have brought to mindset is pipeline discipline, customer pipeline discipline, opportunity pipeline discipline now. We are at a place where our technology is sufficiently mature. Our roadmap value proposition is sticking with our customers. We want to be careful in the types of customers we take on. We want to take on customers who have market leadership or technology leadership, who have the funding to take us through from development to production.
We no longer need to be chasing customers just because we can get development engineering money, but we do not know where they will go in the future for production. We can be a lot more disciplined, and that is what we need at this stage in our company, right? Because we have only finite resources. We want to make sure the customers we pick are the ones that will lead us to production. We are putting a lot more scrutiny on the types of customers and opportunities we take. I was at these three conferences that you see at the top. Those are our top conferences and trade shows for our markets and our technologies. OFC is an optical show. This year, it was in San Francisco. You all know Mobile World Congress, MWC, that is held in Barcelona, and the SAT Show, which is the biggest satellite show in Washington, DC.
All these three events happen in Q1 of a given year. I asked my teams, "What was different this year than last year?" Last year was before I had joined the company. They said, "Hey, you know, once we got ratification that our solutions actually fit the bill, a lot of the bigger customers' potentials are coming and seeking us out at this year's show, whereas at last year's show, we had to kind of go hunting for them." They said, "There is definitely a change as we brought more technology out to the market that the bigger branded names are starting to seek us out, which I think is a very, very positive move because those are precisely the types of customers now we want to attract beyond our early adopters." Okay?
To give you an idea, right now in the prospect stage, which is there's a potential to engage with them, they kind of came to our booths, they talked. Wireless has about 40 targets. Photonics has about 20 targets. Out of those, wireless, 30% are strong brand names. Photonics is a more focused market, so like 75% of them are strong brands. Okay? If you look at then what has progressed from prospects into technical engagements where we are actually having conversations with these customers, potentials, is there a program, is there a program where our technology can intersect, et cetera. Wireless has about 30 engagements. Again, about 30% are strong brands. Photonics has seven engagements, but 70% of them are strong brands, right?
Of course, we are building out our major customers, some of which have existed in the past, but some of which we are also bringing into the fold. Many of them are starting to move into either production or prototype before production, right? That is kind of where we are being very conscious about the customer pipeline we are building rather than being very haphazard about it or desperate about it. This is an important point I want to leave you guys with before we go into the next section because we're at a place where we want to make sure we are here for the long run as these circular trends play out, which means we're going to make sure we are very meticulous about the customers we take. We're not going to be desperate, but we're going to be methodical.
As we navigate this, because with any of these markets, you can't exactly determine when something happens. Things can move. As we prepare for it, we are very fiscally disciplined, financially disciplined, and we are going to tailor our equity and debt financing with long-term strategic partners, long-term fundamental investors, long-term debt financing agents. We want the time horizon for us to show our success, to connect and be kind of similar to our investors and our debt financiers as well. That then allows for meaningful conversation as we navigate through, you know, moves in the market as we adapt and we still deliver solutions. Okay?
That is something for our investors to hear directly from me that that is kind of how we are going to manage the path forward because we want tie-ups with long-term-minded investors, debt financing, and our own customers that are going to be around as these things become real. Lottie touched upon this, but I will say that historically, Sivers has mainly been doing custom work for customers, a custom product, a custom development. We have come to a place where we have gathered enough IP and knowledge about what the market wants rather than customer A or customer B. We are looking at customer A, B, C, D, E, F and saying, "If I build a product like this, A, B, C, and D customers will buy it." I think that is a good business case. Let us make that standard configuration, right?
There is also a mindset shift now to put out more standard product configurations because rather than go one product to one customer, we want to see if a product can actually appeal to multiple customers. It gives us customer diversity. It gives us aggregated volumes and makes for better business cases. That is also a big change in our go-to-market mentality. As Lottie mentioned this year, we will be putting out our broad market SATCOM chipset samples late Q3 from the wireless team. That is something that when I showed you that pipeline, a lot of our prospects and technical engagements are looking for. They are like, "Can I get my hands on your broad market SATCOM chips?" Now we are also starting to figure that out on the photonics side.
Beyond our lead engagements, what are standard configurations we can offer the market as scale-up and scale-out happens, and they all need these laser arrays and light sources. With that, that's the conclusion of my section on markets, technology, and customers. Hopefully, we've taken you one level deeper, but I've hopefully kept it simple enough that, you know, it resonates. Again, I'm here and the team's here for questions later. We have a video to play now. It's after Harish. It's after Harish, right? I'm going to now move this to Harish. Harish is on the screen. Harish is going to take us deeper into the business and technology of our wireless side of the company. Harish, please go ahead.
Thank you, Vickram. Good afternoon to everybody. Thank you for tuning in to our Capital Markets Day.
I'm happy to talk about the progress that is being made within the wireless business unit. My name is Harish Krishnaswamy, and I'm Managing Director of Wireless within Sivers. If we go to the next slide. As Vickram mentioned, the primary target market for wireless is satellite communications. We are finding that the satellite communications market is growing extremely rapidly because of several tailwinds. The first and foremost tailwind is the need for connectivity everywhere. This chart is showing several of the applications that satellite communications is addressing today. Of course, we have the consumer internet application that you see on the top left. In addition to that, we see several other scenarios that are demanding high levels of connectivity today. These include the aviation market. This is the ability to access internet while on flights, whether they are commercial or otherwise.
We see the maritime market, which is the ability to access internet on cruise ships and the like. There are also applications, of course, in the military, particularly in remote locations of the battlefield where the warfighter needs to be able to access internet in an ad hoc fashion. As Vickram mentioned, we also see that the terrestrial telecom industry is embracing satellite communications, for instance, as a means of backhaul to basically communicate the data from the base station back to the data center without having to install expensive optical backhaul infrastructure. There are, of course, SATCOM on the move applications where we have moving vehicles, whether commercial or military, being able to access satellite networks in highly mobile scenarios. The key thing is that many of these applications are actually uniquely addressable only through satellite networks as opposed to the traditional terrestrial networks.
That is the first big tailwind. Secondly, as Vickram mentioned, you know, geopolitical forces across the world are causing every country, every region to invest in local domestic satellite communication networks and capability as that is seen as very important for, you know, strategic national security. Finally, the third tailwind is the lowering of the barrier to launch satellites. What was once the domain of, for instance, NASA and other large federal organizations has now become a capability that, you know, big and small private companies are also able to access. More and more satellites are going up, enabling an increased number of satellite communication networks. If we go to the next slide. The second market that we are addressing is 5G millimeter wave.
This market has, of course, historically been slow to develop, but now we see 5G millimeter wave deployments around the corner starting in 2026. Fixed wireless access has emerged as the killer application for 5G millimeter wave. This is essentially using wireless to bring the internet inside the home over the so-called last mile as an alternative to the more traditional copper or optical internet access. This is really being driven by greenfield markets such as India and other developing nations where that incumbent infrastructure of copper is lacking and hence fixed wireless access is providing a much lower cost, lower CapEx means of bringing internet into the home. If we go to the next slide. When we look at these markets that we are addressing, these are very large markets.
They are actually a $1 billion addressable market for Sivers based on our technology today by 2030. Taking a look at satellite communications, this is a very diverse market. It has many global as well as regional players. That totals to around $400 million of addressable market if we look at electronically steered antennas that utilize beamformers in the frequency ranges that our technology can address. We have several key design wins here that will drive our ability to participate in this market over the next several years. Of course, we've talked about ALL.SPACE and we'll have some more information about ALL.SPACE further down this presentation. We also have design wins with Thorium Space and a leading European SATCOM network provider, as well as several other very important customers in the pipeline. We'll talk about that as well further downstream in this presentation.
Within 5G millimeter wave, as I mentioned, fixed wireless access is the key driver here. The key thing about our technology is we already have standard products that address all of the 5G millimeter wave bands as well as all possible configurations of those standard products, whether it's beamformers for large arrays or highly integrated transceivers for user devices such as CPEs for fixed wireless access. We have some critical design wins with a tier one telecommunications infrastructure provider as well as a leading Japanese infrastructure manufacturer. These will allow us to very actively participate in a roughly SEK 260 million addressable market by 2030. Finally, we have some more emerging markets that we are starting to explore. These include defense and FR3. I'll talk a little bit about each of those.
In the defense market, we find that radar systems as well as electronic warfare systems are increasingly beginning to exploit electronically steered arrays based on commercial beamforming chips so that they can leverage the economies of scale that are driven by the commercial market. These applications can either use our existing beamformers or variants of our beamformers at the specific frequencies that defense applications are targeting. We have some exciting engagements in the pipeline here that will allow us to participate in this market moving forward. On the 5G, 6G side, the FR3 frequency band refers to a range of frequencies that lies between the sub-6 gigahertz frequencies that are used today and the millimeter wave frequencies that are beginning deployment. That range of frequencies runs from about 7-24 gigahertz.
The exciting thing about that range is that it combines the best of both worlds of FR1, which is the sub-6 gigahertz, and FR2. Namely, it has the coverage characteristics of the sub-6 GHz bands while having bandwidth that is competitive with the millimeter wave frequency bands. What's interesting about both the defense and the FR3 markets is that we have received funding via the CHIPS Act under the Microelectronics Commons Program that allows us to develop standard products for both of these markets in collaboration with some of the key customers that we would eventually sell to, including Ericsson, BAE Systems, Raytheon, etc. All of these together form a very exciting $1 billion directly serviceable opportunity for Sivers over the next five years. If we go to the next slide. However, each of these opportunities have challenges today which our technology is uniquely able to solve.
This graphic on the left-hand side shows some of the challenges that SATCOM networks face today. A little later in this presentation, I will talk about how the differentiating features of our products address these challenges directly. The graphic is showing a terminal that is now trying to communicate with multiple networks in different orbits, the low, medium, and geostationary Earth orbits. Different challenges in trying to do this are highlighted on the right-hand side. The first challenge is that today's satellite networks and operators do not actually communicate with each other. For instance, an Intelsat satellite cannot communicate with a ViaSat satellite, for instance. The second challenge is that ground terminals today cannot simultaneously communicate with multiple satellites.
