Welcome to the Raspberry Pi Holdings plc Half-Year Results Investor Presentation. Throughout this recorded presentation, investors will be in listen-only mode. Questions are encouraged, and they can be submitted at any time by the Q&A tab situated in the right corner of your screen. Just simply type in your questions and press send. The company may not be in a position to answer every question it receives during [the meeting itself]. However, the company can review all the questions submitted today and publish responses where it' s appropriate to do so. Before we begin, I'd like to submit the following poll. I'd now like to hand you over to Eben Upton, CEO. Good afternoon to you, sir.
Good afternoon. Welcome everybody to our first half 2025 interim results. I'm Eben Upton, Chief Executive of Raspberry Pi Holdings plc. I have with me Richard Boult, [crosstalk] our CFO, with me.
[crosstalk]
I hope many of you know who we are and what we do. We build computers. We're based in Cambridge in the UK. We've been building and selling ARM-based embedded systems since 2012. We design our products in Cambridge. We build the vast majority of our products in Wales. It's about 140 of us, almost all located in Cambridge, exporting, manufacturing between 7 million and 9 million Raspberry Pis a year. We've made very nearly 72 million of them in the 13 years that our products have been on sale. Roughly 95% of the product that we produce is exported out of the UK. A quick summary of the highlights of the first half. It was a good half.
T he full year, we communicated that we'd seen a recovery in demand from both our existing and new OEM customers, and a normalization of the channel inventory environment. Last year, we had a period of elevated inventory and channel and our OEM customers, which was a significant drag on the performance of the company in the second half of last year. We saw a 9% sequential increase in unit volumes. That's , as I say, underlying demand growth from existing and new OEM customers. An 8% sequential increase in direct units and a 27% increase in direct revenue. I think that speaks to the success of our ongoing strategy to move a larger number of our large OEM customers into direct relationships with us.
A favorable mix of products in terms of unit economics, certainly a better unit economics mix than we saw in the second half of last year. Together, these delivered a 19% sequential increase in adjusted EBITDA. To address a couple of potential headwinds in the market, obviously, we are now in a high and dynamic tariff environment in respect of our sales into the United States. We've yet to see any significant evidence of impact on U.S. demand for our products as a result of those tariffs. Indeed, we believe that in the medium to long term, definitely we will benefit from a differential competitive advantage over many of our competitors whose products are manufactured in Far Eastern venues that are subject to higher tariff rates.
DRAM pricings, t here has been a lot of noise recently around DRAM pricing, particularly around the LPDDR4 and LPDDR4X memory that we use on our Raspberry Pi 4 and Raspberry Pi 5 products. We entered the year with substantial inventories of LPDDR4 memory. We remain actually well hedged into lower densities. We remain well hedged into 2026, through the end of the year and into the early part of 2026. We do have some exposure to market pricing, to contract pricing, not spot pricing, but to contract pricing at higher densities. There is sufficient supply on hand and on order to meet our needs through the end of the year.
Further out, if we continue to see elevated DRAM pricing, which we expect we will do in the first half of next year at least, we have a number of technical and commercial levers that we can pull in order to mitigate our sensitivity to DRAM pricing. It was also a good half for product launches, where last year was a year of platform refresh. We refreshed both large Raspberry Pi platforms, Raspberry Pi 5, Compute Module 5 and Raspberry Pi 500, and also the microcontroller-based products with the launch of RP2350 and the Raspberry Pi Pico 2 series on single-board computers.
This has been a year I think of sustaining innovation. We'll say a little bit more about that later in the presentation. We'll also talk a little bit more about the progress on other strategic programs, including the board-to-board initiative and the progress that we've been making in our second franchise, the semiconductor business. [ I'll go to] Richard Boult for some [financials] .
A financial review of first half of 2025. The graph on the left-hand side, which I think is just important context before we dive too much into the comparison of the various halves, this looks back to the beginning of 2024. It picks up on a quarterly basis, our unit sales. That's the total height of the bars. The gray at the top is really the sum of our Pico and Zero products. These are our generally low-cost boards. I mean, Zero, between $10 and $15, Pico's $3, $4. Below that is our major boards, single-board computers that I think most people would know, and also compute modules. Coming down that, the blue there is the Pi 5. That's the board that we released just at the end of 2023. Volumes coming into 2024 were still very much launch-related.