This creates a challenge for what's called make-before-break scenarios where, as one satellite is leaving the field of view, one needs to establish a link with another satellite before the first connection is actually broken. Satellite terminals are also not able to communicate across different orbits. Obviously, the requirement to communicate with a lower-third orbit satellite that's much closer to you is vastly different from the higher performance that's required to communicate with a geosatellite. As a result, we don't see convergence across multiple orbits. The fourth challenge is there's also no convergence today between military and commercial networks. They each run on a different set of standards. A different set of terminals are built for each. There's also no convergence, as you see in point number five, across the different frequencies.
The two dominant frequency bands that are used for these satellite communication networks are Ka-band and Ku-band. Ku-band is roughly 10-15 GHz, while Ka-band is 17-31 GHz. Some of these orbits are in the Ka. Some of these orbits are in the Ku. Terminals today do not have the ability to operate on both bands simultaneously. Finally, we do not see convergence today between satellite networks and your terrestrial networks such as 5G. All of these things are problems that need to be solved in order to realize this significant growth that we foresee in the SATCOM market. If we go to the next slide. How we are able to address these challenges is through the broadest millimeter wave IC portfolio that we see amongst our competitors.
On this slide, we are seeing Sivers' products grouped across the level of integration on the X-axis and the frequency and application they address on the Y-axis. On the left-hand side, you see what we call high-performance beamforming ICs. These are ICs that basically form the beam across the antenna array. Really, the level of integration is limited to the RF circuitry to form the beam, and the focus is more on performance. Higher output power, better efficiency, lower noise figure. On the bottom, you see the Timberline family of beamformers, which have been designed for ALL.SPACE. On the top, you see the Summit family of beamformers that are broad market products that address the 5G market. On the right-hand side, you see our more highly integrated RFIC transceivers. Here we have more functionality that has been put onto a chip.
In addition to the beamforming, there is also the frequency conversion to baseband, the PLL synthesizer, and the power management IC, power management circuitry. Here, the focus is less on getting the highest performance possible, but more having a highly integrated chip that can be the single chip solution that is needed for all RF analog functions that might be needed on an end user device, for instance, a CPE. On the bottom right, you see the EBRA transceivers. Again, that's custom designed for ALL.SPACE. On the top, you see the broad market transceivers for the 60 gigahertz and 5G markets, namely EDR and Rapenio. These are all the internal code names that we use within Sivers for these chips. If we go to the next slide.
The key thing is that these products that we have significantly lead competitive products on the market in terms of performance. Here you see a depiction of the Gen 1 beamformers from Sivers, which were released to market about four years ago, as well as the Gen 2 beamformers that have been released to market recently over the last year or so. They are shown against the competition on the axes of output power and efficiency. Even at Gen 1, we had a significant performance benefit over competitors, which you see here grouped based on the semiconductor technology that they use. The gray dot represents competitors that use a bulk CMOS technology. The yellow dot shows competitors that use a silicon germanium technology. The green dot shows competitors that use the same RF SOI silicon on insulator technology that we at Sivers use.
Because we have unique IP and circuit design capability in these beamformers, even the Gen 1 beamformer had a significant performance improvement. We are continuously working on our product development, coming up with new innovations and bringing them into our products. As a result, our Gen 2 beamformers have as much as 6-12 dB, which is 4-10 times higher output power and 3-5 times higher energy efficiency than competitive products on the market. What this means for our customers, their value proposition is the higher output power enables higher propagation range. It allows the millimeter wave link to be formed over longer distances as well as overcome barriers such as foliage, the human body, etc.
The higher energy efficiency means that at the same time, our customers' products will consume less DC power, less energy, and also dissipate less heat into the environment. This leads to a greener deployment footprint as well as a lower cost of ownership for our customers. If we go to the next slide. Our differentiation is not just on these key performance metrics of output power and efficiency. We also offer other differentiating features that address some of the challenges that I talked about earlier. For instance, our beamformers, while having this superior output power and efficiency, are also able to form multiple simultaneous beams. This allows terminals that are built with these beamformers to communicate with multiple satellites at the same time and hence achieve make-before-break, as I talked about earlier.
We're also actively working on multi-band beamformers, so beamformers that can operate in the Ka and the Ku-bands simultaneously. This allows our customers to build terminals that, from a single antenna aperture, can operate in the Ka and the Ku-band simultaneously and thus establish links with networks that might be on different frequency bands. Finally, traditionally, SATCOM terminals have a separate transmit antenna and a separate receive antenna. We're working on beamformers that can transmit and receive at the same time from the same beamformer, which would allow our customers to share the antenna or aperture across transmit and receive functions, which again would directly reduce the size as well as the cost of these SATCOM terminals. These differentiating features are critical in our ability to win customers, but also our customers' ability to then win their opportunities with their customers.
If we go to the next slide. This broad product portfolio and its differentiating features has helped us to really grow our pipeline. If you look at the stages on the right of the pipeline, the companies in mass production as well as the design win and design in customers, these are the familiar customers that we've talked about earlier in the presentation. What we see is that now we have a very healthy and robust set of customers that we are in deep technical engagement with today that we're attempting to transition into the design in column. You see as many as 30 customers, many of them in satellite communications. Behind those, we have several other customers, as many as 24 in the qualified opportunity bucket where discussions have begun and are deepening in their technical engagement.
As Vickram mentioned, one of our key focuses has been to concentrate on high-brand value customers that we feel confident can ramp to mass production in the near future. A lot of the names that you see here fall into that category. If we go to the next slide. A few of the customers that we see here are the design win and mass production customers. I'd like to talk about the traction that they're particularly seeing. Of course, our lead customer is ALL.SPACE on the left. They've hit several important milestones recently. Their Gen 1 terminals, which are called Hydra 2 and Hydra 4, are in general availability now and are actively deployed with the U.S. Army and the U.S. Navy.
The background behind that naming is Hydra 2 is able to form two simultaneous beams, while Hydra 4 is able to form four simultaneous beams. These are in deployment today, as is evidenced by the fact that by Q3 of this year, we would have actually shipped 1.3 million beamformers to ALL.SPACE to support all the terminals that they have shipped. They also have a very exciting sales pipeline into the U.S. DOD, which will drive larger orders for us later this year from ALL.SPACE, which will allow us to then supply beamformers for that growing pipeline. On the 5G side, we have our Tier 1 telco infrastructure provider. We're actively working with them for them to build their Gen 1 product today, which would use Sivers' broad market 5G beamformers and RFICs.
The Gen 1 product would use the Summit Family of beamformers as well as the Rapenio RFIC. That product is expected to ramp to volume in 2026. At the same time, we have a $5.4 million NRE contract with them to design a next-generation product in the GlobalFoundries 22FDX process, which really combines the best of both worlds of our beamformers and RFICs so that next-gen product achieves the output power, efficiency, and noise figure, the performance metrics of the beamformers, while also having that high level of integration that we see in our RFICs, meaning they include the up-down converter, the synthesizer, the power management, the analog baseband circuitry, etc. Finally, I'd like to talk a little bit about Thorium Space. That's our second in-line customer in SATCOM.
The unique thing about our engagement with Thorium Space is that the engagement includes both beamformers for ground terminals as well as for the space payload. That would mark our first foray into designing chips for space. The first ES1 ICs, engineering sample 1 ICs for both the ground and the space segment, are currently in manufacturing today. We will sample them to the customer in Q4 of this year. Meanwhile, we have seen that Thorium Space has signed numerous strategic partnerships with satellite network operators, which builds confidence in their volume deployment plans a couple of years down the road. Let's now go to a video interview with Paul McCarter, CEO of ALL.SPACE, to talk a little bit about their technology, their initial deployments, their customer traction, as well as the value proposition that our technology brings to ALL.SPACE.
Hello. My name is Paul McCarter.
I am the CEO of ALL.SPACE. I was asked to do a quick video for you on the background of ALL.SPACE and why we chose Sivers. Now, I believe you will have watched the Seraphim video to give you a background on satellite communications and why right now phased array technology is pivotal to the expansion in communications going through GEO, MEO, and LEO orbits. Basing on that, just to give you a quick background on ALL.SPACE, about two years ago, the U.S. DOD ran two competitions, one for the U.S. Navy, one for the U.S. Army. They were to find a critical piece of technology that could enable resilience in their satellite communications. They searched the world, and in both of those independent technical teams, they chose ALL.SPACE for that capability.
The reason for that is because we currently are the only providers of technology that can do simultaneous full duplex, full bandwidth links from a small aperture phased array to satellites in any orbit. Part of that is the fact that we have very, very carefully selected the component providers in our electronic supply chain. And Sivers is a critical part of that. What that's led us to is the R&D programs are converting to programs of record in the U.S. DOD. The U.S. Navy program has already converted in January of this year to the commercial broadband satellite program of record. We have delivered 32 units to them already in 2025. They will be expanding that, and the numbers for that IDRQ contract will be going up.
The U.S. Army has several programs of record, which we are now in a really good position to win and will be ordering units off us in this coming year. That pipeline that we have right now, since we've demonstrated the capability in November and December of last year, is in excess of $200 million worth of opportunities. However, that is nothing in comparison to the programs of record that they require, like next-generation command and control, the family of terminals, because all of the old gimbal technology that is based on both land and sea needs to be swapped out for new phased array technology. Now, that's just talking about the U.S. DOD.