The green is what we call the Raspberry Pi 4, continuing to sell significantly, despite the new Pi 5. That's really a function of this industrial demand that we'll talk about elsewhere, that 70%- 80% of our boards are actually used by companies either to build into products that they sell or in their own factories, in their own laboratories. Very much a demand that actually is quite sticky because of that. The orange, Raspberry Pi 3, that's something that we released in 2016 and we're still selling pretty significant volumes, which are coming through. Compute module, which is the credit card-sized version, if you are building something where you are limited by space, that contains all the Raspberry Pi-edness that you need of a Pi 4 or Pi 5, and therefore you can fit it into a product. That's very much an industrial product.
As you can see, coming into 2024, Pi 5 volume is very high initially on the results of launching the product and all that launch demand. The compute module volume was also high, because in 2022 and 2023, there had been a period of really quite chronic supply shortages, and we had seen an awful lot of back orders finally, the last part of which was fulfilled in quarter one 2024, really for compute modules. You have a very good first quarter for those reasons. What came out of that at that time was an awful lot of our people in channel actually consumed, added large quantities of inventory, indigestion, which is something that you may have seen right across the electronics sector really through 2024, an awful lots of inventory build and channel stockpile.
I think the nice part is really into 2025, we have seen that now back to a normal level of demand, that industrial demand, which is to a first approximation, is the part under that red line, the sum of those, Pi 4s, Pi 5 CMs, et cetera, after it dipped from early on, has now risen, has continued to rise through the period. I think that has been the sort of strong theme that we want to get over to people really in the first half of 2025, is that recovery in demand back to much more normal- paced levels, almost to the point where we're seeing a little bit of order backlog in some cases as we step up our production to match that. Key metrics that we talk about, unit volume in the first half is about 3.6 million.
That compares to 3.7 million a year ago, but as you can see, quite a bit has happened into the intervening period. In half two, we did about 3.3 million. Very much that 3.7 down to 3.3, up to 3.6 trajectory. Gross profit per unit, key metric as far as we're concerned, the profit that we make on each of those boards, $8 in the first half of 2025 compared to $8.3 in the first half of 2024. Talk about that a little bit more on the next page. That comes together to give us a gross profit of $33.2 million, together with some profit that we make on accessories that are attached to those boards and also on semiconductors, which I think Eben's mentioned and we'll talk about more. This is where we sell chips specifically.
Overheads were generally well-controlled through the period, giving us an adjusted EBITDA of $19.4 million, about a million lower than it was a year ago, as a result of still coming through that very strong period at the start of 2024 before the volumes had picked up. Operating profit, which is after depreciation and amortization, amortization has increased by about a million after we released a product in the summer of 2024, which increased our amortization on that development work, and hence has impacted the profit by the amortization of that. Finally, cash, we finished the half at $34.3 million. That's down on the $45.8 million that we finished last year. Again, we'll talk about that on a subsequent slide. Units by half. We talked at a very high level about units by quarter. We also sell products through [tourists]. We sell them directly. We sell to OEMs ourselves.
We also sell them through a network of resellers worldwide. There's about 114 of those across 75 countries. They are our sort of virtual sales force in many ways, a mixture of people from big distributors down to mom and pops who are huge advocates of our brand worldwide. We also sell through our licensee, Premier Farnell, which is the darker R there. As you can see through the period, we had strong volumes back at the end of 2023, with the launch of Pi 5 into the first half of 2024. T hat gradually attenuated on the licensee channel. The volumes that we're selling on the direct side have continued to appreciate through the period, half- on- half, the part we sell directly. Gross profit per unit was $8.30 per board, each of those boards we just talked about in the first half of 2024.
We made a further $1.10 of profit from accessories per every board we sold. That dipped in the second half of 2024. We were selling particularly a product called Pico, which was newly launched, Pico 2. That bolstered unit sales, but at $3, $4, as you can imagine, the dollars per board was really quite slight. Whereas in the first half of 2024, we sold a lot of Compute Module , which was a pretty high volume value board with good margins. Into the first half of 2025, we got to $8 per board of profit, down slightly on a year ago. Two factors there, less of the Compute Module volume that we'd seen in the first half, offset by the fact we're paying now $15.98 for the Broadcom chip that goes in the Pi 5.