As we all know, NATO and Europe will have to upgrade its defensive over the next decade, and there will be a huge amount of investment coming from investment in things like the U.S. DOD for multi-orbit technology capability. We are in a really good position from a defense market perspective. Now, just to add to that, the technology itself isn't defense-oriented. It's dual use. We are in conversations right now with airline operators who wish to take the technology, do investigations, and put it onto aircraft so that they can provide the next generation of multimedia delivery whilst people are on flights. It's a huge opportunity right now as we expand into this GEO, MEO, and LEO orbit capability. We were asked the question, why Sivers?
Just as the U.S. DOD did their sweep of the market, we also did the sweep of the market for who we needed to partner with to provide what is a critical part of our technology. When we looked at this, we believe Sivers has got a differentiated technology. It has got a higher power capability, and that leads to a better efficiency for how we use it with a lower power consumption. All of that adds to less cooling requirement, which is a very important part for the electronics. When you add it all up into a business case, it means that the lower cost of ownership gives us an advantage for the complete unit as we move forward. We selected our component suppliers because they give us an advantage.
We believe Sivers is one of the best in the world at doing what they do, and we work with them to ensure that we can therefore deliver our simultaneous multi-orbit capability. I'd like to say thank you to Vickram and his team. They've always been there for us, and they work tirelessly so that we can all have great trading opportunities and move forward with the best technology. Thank you. I hope you have a great day.
As we look to the future, we also have a very robust product roadmap that will allow us to build upon these initial successes. First, I'd like to highlight Cloud Chaser, which is the internal code name for our Ka-band standard product beamformer for satellite communications.
This is currently in manufacturing today and will be available and sampling to customers in Q4 of this year. Many of the customers that you saw in the pipeline slide that I showed earlier are eagerly awaiting this standard product beamformer. This will allow us to really take the technology and the differentiating IP that we have and expand the number of customers that we can service with that IP and technology. Early next year, you see the production phase of our Summit family of broad market beamformers for 5G. This is already sampling with customers today. The 5G Tier 1 telco is using it today for product development, but the production release of this beamformer will happen in early next year to support that product ramp of the 5G Tier 1 telco.
There are several other customers behind that customer that are using the chip today and are awaiting the production release. You see a couple of our custom chips, Juniper for Thorium Space, the Ku&Ka dual-band chip for ALL.SPACE, and BIFROST, which is the internal code name for our next-gen 22FDX 5G transceiver for the Tier 1 telco. Those will be released through late next year and early 2027. Later in 2027, we anticipate the release of our Ku-band standard product beamformer for satellite communications, thus giving us comprehensive coverage of the SATCOM ecosystem. Of course, I should mention in parallel, we have more long-term product development ongoing at transceiver for the FR3 frequency band, as well as a self-interference counselor for the defense market for electronic warfare systems.
These, as I mentioned earlier, are funded under the CHIPS Act, and their production release will be later in the timeline, but we're already actively engaging with the customers under that CHIPS Act project to ensure that the product that is being designed best fits their needs. If we go to the next slide. One thing that we've realized over the last few years at Sivers is that although our primary focus is in building and selling chips, it is very valuable for us to climb up the value chain and offer antenna arrays as both reference design and as full products. Here we're talking about products where the chip is integrated on a PC board or in a package with the antennas tightly integrated and connected to the chips themselves.
We find that this is valuable for our customers because it accelerates their time to market and reduces their development cost and risk. For us, it allows us to climb up the value chain and capture a greater percentage of the value of the final end product that our customer would sell. Historically, we've done this in the form of evaluation kits, mainly as test vehicles for our ICs, as well as small arrays for 5G and 60 GHz. We're talking about arrays that have 16, 32, maybe 64 antennas. This year, we've taken the plunge to essentially expand these antenna array offerings to large arrays for both 5G as well as satellite communications. We're talking about arrays with 128, 256, and as many as 512 antennas for 5G mm wave, and then panels with as many as 2,900 antennas for satellite communications.
We see a lot of excitement in the market both to access these designs as reference designs so that the customers have a starting point to design their custom module or panel, but also as complete products which our customers can sort of build a box around. We go to the next slide. We have actually innovated and sort of expanded our scope even further with our recent Intelsat engagement marking our entry into digital subsystems. Here for Intelsat, we are building a product which is called a DIFI down converter. DIFI stands for digital IF interface. What this DIFI frequency converter would do is that it would take the L-band intermediary frequency that comes from the antenna panel. L-band is from roughly 900 MHz to about 1.5 . That is the frequency that comes out of the panel.
The millimeter wave is converted down to that frequency. This DIFI converter will further take the signal from the L-band to digital bits and put it on this DIFI standard interface in an Ethernet protocol. What this then allows is that if this DIFI converter is integrated with the SATCOM terminal, it can have an Ethernet connection to the eventual modem, which allows the modem to be remotely located from the satellite terminal as well as virtualized, meaning you do not need dedicated modem hardware anymore. The modem can essentially be software code running on a server somewhere, on a general processor somewhere. This is an exciting product because it really expands the deployment capability of these SATCOM terminals. Now, the SATCOM terminals can be remote from the modem. It marks an entry of Sivers into digital systems.
It has really created a lot of awareness of Sivers once this announcement with Intelsat came out, in particular at the Mobile World Congress earlier this year in early March. At the SAT Show in Washington, DC in the middle of March, we had a lot of customers coming to us because of the announcement with Intelsat on this DIFI frequency converter. That has really expanded our customer base quite significantly. We are executing this project with NXP as our partner. NXP provides the Layerscape platform, which is the hardware platform on which this DIFI converter is implemented. To hear a little bit more about not just this project, but also Intelsat's view of the SATCOM market and how this DIFI converter can really expand that market, and to end this presentation as well, let us now go to a video by Dr.
Salim Yaghmour of Intelsat to talk a little bit about our partnership. The market is experiencing significant growth and transformation driven by advancement in technology, strategic partnership, and increasing demand for mobile SATCOM solutions. Some of what's happening in the SATCOM market and the innovation and technology, the fast deployment of large NGSO constellation, including the emergence of the direct-to-device NTN services, deployment of the software-defined satellite fundamentally changed the game delivering services when and where they are most needed, and dynamically steering bandwidth and power as needs change, enabling adaptive use cases for various mobility applications, utilizing AI to optimize resources and network performance. The evolution of SATCOM ground infrastructure and following the 5G NTN standard to meet the demand of modern scalable and software-defined networks for supporting flexible multi-orbit, multi-network operation, particularly as satellite constellation scale and services become more dynamic.
Intelsat is building the future of global communication with the world's first hybrid multi-orbit software-defined 5G networks. Intelsat has partnered with Sivers to develop the next generation high-performance digitizer, which is an important enabler for the SATCOM for the smart-edge mobility terminal to provide our customers with the full flexibility to manage terminal configuration and applications. The architecture shift in ground terminals from hardware-centric architecture with purpose-built specific frequency band to a flexible, scalable, ease of integration, and cost-effective offers the ability for the Sivers Intelsat partnership to provide differentiated solutions in the SATCOM market as a new constellation and terminal infrastructure are being rolled out. We are happy to partner with Sivers and looking forward to a successful project in building a unique digitizer solution. We told you we had an exciting program. We're doing a 10-minute break while we're playing a video, Ayar Labs. Stick around.
There are two more sections coming, three more sections coming. We'll deep dive into the photonics business, and we'll look at our manufacturing strategy and the Q&A, which should be exciting. See you in 10 minutes.
We're seeing evolutionary breakthroughs in technology, such as ChatGPT, that are enabling truly transformative outcomes across many areas of our lives. The foundational AI models powering those possibilities have become exponentially larger and more complex, requiring far more GPU processing and memory capacity than can be provided by a single GPU, or node, or even a single rack. Today, when those AI models need to move data across nodes, the nodes are typically connected over expensive, low-density, pluggable optics with low-bandwidth electrical I/O connections inside each node.
That results in massive bottlenecks by forcing your GPUs to remain idle the vast majority of the time, reducing compute efficiency and performance while creating large increases in power consumption and capital and operating expenses. If we're going to get the most out of AI today, let alone the future, we need to eliminate the problems created by traditional interconnects. We need a way to connect nodes at scale so they effectively work like a single giant GPU. Fortunately, there's an answer. Ayar Labs in-package optical I/O composed of Terrify optical I/O chiplets and the Supernova remote light source. Integrating optical interconnects within the GPU and switch packages enables data to be transmitted at significantly higher throughput within each node and across nodes. In-package optical I/O delivers up to five times higher data rates and 10 times lower latency, with up to eight times the power efficiency versus traditional interconnects.
The solution is also highly cost-effective, enabling greater link density to flatten the network infrastructure and reduce the number of performance and power penalizing trips while providing uniform latency when transmitting data across nodes. With more efficient communication, GPU utilization rises significantly, so AI tasks complete faster. That means you can deploy fewer GPUs and switches while reducing power, slashing your CapEx and OpEx for your AI needs today. You can efficiently scale your infrastructure to handle your ever-growing AI needs tomorrow. Discover more about how Ayar Labs is shattering the barriers to AI at scale. We're seeing revolutionary breakthroughs in generative AI, such as ChatGPT, that are enabling truly transformative outcomes across many areas of our lives.
The foundational AI models powering those possibilities have become exponentially larger and more complex, requiring far more GPU processing and memory capacity than can be provided by a single GPU, or node, or even a single rack. Today, when those AI models need to move data across nodes, the nodes are typically connected over expensive, low-density, pluggable optics with low-bandwidth electrical I/O connections inside each node. That results in massive bottlenecks by forcing your GPUs to remain idle the vast majority of the time, reducing compute efficiency and performance while creating large increases in power consumption and capital and operating expenses. If we're going to get the most out of AI today, let alone the future, we need to eliminate the problems created by traditional interconnects. We need a way to connect nodes at scale so they effectively work like a single giant GPU. Fortunately, there's an answer.