Whereas for the first 2 million boards in 2024, we were paying $5 more than that. We've seen, I guess, the cost of the board come down, but the mix of other products hasn't quite caught up with that. Overall, we finished at $8 of margin a board. In addition to those, we were selling accessories, as we've said, which was $4.1 million of margin, $1.10 per board. Also, we sold semiconductors, the record number, which actually was I think about half a million dollars of margin as well from those. Overall, rather a lot of numbers on this slide, apologies, but g ross profit was down slightly in terms of what we made, so down to £ 33.2 million from £ 34.2 million. Administrative expenses, we saw a small increase. We saw a little bit more headcount. We had some additional costs from being listed.
That was moderated slightly by some foreign exchange gains that we saw on the various sterling receivable balances that we had as sterling appreciated. That gave us overall an EBITDA of £ 19.4 million for the half compared to £ 20.9 million last year. Amortization, talked about that, was up about a million. That gives you an adjusted operating profit of £ 13.2 million for the first half. Below adjusted operating profit, we don't include in that number share-based payment costs. We've had two new schemes since the IPO in June 2024. We had a grant of options in June of 2024, and we've had a further RSU grant this May. The RSU grant is much more the standard grant that we would expect to make going forwards.
The options was very much something very much linked to the launch of the IPO, the successful listing of the company and really making sure that all of the people who came through that process, having received the crystallization of their [alto], continue to be motivated and participate in the success of the business. Below that, financing costs did rise. They're up by about a million. It's a combination of some foreign exchange losses. Again, sterling appreciation. We have a couple of leases, which under latest accounting standards, we have to capitalize. Therefore, we had a sterling liability, which incurred a loss, a non-cash unrealized loss of about £ 600,000, £ 700,000. We also have some long extended payables, which again under accounting standards, we have to discount. There was an unwind of the discount over the period of about a million in there. Tax was down.
Tax was down, partly slightly lower profits, but also we had a foreign exchange gain. This is becoming a bit of a theme, but we did see quite significant sterling movements in the period on a tax receivable that we had at the end of the year. We received R&D expenditure credits in addition to paying corporation tax in the usual way. We do do a lot of R&D. Those credits are paid a little bit, or paid this time a little bit after the year-end. We had a receivable, which again, the revaluation of that gave rise to a gain, which is why the tax rate that you can see here is actually quite low at 13%, when the true underlying rate is actually pretty much 25% corporation tax. Earnings per share, we particularly focus on adjusted basic earnings per share.
The adjusted profit after tax was down by a couple of million, £ 2.5 million. A combination of the EBITDA we've talked about, finance costs, share-based payments being higher. We're paying national insurance these days as a listed company, whereas the original scheme pre-IPO very much revolved around capital gain structure and t hen we've got a reduction in tax costs. Overall, profit after tax was down. The number of shares in issue is pretty much 193 million now. The number a year ago is, obviously up to June, i t's really the pre-IPO volumes. It's therefore a very unrepresentative comparative. At IPO, we issued about 10 million shares as part of just over 10 million, as part of a 40 million primary. Also, the crystallization of the LTIP as well moved that significantly. I think 193 as a denominator into the future is realistic.
The numbers for last year are difficult to parse. Just to talk about our balance sheet, particularly inventory, we've held that flat from half year last year through the year-end to now. The mix within there I think is quite significant. At June 2024, we had a lot of finished goods. As we talked about on that first slide, there was a slowdown in volumes, quite significant and quite rapid. We had production that we then had to work to reduce and wind down. Through that period, we had an overhang of inventory. We [very much] said we would take steps. We were able to work closely with our manufacturing partner, Sony, to slow down that production, which was successfully done, so that we've come down with finished goods at around £ 40 million for the end of this half.