Ayar Labs in-package optical I/O composed of Terrify optical I/O chiplets and the Supernova remote light source. Integrating optical interconnects within the GPU and switch packages enables data to be transmitted at significantly higher throughput within each node and across nodes. In-package optical I/O delivers up to five times higher data rates and 10 times lower latency, with up to eight times the power efficiency versus traditional interconnects. The solution is also highly cost-effective, enabling greater link density to flatten the network infrastructure and reduce the number of performance and power penalizing trips while providing uniform latency when transmitting data across nodes. With more efficient communication, GPU utilization rises significantly, so AI tasks complete faster. That means you can deploy fewer GPUs and switches while reducing power, slashing your CapEx and OpEx for your AI needs today.
You can efficiently scale your infrastructure to handle your ever-growing AI needs tomorrow. Discover more about how Ayar Labs is shattering the barriers to AI at scale. We're seeing revolutionary breakthroughs in generative AI, such as ChatGPT, that are enabling truly transformative outcomes across many areas of our lives. The foundational AI models powering those possibilities have become exponentially larger and more complex, requiring far more GPU processing and memory capacity than can be provided by a single GPU, or node, or even a single rack. Today, when those AI models need to move data across nodes, the nodes are typically connected over expensive, low-density, pluggable optics with low-bandwidth electrical I/O connections inside each node.
That results in massive bottlenecks by forcing your GPUs to remain idle the vast majority of the time, reducing compute efficiency and performance while creating large increases in power consumption and capital and operating expenses. If we're going to get the most out of AI today, let alone the future, we need to eliminate the problems created by traditional interconnects. We need a way to connect nodes at scale so they effectively work like a single giant GPU. Fortunately, there's an answer. Ayar Labs in-package optical I/O composed of Terrify optical I/O chiplets and the Supernova remote light source. Integrating optical interconnects within the GPU and switch packages enables data to be transmitted at significantly higher throughput within each node and across nodes. In-package optical I/O delivers up to five times higher data rates and 10 times lower latency, with up to eight times the power efficiency versus traditional interconnects.
The solution is also highly cost-effective, enabling greater link density to flatten the network infrastructure and reduce the number of performance and power penalizing trips while providing uniform latency when transmitting data across nodes. With more efficient communication, GPU utilization rises significantly, so AI tasks complete faster. That means you can deploy fewer GPUs and switches while reducing power, slashing your CapEx and OpEx for your AI needs today. You can efficiently scale your infrastructure to handle your ever-growing AI needs tomorrow. Discover more about how Ayar Labs is shattering the barriers to AI at scale.
All right, that was a very quick 10 minutes. Welcome back. The last stretch of this day, Andrew, my dear colleague and CTO, is going to talk us through business and technology.
Okay, thank you for the introduction, Sander. Good afternoon to everybody in person and also online.
It's great to have the opportunity to talk in a bit more detail about the technology that we have within the Photonics Business Unit here at Sivers. Primarily, I'm going to focus on the opportunity, which is by far the biggest opportunity that we have in Photonics, of supplying laser devices into the AI infrastructure and data center market. Next slide, please. Okay, we've already heard in a lot of detail from Vickram and also from Tim about what's happening within AI infrastructure. Let me just quickly get my perspective about the sort of broader landscape and what's happening here in the developments around AI infrastructure. Obviously, over the last couple of years, we've seen sort of a rapid increase in the large language AI model size.
You know, we likely will see models in excess of 100,000 GPUs in 2026, which has scaled dramatically over the last two years. You know, that is creating sort of serious bottlenecks where interconnect and high bandwidth memory just really cannot keep up with the sort of computational advances that we've seen from the likes of NVIDIA and AMD. You know, we have a very large sort of mismatch that's been created, causing significant issues in terms of the system network performance and also in the raw energy requirements. These models really have scaled from, you know, millions of parameters into, you know, hundreds of billions of parameters recently. You know, ChatGPT 4 is an example, Gemini, xAI, etc. Fundamentally, they require sort of massive data movement between the computation units and the high bandwidth memory. There are major issues here.
Today, the GPUs, as we've heard, you know, a few times today, you know, they're connected fundamentally through copper within the data racks themselves, you know, high-speed electrical links. There's a number of different protocols that are used here: NVLink, InfiniBand, PCIe, Ethernet, etc. You know, these links really have sort of got, you know, finite bandwidth and very much non-zero latency. We've heard about idling of the GPUs, which can significantly degrade the system efficiency. Really, data exchange between the GPUs is a major bottleneck and a major issue, especially as we move to the larger models where there's, you know, very high levels of parallelism.
The other issue, of course, is power and the energy per bit, you know, for these electrical links is currently very, very high, you know, several tens of picajoules per bit to transfer from one GPU to its neighbor. Again, this very much limits the scalability of these networks. You know, some of the media coverage that is shown here, you know, tries to sort of highlight this and really sort of quantify some of the issues that we face. You know, we have, for example, Sam Altman talking about, you know, very, very subtle small inefficiencies within these models costing millions and millions of dollars. This is why any sort of incremental improvement that we can make to these networks has got a massive cost benefit to the users of these networks.
If we also look at, you know, some of the other sort of diagrams that we have on this chart, you know, top right, we have the power consumption for all of the AI, you know, data centers. You know, it's approaching the sort of energy usage of the U.K. grid. You know, it's an incredible statistic. That's really driving, you know, people, you know, taking on some pretty radical approaches to powering these data centers in the forms of sort of small nuclear reactors. You know, it's an incredible thing to consider that people are doing this. This is a very real activity in the sort of industry today. Where does Sivers fit into all this? Fundamentally, our laser technology is, you know, focused on improving the GPU interconnect performance, both in terms of performance, but also in terms of efficiency and system latency.
Next slide, please. What we're showing on this chart here is really sort of direct evidence of the sort of industry acceptance that optical interconnectivity is critical for the success of AI. We have some quotes from NVIDIA, you know, obviously the main sort of provider of the infrastructure for these networks. We have one of the users, OpenAI, and then also from Lightmatter. Lightmatter are a very disruptive startup in this space, bringing sort of, you know, very sort of novel optical connectivity into the sort of data centers. There is very sort of wide acceptance, you know, that this adoption of optical interconnectivity is really critical for the expansion of the AI networks in the future. Next slide, please.
If we go on the premise that, you know, replacing copper today, the copper interconnects that are in the data racks today with optics, that's quite a broad statement. You know, optics can mean lots of different things. It can be indium phosphide, you know, integrated modulators. It can be silicon photonic Mach-Zehnder modulators. But really, there's a clear lead technology, which is silicon photonics micro ring resonators co-packaged with GPUs or ASICs. And they're becoming the very sort of clear lead technology within this industry. There's very much wide acceptance that this is the approach that's going to be the winner as we deploy this, you know, sort of optical interconnect technology over the next few years. That's because of a few sort of key advantages of this technology. The primary one is energy efficiency.
This technology is the only solution that really gets down below the sort of industry target of 5 picajoules per bit. We mentioned copper; copper is stuck up at around 50 picajoules per bit. Traditional pluggable optics, even LPO, which is the sort of linear pluggable optics, are stuck around 15 picajoules per bit and do not have a roadmap below 10. We have a clear sort of, you know, solution here to get down to the industry targets. One of the other key benefits and advantages that we have here is the potential for very high channel density through a sort of scalable, highly parallel architecture. What does that mean? That means basically multiple signals can be sent down each fiber, what is traditionally called WDM, you know, providing routes to 16 terabits bandwidth. I will explain that in a second as to how that is actually realized.
Pluggables today, as Tim mentioned a week while ago, you know, are typically shipping at 400G or 800G. They have got a clear roadmap to 1.6T, maybe 3.2T, but nothing beyond that. You can see where pluggables start to run out of steam. The micro ring resonator approach really has got, you know, the opportunity to take it to a completely different dimension. The other big advantage is, you know, this stuff is produced in high volume silicon fabs. Okay, you know, the two main companies delivering this technology today are GlobalFoundries and TSMC, who are, you know, obviously some of the leading silicon foundries out there today. They are both sort of, you know, working on very, very similar technology.
GF may be slightly more advanced today, but TSMC will almost certainly catch them up in the next sort of 18 months. Quick click, please. Let's explain how we get up to 16 terabits of bandwidth with this technology. We do this by using multiple wavelengths. Sixteen wavelengths, if we just do a couple of clicks, you can see 16 wavelengths. Click again. We put those 16 wavelengths down 16 different fibers in a fiber array. Then what we can do is modulate each of those individual 256 carriers at 64 Gb. That aggregates up to a 16 terabit bandwidth signal. This is described in the industry as sort of highly wide, slow approach, but actually it's very, very, very wide. Sixty-four gigabits is not particularly slow.
64 Gb is what can be done today with CMOS resonator technology, but that has got the potential to go up to 200 Gb This technology is not fundamentally limited to 16 terabits bandwidth. Incredible bandwidth density that the pluggables just cannot match in the future. Next slide, please. What do we do? Where do we fit into this great sort of ecosystem? We are supplying the laser sources for CPO, as Vickram mentioned earlier. We currently are the technology leader in this space. You know, we have, you know, developed this competitive edge built on sort of 25 years of advanced laser development across a few different markets, including sort of telecoms, PON networks, sensing applications, including consumer biometric diagnostic sensing and also LiDAR sensing. That is a couple of the outposts that Vickram touched upon earlier.
They could, of course, be very interesting in the future. Really, we've sort of leveraged the technology that we've developed to focus on this great opportunity. Today, we've got, you know, significant strength in the chip design IP. We've got patented technology around this array format. We've built upon sort of very well-established process technology that we have here in-house in our Glasgow facility. We use traditional ridge waveguide laser architecture. It's simple, proven, reliable technology. It's used today in other parts of the network. We have a couple of, you know, quite novel approaches that give us the edge across our competition. We've etched facet designs. This is leading to sort of highly scalable volume manufacturing. We also do on-wafer optical coatings and on-wafer optical testing.
This allows us to, you know, have sort of industry-leading yields, which can obviously allow us to drive the cost down and margins up. Today, we're doing chip development and prototyping here in our internal Glasgow fab located right underneath where I'm sitting today. We have a process transfer underway to a high-volume Asian fab in Taiwan that Alex is going to cover in a bit more detail in the later presentation. Next slide, please. I like this slide a lot. This really shows the architecture in action. We're using an Ayar Labs schematic, but it's a sort of beautiful representation of how our laser technology fits in with the Ayar Labs co-packaged optics right into the GPU itself. The core of this architecture is the GPU. This could be a Blackwell, next-generation Blackwell GPU chip from NVIDIA.
What you can see is it being surrounded by these silicon photonic chiplets. The four gold boxes, which are manufactured in GF. What's happening here is they're connected through a high-speed interposer that they both sit upon. You have a very high-speed electrical signal coming from the GPU over to the CMOS chiplet, which then converts it from the electrical domain into the optical domain as locally as possible. What that basically means is you're transmitting high-speed RF signals over millimeters as opposed to meters within a traditional data rack today. That's where most of the energy saving actually comes from. We then have our laser. Our laser really powers this entire optical circuit. Our laser is sitting inside a remote module. All of the laser wavelengths are multiplexed and split into the multiple fibers.
You have fiber arrays coming into this co-packaged optical module. In the future, these modules, the GPU modules, there'll be no high-speed interfaces. It will be all optical fiber connectivity. There'll be a DC power supply, of course, but all the data coming off the module will be in the optical domain. This really facilitates, you know, bidirectional connectivity from one GPU to the next in the scale-up architecture that we've mentioned a couple of times already through the presentations. The other important thing to point out here is the lasers are remote. There's a couple of good reasons for that. You know, fundamentally, the GPUs are generating huge amounts of waste power. They're liquid-cooled, you know, to get them down to sort of, you know, 100 degrees Celsius temperature. The lasers don't really like to operate at that temperature.
Having the laser remote allows it to be operated in a much more benign temperature environment, which improves its efficiency, which is important from a power-saving perspective. It also improves its reliability. You know, the laser can be in the front of the rack, it could be on a tray as we looked, as we saw earlier from Ayar Labs, or it could be at the back of the data center. You know, the fibers are effectively lossless. The laser has the potential to be put anywhere within the data center. The other advantage is if the laser does fail, which is very rare, actually, you can unplug it and put in a replacement. You're not having to throw a $40,000 Blackwell module into the bin if it fails. This is the true essence of co-packaged optics.
This is a really nice slide that, you know, really demonstrates what is co-packaged optics. You know, it's getting those silicon photonic chiplets right in beside the GPU, being powered by a remote laser source and providing, you know, true optical connectivity from one GPU to all its neighboring GPUs in the network. Next slide, please. So we've mentioned Ayar Labs, and this is some of their sort of latest, you know, configurations of, you know, multiple, you know, photonic chiplets surrounding this sort of GPU or ASIC, if you will, they've got 10 in this particular example. They're not the only people working on this. If we just click on, please. NVIDIA themselves are obviously now very active and publicly active in this space. At the GTC conference, we saw Jensen trying to untangle the optical cables, which was quite amusing.
You can see they're adopting a very similar approach, surrounding these GPUs with multiple silicon photonic chips. This is really exciting for the industry. The other key thing is that they've openly stated they're using micro ring resonators, exactly the same approach as Ayar Labs. This is an incredible endorsement of this technology and how it's going to scale in the future, both for scale-up architecture, but also for scale-out architecture. Just click on, please. Both of these companies are working with both of the silicon photonic foundries. As I mentioned, GlobalFoundries may be slightly more advanced today, but TSMC will catch them up very quickly. They're very, very mature, high-volume foundries supporting this silicon photonic micro ring resonator landscape. Next slide, please.
We also have an optics industry MSA consortium. That's a multi-source agreement consortium defining the laser standards specifically for this ring resonator architecture. Now, we have a number of promoter members and, you know, there's some very large tier one companies in there, but we're also in there as well. We've been working within this consortium for about four years now, defining the laser standards that's going to power the micro ring resonator architecture. That defines things like wavelength grids, frequency grids, optical power requirements. We've been a sort of core partner in this process and really shaped and influenced, you know, where the standards are today. On the right-hand side, you can see all the sort of observer members of this MSA. You can see all the big players within the industry.
If you just do one click, you know, you have Nvidia highlighted, AMD, of course, very active in this space. They're extremely interested in this technology. And also Inphi, that Tim mentioned earlier, now part of Marvell, also very, very interested in this approach. You know, this is a great endorsement of the whole architecture. You know, strong involvement from industry leaders. And we are very much at the heart of that right now. Next slide, please. When do we go to volume? That's probably the key question. Again, Vickram alluded to this a little bit earlier. 2027 is the target date for volume production. This year, you know, we're starting qualification. You know, we've got the designs that are frozen. The technology is developed. We're starting the qualification program. 2026, we complete qualification.
We start to do pilot ramping up into production for 2027. A very, very exciting time for us. To put into perspective, you know, what's the size of the opportunity here? In 2028, the AI industry GPU shipments is forecasted to be 19 million. If you look at the architecture that we've looked at over the past few slides, you can see that any given GPU is surrounded by multiple silicon photonic chiplets. Ayar Labs showed 10. Nvidia showed 32. The typical architecture in terms of powering from the laser is that one laser is driving two chiplets at the moment. You can see in terms of quantities, if you take the Ayar Labs approach, then we need at least five laser sources per GPU. If it's Nvidia, then that's going to be 16 laser sources per GPU.
So the potential, you know, market size for this is, you know, approaching hundreds of millions of laser sources per year as we go through 2028 into the future. So an extremely exciting, large, addressable market, you know, when this really starts to become deployed in the data centers. Next slide, please. I think this is really my final slide just to sort of summarize. So, you know, for sure, we've got these sort of mega trends in the sort of AI infrastructure driven by, you know, rapid developments in compute performance. But ultimately, the demand for lower power consumption. The current IO solutions, whether it's copper within the racks, pluggable between the racks, they just cannot deliver the required improvements in the future. They do not have a roadmap to do that. This is why there's intense interest and activity in the ring resonator architecture for the CPU GPUs.
Remote light sources powering the rings are using proven multi-wavelength laser arrays. We have the technology ready today to deploy. And today, you know, Sivers is producing the leading chip performance in the market for the leading customers in the market. So an incredibly exciting time for us. I will just finish off by introducing Mark Wade. Mark is the founder and CEO of Ayar Labs, one of our partners, of course, our key partners in this space where he's going to talk about their technology and the partnership that we have together and how that's developed over the past few years and where that's going to go in the future. I think after that, we're going to hand over to Alex, the Photonics MD, who's going to take a look at the sort of manufacturing scale-up across the business. Thank you. Hello.
My name is Mark Wade, and I'm the CEO and co-founder of Ayar Labs. Today, I want to talk a little bit about what's going on in large-scale AI computing and how that's driving a new need for a new paradigm of higher performance optical connectivity. We all know that AI has changed our lives just in the last few years since ChatGPT was announced. Underneath the hood there, there's been a revolution in computing. We've moved from a paradigm of general-purpose computing from the last few decades now to very specialized computing that's underneath the AI application. In this application, it's basically very large matrix multiplication and mathematics. What's happened is the AI models in these big neural nets have gotten so large that they no longer fit within one computing or one GPU's memory.
You have to spread that model across a large system. This is where you hear about hundreds, thousands, tens of thousands of GPUs being used in these large-scale AI data centers. The first thing to realize is that as the problem of AI spread across all these GPUs, we entered into a regime where connectivity between these GPUs is incredibly important. On some of the largest models that are running today, you have a huge bottleneck in how much data and how much bandwidth you can exchange between all of these GPUs and Axelerators in the large-scale AI system. There is a drive to say we need a new kind of technology and a new technology roadmap to help drive more bandwidth, more connectivity across these large-scale systems. Today, the AI data center is broken up into two parts of a communications network.
There's a so-called scale-out network, which looks more like traditional networking. Ethernet or InfiniBand is used in this portion. That part of the network is already using optical communications. It's inheriting those solutions from the more traditional data center networking set of products and applications. The other portion of the AI compute network, we call it the AI scale-up network. This is the part of the communications fabric that is running the core AI computation. Today, that portion of the system is using electrical communications. Historically, it's always used electrical communications. This is where the core bottleneck comes in and where Ayar Labs is working to solve the communications bottleneck. It also connects to why our partnership with Sivers and work that Sivers is doing to bring forward high-volume laser array technology is very important.
What's happening in the scale-up fabric is as you try to connect more GPUs in that fabric, today you're limited by the bandwidth and distance that you can achieve with electrical connectivity. If we can bring a new generation of silicon photonics and co-packaged optics into that portion of the network, not only is it an incredibly exciting market opportunity because the volume, the number of units that are needed in that portion of the system is much, much higher than what's needed in the scale-out portion. It also solves a huge problem. It gives AI companies, AI researchers, and people building products and solutions on top of large-scale AI a very exciting future. I want to talk a little bit about what Ayar Labs does and the products that we build and how that connects to Sivers.
Ayar Labs builds optical IO connectivity products for large-scale AI systems. That is comprised of two things. There is an optical chiplet side of that, which is the electronic photonic piece that does the very high-density, high-performance integration of advanced node transistors and a new generation of silicon photonics. There is the laser module piece that provides the laser light that connects into the optical chiplets to power up all of this new optical IO bandwidth. On the CMOS side, we partner with leading foundries like GlobalFoundries, TSMC, Intel Foundries to work on this new generation of densely integrated electronic photonic optical chiplets. On the laser side, our partnership with people like Sivers is incredibly important to us. Sivers is working on a unique DFB laser array technology. We have been able to accomplish a number of world firsts in partnership with Sivers.