That's probably about as low as we can realistically go, while meeting the demands of customers. As I said, we have an order backlog at the moment. There's a possibility that might dial up by a few million as we get into a slightly better, more liquid state on our finished goods inventory. Through the period, our component inventory, the other parts, so these are the parts that we sell that go into Raspberry Pis. We sell them to Sony. It's really a combination of memory and processor chips that we buy from Broadcom, that has risen through the period. Memory has stayed broadly steady through this period. We are still covered forwards about a quarter. We'll talk about that a bit more later.
The components, which is very much the Broadcom chips, after the real chronic effect on the business that we had in 2022 from just not having enough of these processor chips, we are quite deliberately under a long-term agreement that we have with Broadcom, building that up. That has picked up to, in total, about 117 million, memory. That's memory and processor chips at June. What we see is that inventory, we think will probably stay at these levels now through this year-end and into next year. It may even come down a little bit next year. On the other side, we have payables that are aligned or allied to those particular component purchases. At the end of the year, that was about £ 96 million a s you can see, and is now about £80 million.
That reduction is really the thing that has impacted cash flow this year, not the inventory levels, which I think my colleague Mike Buffham has really controlled very tightly. Within that, there is obviously what we call standard payables and t hen there are extended payables. These are where we've had people offer us, particularly as an inducement, really to hold more memory, to hold processor chips at certain times. They have given us extended payables. We have the benefit of some £52 million at the end of 2024. That has come down to £22.9 million now. That's what's caused really the working capital outflow in the period. That £22.9 million, we believe, will come down to round about zero by the end of the year.
Payables by then will be much more on a clean, round about 30 days- 45 days of cost of goods sold, having paid off really those extended payables by that stage. We do see a bit more deterioration in the cash by the year-end as a result of that. Clearly, once that's reduced down to zero, we don't see any more. Going into 2026, what we see really is with that inventory steady and with the payables fairly steady, we shouldn't see any great accumulation into working capital. It should be quite negligible. Therefore, we would expect over the course of 2026 particularly , we'll spend some on CapEx. We continue to spend about £20 million a year on CapEx to build new products. We'll continue to do that. That is very much the essence of the business.
We'll pay a little bit of tax, but the remainder should really drop through as cash. We should see cash start to improve. We probably already described all the key elements on that cash flow. Just to summarize, we finished the year at £45.8 million of 2024. We made some profit. We spent a little bit on inventory. We spent more on payables. You can see about £18 million. CapEx for the half year was £ 9.5 million, very much in line with that £ 20 million for a full year, and bringing us really by the end of the year to £ 34.3 million. As I said, I think that will drop down a little bit further because of really removing the last of those extended payables offset by the natural cash generation of the rest of the business.
That should bring us down to a cash number at the end of the year, somewhere in the order of, we have various analysts out there with £ 15 million- £ 20 million numbers we're comfortable with. A little bit of some cash reduction before seeing really a return to positive cash flow in 2026, and therefore some increase in cash over that period. Just to touch on our business model and the key components within there, our revenue model, we have really three ways that we get to customers. We sell to some customers directly, and typically we have a margin, say of $15. These are sort of platonic numbers, but to give you a sense of the ratio of profitability between each of these routes to market. A significant majority of our sales however go through resellers. This is this network of about 100 people worldwide.
We generally make a margin of about $10 when we sell to them, before they sell to the end customer and they would pick up the $5 differential. We also have this partner, Premier Farnell, who make and sell about 25% of our unit volume. In that case, it's $5. More importantly, it's probably that ratio. It's one to two to three in terms of the margin that we make. Clearly, as we move between those channels to customer, that is one of the drivers in the background of the improved profit per board that we've seen over time. The other parts of unit economics in terms of our boards, costs, we have processor chips that we are buying almost entirely from Broadcom. We have DRAM, so memory that we're using in those processor boards.
That , we'll touch on later, is obviously something where there has been some fluctuation in prices and other components. This may be diodes. It may be the printed circuit board, the green board on which all of our SBCs and compute modules are based, and capacitors and diodes and a lot of the USB cans, et c, other components. It's an area where typically we will have more than one supplier, so we have some pricing pressure that we can control. W e have manufacturing, which is almost entirely through Sony and we control the relationship with them. How do we move the margin up? What's the margin upside? As we move up memory densities, if somebody buys a Raspberry Pi with 8 GB of memory compared to 4 GB, so it's fair to say we do make more profit on that higher memory, very similar to Apple.