We've used their array technology to build our 8-wavelength and 16-wavelength laser solutions. We've showcased some of those products at a variety of leading industry conferences. The array technology that Sivers is working on allows us to build a very elegant integration of these laser devices and integrate them all into one chip to really streamline the manufacturing and the kind of scale of manufacturing you can get to and the efficiency of that design. As we add more lasers to that laser array, it allows us to scale our overall bandwidths. There are a few ways that Ayar Labs scales its performance and bandwidth. One of them is by increasing the number of laser wavelengths per fiber, which we call wavelength division multiplexing.
Now, one thing that I'm very excited about that Sivers is working on is taking some of the core innovations that they've accomplished in their research manufacturing environments, such as the fab in Glasgow, and transferring those capabilities over into a true high-volume, commercial-ready production foundry environment, such as what they're doing at WIN Semiconductors. We think this is going to be very important to addressing the scale and the overall volumes that are going to be needed as these laser array technologies and laser products go into the high-volume AI scale-up fabric. Ultimately, the industry needs products in that market application that are in the tens of millions, maybe hundreds of millions of units per year. We're incredibly excited about what's going on there. Ayar Labs is looking forward to continuing to work with Sivers for many years to come.
Thank you very much and very exciting. The last presentation before Vickram comes back with his closing remarks is Alex, who's going to walk us through the manufacturing and production strategy. Next slide, please.
Hello. Hi, good afternoon. And thanks for joining the call today. My name is Alex McCann, and I am the Managing Director of the Photonics Business Unit. In this presentation this afternoon, I'll provide a brief overview of the Sivers Semiconductors manufacturing strategy that will ultimately allow us to scale the business in a capital-efficient manner. Looking back in the industry in the early 1980s, the common mantra was that real men needed fabs. That indicated a collective thinking at that time that you needed a wafer fab to be in control of your destiny and to be successful.
However, since the introduction of commercial large-scale wafer foundries in the early 2000s, the dominant model in the industry has become what we call an asset-light model, where the users continue to provide product design innovation, and the wafer foundries provide the manufacturing service. As a result, the reliance on captive, CapEx-heavy wafer fabs is reduced significantly. What are the advantages of this model? There are many advantages to moving to an asset-light manufacturing strategy. The primary advantage is the significant reduction in CapEx, given that the CapEx for capacity flex or capacity scale-up is provided by the foundry itself.
To put that in perspective, to scale up a wafer fab from an R&D or a low-level production volume basis to the volumes that Andy mentioned, whether it be tens of millions of units to hundreds of millions of units, is essentially the equivalent of $10 million-$100 million in capital avoidance going to this model. You can result in, ultimately, you can result in the business scaling with a much more capital-intensive structure. In addition, the users of our foundry service gain access to a proven high-volume manufacturing operation with best-in-class yields, best-in-class cycle time, and best-in-class product reliability.
Also, users of our foundry service can tap into world-class mature supply chains and demand management and demand planning business processes with good inventory control with established suppliers and having additional benefits of additional inventory control and cost management. Next slide, please. The Photonics Business Unit is in a transition mode at this point to an asset-light model. We are repurposing the Glasgow wafer fab from being our primary production facility to a model where this facility would be more focused on R&D, on photonics product design, on new product development, on new process development, on pre-production release for volume products ultimately, and will run low-volume early-stage production in that particular facility. The transition to the asset-light model means that we will partner with WIN Semiconductors based in Taiwan. They are focused on high-volume production. They have proven high-reliability platforms.
We will match the technology between the Glasgow fab and the WIN Semiconductors fab. There will be, in the future, ongoing technology transfers from Glasgow to WIN Semiconductors as we develop new process technology and new product technology. Ultimately, WIN Semiconductors has been chosen to become our volume manufacturing partner. Next slide. WIN Semiconductors were founded in 1999. They have been around for quite a long time. They are located in Taiwan. They are currently the largest pure-play compound semiconductor foundry in the world with four advanced 150-millimeter wafer fabs, all in Taiwan currently. As Vickram mentioned, when we talk about de-risking for geopolitical issues, they also have the ability to build facilities in Europe or the U.S.A should that need arise. They currently have one dedicated wafer fab for producing optoelectronic and photonic products. In 2024, they have annual revenues of $540 million.
Typically, their major customers today would be Apple, Broadcom, and Lumentum, so marquee names in the business. Their mantra is that they provide compound semiconductor solutions from RF to lightwave. A very established wafer foundry playing in the compound semiconductor space. Next slide. If you move on to the Wireless Business Unit, they're currently utilizing an asset-light model and have done so since their initiation. They're using GlobalFoundries, who is a leading manufacturer of essential semiconductors for a whole variety of different technologies. They're currently the third-largest silicon foundry in the world with about $6.8 billion in revenue in 2024. Major customers for GlobalFoundries would include marquee names in the industry such as Qualcomm, MediaTek, NXP, and AMD. They currently have 11 200-millimeter wafer fabs in Taiwan with a 7 million sq ft footprint, so a behemoth of the industry.
They have a global presence in the U.S., in Germany, and Singapore. That provides the geopolitical de-risking that everybody in the current political climate is concerned about. We use two of their advanced RF process technology nodes, which are the best in the industry, including 45 RF SOI and 22 FDX. That concludes my talk on the manufacturing strategy. I'll pass it back to Vickram.
All right. That was bringing us very close to the end. It's just me between the rest of the presentations and Q&A. I'm just going to conclude with one slide.
Thanks, Alex. Again, I just want to kind of bring us back to what we started with, right? We have a lot of current investors, potential investors, analysts listening in. Investing in Sivers Semiconductors, it's still the question on the mind.
I want to come back to what I said at the beginning. It's a pretty unique opportunity to participate in two of the industry's hottest global secular trends. I'm going to leave you with the five twos. Two of the hottest secular trends, we are playing in two markets with tremendous momentum. That sums up to a $2 billion serviceable available market. We got two highly different technologies, differentiated technologies. You heard about both, our wireless and our photonics technology, both from our business leaders as well as key customers and our analysts. Here, we see growth in the coming years, also in two waves. That's the five twos I want to leave you with. The potential here is to triple, quadruple the business, or even do better at very attractive margins over the next four to five years.
I want to go back to what Tim said during the analyst section. As a prospective investor, you look at this opportunity, and this opportunity is very relevant to Europe and the U.S.. We believe the company is undervalued right now because it's not yet fully appreciated what we are bringing to the marketplace. This is the time for you to consider participating in the company because within the next three, four years, as we deliver this, there's some tremendous market valuations at stake. It is an attractive opportunity for you to participate in that journey with Sivers starting now. That is my rally cry. I believe it's a very compelling thesis, and I hope to talk to many of the investors that are on the call in follow-ons and happy to provide more details on our plans. Thanks, everybody, for spending several hours today with us.
Hopefully, it was educational and you learned more about the company, our focus, and where we are going next. Now we're going to open this up to some Q&A.
Thank you, Vickram. Yeah. Maybe we can start with a few questions from the crowd here.
Okay. Hi. We are here from Carnegie. The first question, can you describe a little bit about the process that you tried to implement in the company? Now it's on. I hope. Okay. We are here from Carnegie. The process you tried to implement here from idea generation, product development, technology development, and sales. How do you want to see that process going forward in relation to maybe what it has been in history?
Yeah. Going back to one of the points I made in the past, it was technology-seeking opportunity.
It was more about having a discussion with the customer. It was an interesting project. You cannot do it for a certain engineering development revenue. Now that we're becoming more market-aware in our focus areas, now we are looking at it as a cross-functional discipline, which is sales is sourcing all these interests from the customers. We are looking at the IP basis we have created. We're trying to understand what's the broadest intersection of the two, and that being supported by customers who have deep pockets and a track record. That's where the process is going to, which is going to result in more of a standard or near-standard product configurations to a broader set of market customers rather than the one-to-one of the past.
That is really the fundamental tying up of all these cross-functional disciplines and using our market awareness to provide product roadmaps that are more broadly attractive. That is kind of the process in play. It is still going to take a few cycles for us to get there because we are still working through some custom projects. As we get more market-aware, I think that is the process that is going to end up in play.
Thank you. Looking at your two BUs, wireless is already on track for good product revenues. What would you see be the trigger point for photonics also getting the product revenues up? Why is that market not taking off already? Because we have a lot of data centers and a lot of AI already in the space.
Yeah. You are absolutely right. Wireless is on a growth path, and it is going to continue to go.
I've managed businesses in the past where some businesses are growing and supporting other ones that are incubating. Then when they hit the inflection, they are able to support the other businesses. We are on one of those combinations right now. When you look at why has it not yet triggered or why has that inflection already not happened, you look at the current momentum of traditional data centers. They're built out in a certain way. There are large incumbents. There is that momentum that kind of continues to take it. If I were a pluggable guy, let's say I'm in pluggable optics, part of me says, "I got to be part of this future." Part of me probably says, "I might lose content from my pluggables.
How do I continue this journey?" It is an essential tension when you have established incumbents in old architectures, and then we got a new architecture coming up. Early catalysts are the likes of Jensen from NVIDIA, the hyperscalers spend, etc., that are starting to say, "Hey, this is not just some fancy thinking. We're putting some serious money behind this, and we are saying you got to solve this bottleneck." That is the first piece of putting more seriousness into the inflection point. They are not stopping there. They are also in the value chain. They are asking everybody to get ready with their supply chain robustness, product qualification, manufacturing readiness. We see the right early signals and also from the big guys. It still takes time to get to the types of deployment they are talking about. That is why it is taking a little bit of time.
That is why the combination of the businesses we have is allowing us to continue to invest in that inflection point while leveraging the benefits of a wireless business that is growing well. That combination is helping us take that journey to the inflection point. We will be watching it every three, six months. We are going to watch, "Hey, is the inflection point gathering strength? Does it mean it is six months earlier or six months later?" That is the essential tension you have. There is an established market that is doing things a certain way. Then thought leaders are saying, "Hey, the wall is coming. You got to change."
Thank you. Regarding your pipeline,
can you just identify yourself and then?
Yeah, sure. Jacob Benon here from Redeye. Thank you for the presentation.
Regarding your pipeline of customers that you showed in the presentation, I think it said some 30 technological engagements within wireless and 7 within photonics. Considering that you're now being much more picky regarding what customers you choose to work with, I wonder what criteria you use to evaluate which of these potential customers to work with. You mentioned mainly deep pockets and track record, but is there anything else here? Yeah, there are a couple of other things also, right? I mean, not everybody in the top of the funnel makes it to the production line. There's a lot of loss as well. There's a certain conversion rate, right?
Absolutely, they need to be financially strong, and they have impact on the market through either their past or their current thinking.
At the same time, I also am cognizant, and the team's also cognizant that we can't do 30 custom products for 30 technical engagements. We are also looking for similarities in needs. What we might do, for example, I'm not saying this is what's happening, but let's say 30 of them have come in asking for products, and we find out that if we do two products, we could get 18 of them. We might just say, "That's the best return of money because any additional engagement is actually going to dilute us." We might say, "Let's do those two products and make sure those 18 customers are happy," right? That's the mindset change that's happening right now.
It is a combination of those three things it is looking for, "What can I build that services multiple people?" Those multiple people still should need to be financially strong as well as have the strength in the market to take something through to production and put it out there .
Thank you. Another question for me then on the topic of customer engagements. To capture a meaningful market share of your $2 billion serviceable available market, do you feel confident that your current customer engagements, including those in the pipeline that you showed, will be able to take it there?
Yeah, especially with the focus on the types of customers we are looking at in the markets we are in, we are making sure our focus right now is on strategic customers with that type of market presence or market share.
If we get our share of those customers, it does allow us for that type of market share for ourselves. At this point, I would say our focus is on the strategic bigger customers rather than the long tail. At some point, we'll get there, but right now, we need to get enough of our share of mind and share of wallet of the material strategic customers in both of our businesses. The names you saw from Harish or Andy, these are all big names with big market shares or market presence. Final question for me relating to something you mentioned earlier in the presentation. You have made many improvements in terms of your working capital structure for the past few quarters, and many of the current deals that you have announced have included upfront payments as well.
Can you give us some more color about how you have managed to do this in the first place and if there are any other ways you can work to optimize your working capital structure going forward?
I'll make a couple of comments, and then, Lottie, if you have some things to add, please do. Number one, absolutely, it's contract discipline. It's something that we did not pay as much attention to in the past when we took on certain contracts because we were eager to get developments going.
If these are customers with the ability to front cash because they are cash rich, they're financially strong, then of course, there's a discussion to be had on, "Hey, look, for us, it's important for us to get our working capital to get these things done." We use that as a basis to sit with our customers and put down contracts that make sense for both, but also allow us not to get in a negative cash flow situation within projects. That's one just block and tackle discipline. The second thing is not ending up with too many customers that have cash flow problems on their own because that then becomes a trickle effect onto us. That's another type of filtering discipline as well, right? Those are two things.
Of course, we are very careful in how we take on equity and debt financing as well because some of these things take a little while even with the right contract structures. In the meantime, we want to have that right partnership with our financing agencies. It is a combination of those things. Of course, adding more and more product shipments allows us for repetitive business, which again improves our cash flow situation going forward. Meaningfully shifting to product shipments, that is going to continue to remain a huge focus for us. Lottie, do you want to add anything to that?
No, I think it was perfectly described.
Okay.
That was all for me. Thank you very much.
Thank you. Thank you. We will take some questions that were submitted online. Martin says, "Thanks for a great presentation, compelling case. What about your—" Hang on a second.
What about your cash position, credit facilities, etc., and the current and future burn rate?"
Lottie, do you want to take that? Do you want me to?
No, I mean, I can build on what you mentioned in terms of managing cash and working capital. I mean, we ended Q1 with a cash position of SEK 73 million. We are not going to come out with any guidance today or new information on our cash burn. As we showed here in the presentation, obviously, it is a focus area that we have. We, over time, have been moving in the right direction the whole time. The first step is obviously to be EBITDA positive, which we have shown we can be for longer periods of time, and then becoming cash flow positive. The discipline in capital and CapEx expenditure and improving working capital.
We have a good plan. This is also a business with large projects and large in and outflows in those. To as much as possible balance those flows will obviously minimize our need of capital.
I would also add on top of what Lottie says, just three things to keep in mind. Number one, as we grow the top line, which we are on track to continue to do, the top line takes care of the bottom line as we go along. Because it's not like we're inordinately adding resources and putting a lot of below-the-line spend on ourselves. As our top line expands, this is a problem that auto-solves, number one. Number two, as I mentioned, we are being much more disciplined about our cash flow.
You guys have seen over multiple quarters, while it's never a straight line, our adjusted EBITDA metric continues to improve, which should show you confidence that we are taking strong control of how our spend is and how we are managing our financials. As top line grows and we continue to exercise this fiscal discipline, this problem auto-solves. We are now on the cusp of turning those corners in the coming quarters. That's really our plan.
Okay, thank you. One for the photonics business. When do you think the market for photonics wearables tech could arrive from SAMI?
It's a fair question. The question basically talks about, just to repeat, when will photonics take off in wearables? These things involve humans and large-scale human trials, especially if you're trying to solve tricky biometric challenges.
In terms of what we can provide as hardware platforms, physical platforms for our customers to go then conduct these long-term large-scale trials, we have done a lot of things already for interested customers. Now it depends on how those trials go because those trials involve not just the massive field trial population, it also involves tremendous algorithm developments, etc., that go along with it. Again, as I say, it's one of those outpost markets where we make sure we're in the fold, we have provided the hardware platforms, but after that, it's one of those long journeys that has to happen in the world of biometrics and wearables. As we know more, as our customers choose to tell us where they are, we will bring information to you guys. These are always long trials. That's one of the reasons why it's an outpost.
It's not a focused, expansive market opportunity for CEOs right now. Okay. Jonathan, can you compare and contrast your beamformer with Anokiwave and Renesas? Do we have Harish online? Yes. Harish, did you hear the question? Did you get the question?
Yeah. Yeah. Okay. Yeah, absolutely. Yeah, happy to do so. So that's a great question because you're highlighting two of our biggest competitors on the market. Anokiwave, of course, has recently been acquired by Qorvo. And so that's now a part of the Qorvo portfolio. As I mentioned, they are actually contained in that comparison chart, although I anonymized the name and sort of used a proxy of the semiconductor technology that they use. In general, our beamformers have superior output power and superior efficiency to the competitive beamformers that you mentioned. Superior efficiency is fundamental.
Even if you have more output power than the competitors, for instance, one can use a larger antenna array using a competitor beamforming IC, so just more antennas, and that way you gang up more of the power. Of course, the solution would be larger and cost more. When you have better efficiency, even when you do something like that, the DC power consumption, the energy that you draw from the wall plug would be lower and less heat would be generated in the solution. There is no way to work around that. As a result, we find that whenever we're in competition with such folks in any new request for quote or request for proposal from a customer, we're always the ones that are selected over these competitors that you mentioned. The ALL.SPACE RFQ dating back four years ago was an example of that.
The newer RFQs that we've received, including the one from the 5G tier one Telco infrastructure provider, is just another example.
Okay, thank you. What other companies are developing laser arrays similar to your main photonic product? Can you compare and contrast yours with theirs?
Yeah, once again, they don't let me in the labs of these competitors. From everything that we've had with respect to our conversations with our customers, etc., when it comes to the power levels that our customers need for their architectures, we seem to be the only ones that are capable of providing laser arrays in the formats they want at the power levels they want. There may be a company or two which does laser arrays, but they are at much lower power levels. That's what we've heard.
From a power level and an array format, we seem to be leading. We never try to take our eye off the ball. Of course, people will be willing to catch up. The other thing to note is this market is not just for CEOs. It is too big for it to be having only one provider of laser arrays. There is space in the market for two to three vendors. Each of them will work on their own to get their roadmaps up. Of course, having the lead allows us to go and seed and deploy from the early moments onward.
Okay. Another question from Scott. At OFC, CPOs continue to gain momentum, and CPOs is a key element of driving that adoption. Once commercialization is reached in 2027 and 2028, how quickly do you expect the transition from LPOs to CPOs?
What kind of share do you expect exiting the decade?
Two things to note. LPOs are today talked about from a scale-out perspective, which is connecting the switches, going from pluggables to LPOs, etc. One thing to remember when Ayar Labs CEO Mark talked about scale-up, scale-up is all about connecting many GPUs to many GPUs. The volumes there will be much higher than scale-out architectures, which is switches and LPOs and CPOs. The better way to think about it is not so much as to how much of LPOs will convert to CPOs, because that is still the smaller volume element, but it is more about how much copper will migrate to optical in the GPU to GPU connects. That is kind of where we look at different scenarios. I mean, nobody has a crystal ball.
But even at a 5-10% conversion rate, by 2028, 2029, you're talking about that billion-dollar SAM I'm talking about on light sources. And then anything on the scale-out front could be 10-15% on top of that. So the reality is think about it from the framework of the many to many GPU connections and how much of copper will move to optical in the future deployments. Okay. A more straightforward one. Isn't the CHIPS Act ended by President Trump? This is a common worry or concern. I want to give you, and again, I'm not sat inside President Trump's office, but I've really tried to piece together what is he trying to achieve. Here is my view on it. I do think there's a lot of substance behind this view.
To do that, let me do a little bit of education for the audience. The CHIPS Act has two pots of money. The U.S. CHIPS Act has two pots of money. One of the pots is tens to hundreds of billions of dollars for companies to incentivize them to build factories in the U.S., like a TSMC building a factory in the U.S. or an Intel building a factory in the U.S. That is one pot of money. The second pot of money is in the tens to hundreds of millions of dollars. That is all about, hey, if I have promising technology in the U.S., I want companies to come together and commercialize it so that the U.S. still has technology leadership for newer generations of products and solutions. There are these two pots of money.