Custom boards, we do design some boards for customers. Frequently, when we do that, they get a price reduction, but often to date, every one of those that we've done, we make more profit as well because clearly the board contains just what they need. It doesn't need to contain everything else. We have seen a way of appreciating volumes there. Variants, what do we mean by this? Obviously, we've talked about the classic variant, which is more memory, but also you can have eMMC memory rather than having a plug-in. Also, you can have things like temperature, so higher temperatures. We also sell accessories. We made about $1.10 for every board in the first half. We made a margin on accessories. That can be things like power supplies, screens, AI accelerator hats, a number of things.
What we are seeing rather nicely is on things like power supplies and particularly keyboards. People are buying those in their own right now as a piece of quality engineering, rather than just as an attachment to a Raspberry Pi. We're seeing a step up in those. Overall, there's a lever in terms of managing the economic switches. We have scope to increase, to change pricing as appropriate. Sometimes we've changed it down in the past.
We have increased it. It's not something we do lightly. We are very much focused on that long-term establishment with customers. That predictability is important to us. We are clearly focused on making sure that Chinese competitors do not come in if we overcharge, and get a toehold in the market. That gives us a chance to meet our objective of growing gross profit per unit over time. I'll p ass it back to Eben there.
Good. Okay, I 'll give a quick update on the progress of some of our strategic activities. As I said, this has been another strong year for product launches at Raspberry Pi. Very different character of 2025 versus 2024. In 2024, we refreshed our core platforms, Raspberry Pi 5, Compute Module 5, Raspberry Pi 500, the big end, and then Raspberry Pi Pico 2 and its variants, powered by the new RP2350 microcontroller. 2025, in contrast, is really a year of sustaining innovation. It's about identifying gaps in our product lineup, about addressing specific pieces of customer feedback and delivering, finding places where we can deliver incremental improvements to our existing products. I've got a couple of examples here on the slide.
Extended temperature Compute Module 4, historically, our Compute Module products have been commercial temperature-grade products rated from - 20 degrees to + 70 degrees. By extending that temperature range out to the range of- 40 degrees to - 85 degrees, we expand the range of target applications to the platform. We had customers who wanted to design with Compute Module, but were unable to because of the lack of an industrial temperature-grade variant. It's also a good example of a non-density-related way in which we can charge more for our products. When we provide people with that facility, we can charge more money and it delivers incremental margin to the company.
The second example, Radio Module 2, we have a lot of OEMs designing prototyping roughly with our wireless Pico variants, Raspberry Pi Pico and Raspberry Pi Pico W and Raspberry Pi Pico 2W, and they're wanting to go to scale as direct semiconductor customers. When they do this, they need a solution for wireless connectivity. Radio Module 2 is effectively the radio subsystem for Raspberry Pi Pico W products, but packaged in a way which makes it more integral into OEM products. I think a sort of the tradition at Raspberry Pi is not to talk in great detail about future product launches, but we have a number of things coming up in the second half. We have an imminent major platform release targeting the education enthusiast and enterprise market.
Last year saw the launch of our first artificial intelligence machine learning acceleration products, the AI camera, the AI Kit, and a pair of products called AI HAT. The fourth quarter will see a new product in that product line, a new AI product with enhanced support for more modern workloads, for generative workloads, including LLMs. We have a cost-engineered Compute Module product for the Chinese market, actually launched this morning in Shanghai. It's the first time that we've done a product of this sort. It's the first time we've done a specific geographically targeted product, and it reflects the extreme price sensitivity that we see in the modular computing market in China. Finally, on the software side, we'll be adding new features to Raspberry Pi Connect. Recall that Connect is our product offering, our IoT cloud offering to our OEM customers.