The pot that President Trump has a challenge or a problem with is this big pot of tens to hundreds of billions of dollars. His primary challenge is, you know the opportunity is there in the U.S. You are companies with a lot of money. Why am I subsidizing you to build this factory in the U.S.? You should put your money and build it here because I have business to give you. That is where he wants to squeeze those companies like TSMC or GlobalFoundries or Intel saying, why am I giving you tens of billions of dollars? You make enough profit. You go build a factory. That is what he is challenging. We have not yet seen any signs of him saying, I do not want the U.S. to stay ahead in technology. I do not want U.S. technology to go into new products. He actually wants that.
This pot of money, I haven't seen any pressure on. Guess which pot of money we are getting our U.S. CHIPS Act funding from? I'm not planning on building a factory like TSMC in the U.S.. I want to commercialize technology that's relevant for next-generation solutions. Hopefully that clarifies that it's not that the U.S. CHIPS Act is dead. He wants companies to build factories to use their money, but he still wants to commercialize promising technology that's available in the U.S.
Okay, thank you. One round for photonics. Don't you think that a high one-digit market share is low if you consider the position CEOs have in the AI photonics business? We are talking about timelines, right? When I look at the next four to five years, I'm also looking at who are my competition, right?
Yeah, the future could be rosier than that.
We also have to have a plan first we deliver on. Then there are upsides to the plans. Some people might think it's conservative, but I want a place where I can go execute too. If we can build upside on that, we'll happily take it. Okay, very clear. Did you provide any revenue expectations for 2026 and 2027? Yeah, we don't provide forward-looking guidance like that. I think on the slide that you're seeing on the screen, I've given you my view of what the CEO's potential is, given the size of our teams, given the number of projects we can take, given the number of products we can release as to how big this company's business can be in the next four to five years.
I'm sure for the mathematically inclined, you can go in there and calculate whatever your growth rates are, etc. This is just my view of the CEO's potential that I'm showing here rather than the company issuing guidance in the outer years.
Yep. Do you use your new share issue money now and save parts of the loan facility for later, or do you use and pay interest for the entire new loan from the start?
The loan that we have taken, we pay interest from the start.
Yeah, and maybe just to clarify, it's a refinancing. It's not a new loan. It's replacing an existing loan that was coming to an end in May.
Exactly. That's a very important point because I don't want the audience to think we're just stacking on debt financing. We are refinancing the debt line we had.
Lottie and I will continue to work on the right balance of equity and debt financing we will use in the company. As you have seen in the press releases we have put out, the new loan facility that has been used to refinance the debt that came due is a very attractive loan facility. It is with a bank that is fundamentally aligned with the long-term view we are taking. That is why this is a three-year facility. It is a discussion among like-minded partners on, okay, how do we adjust this in the outer years? What do we move up? What do we move down? What do we want to do? It is a strategic debt financing partner we have right now that we have used to refinance the debt that came due. A very important distinction.
The next one is from Scott.
It sounds like Cloud Chaser and Summit Gen Two drive the wireless opportunity within the next two years. Within Summit, how quickly do you expect volume to ramp? Is it specific to any particular geography? What is the pricing?
Harish is there, but I'll just make a couple of comments and then I'll ask Harish to add. We do not discuss pricing on our part. That is not something I can give guidance on. The product that is to be released in 2026 has attraction in multiple geographies. Of course, the way it might roll out is depending on which geography comes up first and which geography is next. That product configuration from our customer is seeing wide interest across geographies. Harish, is that a fair statement to make?
Yeah, that's correct. Cloud Chaser and Summit, you're correct that they are broad market products.
That's the key thing. They're not custom products developed for any specific customer. A lot of the customers that are in our pipeline on the SATCOM side, as well as on the 5G side, are either using Summit today in the sample phase and are building a product that will ramp to production or are awaiting Cloud Chaser so that they can start designing a product around it. Indeed, the SATCOM frequency band that Cloud Chaser is addressing, the Ka-band, is pretty widely standardized across the world. There is no difference between the U.S. versus Europe versus Asia, etc., in terms of Cloud Chaser's ability. Indeed, as I mentioned, SATCOM has a strong geographic aspect to it where every region and every country is investing in local SATCOM technology as well as SATCOM networks.
Similarly, in 5G, we're talking about fixed wireless access. This is a product that would be used towards both base stations and CPEs by the leading customer as well as several others. These are products that are designed to cover the entire 24-29 gigahertz band, which would then make them applicable in Europe as well as in India, in Japan, and all of the other regions where we're starting to see fixed wireless access roll out in 2026 and beyond. Yeah. Not to forget, we heard from the ALL.SPACE CEO that they're seeing tremendous traction in their pipeline as well. In terms of revenue growth, there will be contributing elements from that side as well in addition to these broad market chips. Something to keep in mind.
Okay. In the presentation, we talked about these outposts markets.
Question is, do you think these outposts markets have potential to wake up in the near future?
The very reason why I maintain the outposts is because I'm looking for any such strong signals, but with very carefully orchestrated engagements. If any of those suddenly the spigot turns on, then of course we look at, okay, how best to take advantage of that opportunity. We have been very calculated about that. Each of those has the potential to become something strong. Otherwise, I wouldn't have picked it as an outpost in the first place. That is kind of the mindset we have is, how do I spend the least amount of money in the outpost while still getting a good read on what's happening in that market?
The moment I know that something is big, something's about to happen, we need to figure out how to quickly make sure enough resources are put there to get our returns on it.
Okay. Is there any news on your old 4D LiDAR customer, FMCW? Are you still their number one supplier for future products or products? And do you sell anything now? And are there products on the market yet, or is it still in the future? Please tell us more.
Replay the name of the customer again here. Is it 4D what? 4D LiDAR? 4D,
your old 4D LiDAR customer, FMCW.
Yeah, at this point, as I said, I've shown you what my focus markets are and what my outposts are. By definition, what's not there is not in a relevant portion of my focus. That's how I'm going to answer that question.
Because the moment I keep getting asked, is there a seventh market? Is there an eighth market? Is there a ninth market? We go back to the old days of trying to do everything for everybody. What I speak about is where the action is right now. That is how the audience has to think about it.
Can you please review the exact amounts and terms of your strategic debt partner? I think the information that we wanted to share with the audience has been shared through a press release yesterday, along with the original press release. That is the information we are sharing that gives basically all the details about the loan facility that we have taken on. What basement companies are supporting this DC standard? Can one run 256 QAM on a virtual modem, or is hardware required?
Harish, you might be best positioned on this.
I think the question, if I boil it down, if you move to the DC-based remote radio approach, does it run into a bottleneck at a certain size of modulation or something, or is it usable?
Yeah, that's what we've been drawing. Yeah, no, that's a great question. It depends on the channel bandwidth that you're trying to run on the DC standard. I think it's in our press release, but the digitizers that we are building are supporting 250 MHz and 500 MHz, single channel and dual channel. At the extreme end, at 500 MHz and at an extremely high constellation depth, I need to check actually if 256 QAM is that depth or it is beyond 256 QAM, rather at 1024 QAM.
At that point, there is one extreme end of the highest bandwidth and the highest constellation depth where one runs into data rate issues on the DC standard. I mean, the beautiful thing about the DC standard is that with this virtualized modem, one can now sort of address a very wide range of applications. There are certainly a very large number of applications where that extreme data rate is not required. For the configurations that we are building, there is a very robust set of applications that will benefit from that.
Okay. Thank you. Two follow-on questions on the debt partner. Why cannot you release the name of the lender and when will you draw the money?
It is something that is a conscious decision we have made. At this point, we are not ready to share who the lender is. As I said, it is a U.S. bank.
It's not an investment bank. It's a traditional bank. What was the second question?
When will you draw it?
Lottie, maybe you should talk about the...
It has already been drawn and we used the full amount to repay our original debt.
Again, going back to the whole idea, this is refinancing the debt, which means what we had has been taken away and replaced by this. It's pretty obvious where that money went.
Who are the other laser array providers by name? Can you comment on them architecturally?
I think we talked about this already. As I said, there might be one or two other players.
As I said, I don't get into their kitchen to know who's offering it, but nothing off the power levels and the array formats that we have been asked to deliver by our customers, we have seen from the others. Can't say much more than we haven't seen anybody else provide in the formats and power levels that we have shown.
Okay, Vickram, has there been any interest from other companies of a buyout of Sivers lately? Can you comment, sir?
I can't comment on any of those things.
Okay, a technical one. Can you discuss the technical capabilities of your various products? For instance, some of us know what Anokiwave provided in terms of both Ku and Ka-band beamformers. What bands do you support and how large a matrix? Similar for your digitizer, how large a band at what sample rate? And who is the specific competition?
I think Harish addressed both those questions. He talked about the DC bandwidths that we are supporting. That was already mentioned. Harish showed the roadmap where we have Ka-band products. We are bringing along Ku-band products as well for our beamformers. In the future, if there are other bands that are needed, we have the capability to. Right now, our focus is Ku and Ka-bands, which is where most of the action is.
Very clear. When will we know the name of the second photonic customer? Similar to Ayar Labs?
When I feel comfortable where the partnership goes and then I can talk about it and the other customers are also willing to talk about it.
I am at the end of my list of questions. That was a good bunch.
We had questions from the live audience too. We are at 5:25, so it went a little bit longer, but I hope that all the sessions were very useful and there was a lot of information that we provided. Happy to take any follow-on conversations with analysts, investors as the days go by. Thanks again for attending our Capital Markets Day. It was extremely invigorating for me to have this audience and talk to them along with my entire team. We thank you all for taking time out of your busy days and spending it with us here. Thanks, everybody.
Thank you.
Thank you.