We'll be rolling out a range of new features for the Connect platform, including support for over-the-air updates, secure our over-the-air updates that are compliant with the range of new regulatory requirements that are coming in in the IoT space. The board-to-board initiative is now a year-old initiative. We initiated it in the aftermath of the IPO, designed to reach out to senior decision-makers of industrial OEMs, with the aim of generating individual design wins, which moves the needle. Recall that our typical median OEM customer buys under 1,000 units a year from us. We have a very broad OEM customer base. Really, the goal here is to go out and incubate design wins, which are in the 100,000+ a year range. R eally three elements to this work. First of all, discussing and identifying barriers to scale the use of Raspberry Pi technology.
When we go out to these industrial OEMs, we generally discover quite widespread use of Raspberry Pi in prototyping, in industrial automation, production automation and test automation roles. Some of those OEMs then are also using Raspberry Pis inside their products, but a number of them are stuck at what we would consider to be an industrial role in their embedded engagement. We try to understand why that is when that happens. Often, we'll encounter misconceptions, misperceptions about our business or about our products. Sometimes we will identify genuine gaps in our product offering, which are acting as barriers. The industrial temperature range Compute Module 4 is the first example of us responding to some of that feedback from this program. Often, we encounter purely internal barriers to adoption within these organizations that we can help to address.
We then work at a senior level to develop a shared understanding of where the opportunities are inside these organizations, and to assist in matching those opportunities to elements of our product catalog or to capabilities inside our organization. Finally, we work with the engineering function in the organization to design products using Raspberry Pi, and then to bring those designs rapidly to scale production. I say rapidly. L ike all design activities, these are inherently multi-year efforts, generally 12 months- 36-month efforts. When we do this, I think you'll see us address these opportunities with a mix of standard products and customized products. Historically, we've only provided customers with access to our customization pipeline once they reach a certain scale with our standard products. The scale of some of the opportunities we're seeing in this space means that many of them will qualify on a volume basis.
Many of these opportunities will qualify for some level of product customization out of the gate. What exactly are we finding when we talk to these OEMs in the way of barriers? We're finding organizations which have great domain-specific engineering capabilities, and great enthusiasm for the idea of introducing connectivity and introducing intelligence into their products. I think we do see this series of common barriers, and they're not really just barriers to the adoption of Raspberry Pi technology. They are broader barriers. They are barriers to the addition of intelligence connectivity to products. What are the barriers? All the hardware and software platforms that people need to work with in order to add these capabilities are becoming progressively more complex to work with over time.
Both the hardware and software elements of the platforms in isolation, and also the co-optimization of the hardware and software components of the platform to meet the required price performance point. We also see an increased focus within these organizations on security and on reliability. Some of that is reputational and desire not to put products into the wild, which are going to misbehave when they're there. A significant amount of that is also now regulatory. There is an increasing degree of regulatory scrutiny on the security behavior of connected distributed intelligent products. Obviously, the usual commercial considerations apply, but also some more recent concerns, particularly around supply chain integrity in the context of a more complicated geopolitical and tariff environment, and a continuing and I think increasing difficulty in recruiting and retaining the design engineers within their organization, which is required to successfully land these features into products.
Obviously, we see Raspberry Pi as an answer to many or all of these issues, both through its ease of use, its highly integrated software and hardware solution. We are delivering people turnkey software on cost-effective hardware. Obviously, to that supply chain point, almost every Raspberry Pi product and certainly all core Raspberry Pi products are manufactured in the U.K. As we focus on these potential high-value wins through the board-to-board program, I think it is important to remember that, as I said, our median OEM customer is under 1,000 units a year. It's important not to allow the broad market to wither as we do this. The health of this market really is key. It is the thing that's powered the growth of Raspberry Pi to date, and it will be one of the components that powers the growth of Raspberry Pi over the next decade.
What are we doing? We're investing more in physical events, primarily trade shows. We have a significant presence this year at six trade shows, up from three last year and also virtual events, in particular first and third-party webinars, with particular emphasis on non-English language content. We're making better use of our communication assets now, better use than ever of our communication assets, the website, our various newsletters, our social media reach. We're working with partners, especially in China and the Far East to distribute our messaging in places where there may be language or cultural barriers, which would normally make it hard to address that from head office. Lastly, more scalable early access programs.
If you look at the early adoption curve for Compute Module 5 and for RP2350, and compare that curve with their predecessors, Compute Module 4 and RP2040, you see a much more rapid adoption curve for what's driving that work that we did in the 6 months- 12 months before the launch of those products, to get early access samples into the hands of prospective OEM customers. I think you'll see a lot more of that from us in the future. Some case studies. We love case studies. Trying to do some case studies every time we do some results. They really bring our customer value proposition and our competitive advantage to life. The examples here come from a recent visit to Brazil by Roger Thornton, who runs our applications team in Cambridge.
What's interesting is that , why have we picked Brazil for this? South America traditionally is probably the last market for us to invest significant first-party effort in. Yet as we start to do that, what we're encountering already is quite a fairly well-developed ecosystem of OEMs using our products at the heart of their products. Three examples here. Prodata, they make transport products.
The yellow box you see here is an intelligent ticket gate product for public transport, apparently dealing with challenges associated with people borrowing their grandparents' free bus passes in order to get free rides on Brazilian buses. Safe Farm, Agritech products of precision agriculture, and then Excel fuel station automation. These are very much in the wheelhouse. This is very much the heart. These sorts of applications are the heart of industrial and commercial OEM applications of Raspberry Pi in Europe and in North America.
It's fascinating to see the extent of which already, even in the market in which we are only beginning to apply first-party effort to, there's already a fairly rich ecosystem of people using our products. This is the first time we've included a slide on semiconductor products in a results deck. Electronic products and semiconductors are positioned within the business as mutually supporting franchises. The semiconductor products that we make allow us to make better and more compelling electronic products. The electronic products that we make are then the shop window for the semiconductor products. In the first place, the design engineers who are prospective semiconductor customers encounter our semiconductor products. We have this out. We have an aspiration over the next decade for semiconductors to grow within the business, to the point where they make a roughly equal contribution to the economics of the business.
We're a long way from that to date, but I think we passed a very significant milestone in the first half. This was the first half year in which we sold chips [crosstalk] the first time. Now remember, these are $0.50 chips versus $50 boards. Financially, this is currently relatively immaterial to the business. From a strategic perspective, I think it is important that more than half of the computer experiences that people have had with Raspberry Pi, have been semiconductor experiences as distinct from board experiences. Largely driven by RP2040, the product we launched, t his is an interesting indication of how long the design cycles are in this business. This growth is largely driven by RP2040, the first semiconductor product that we launched in January 2021.
We shipped that up 69% sequentially, with a number of quite large design wins coming through in the half. We have one example here. WIZnet, a very Taiwanese networking product vendor, they build Ethernet chips. What they were able to do with RP2040 was to combine our RP2040 silicon, their Ethernet controller silicon, to generate an intelligent ethernet controller for embedded IoT applications. I think it's a wonderful example. Really, this is the story of Raspberry Pi. I think more broadly, it's a wonderful example of how when you produce a high-quality product like RP2040, you can't necessarily appreciate it, but when you put it in the market, what the market will do with it, you have to remain nimble. You have to remain agile, and you have to remain willing to double down and invest in the applications that the market finds for you.
RP2350, our second-generation microcontroller, obviously earlier on the adoption curve. We launched that in August last year. We put it in general availability in March this year. I think we put it in general availability a little ahead of schedule, really in order to help us to manage the very high levels of customer interest in that product. We shipped an updated revision of this product, so-called A4 silicon, in August this year, addressing a number of issues with the A2 silicon. We're seeing very good adoption. Inquiries from tier-one consumer electronics companies, I think a difference in character of the people, as this is a more full-featured product. I think that's then leading to a difference in the varieties of OEMs, who are considering a design to their products. We've always found that the release of a second product brings credibility to the organization.
Raspberry Pi 2 in 2015 was a key moment in the history of Raspberry Pi, because it was the point where we went from being a company with a point product to being a company with a product line. I think you see that as in many other ways, you see some of those lessons from the board-level products, the electronic products being recapitulated in the semiconductor space. The second product brings the credibility of a product line, and it also leverages the momentum that is built up in the first product. We're very excited, very optimistic about the future of RP2350 and of the semiconductor business more broadly. Finally, to the outlook for the second half and beyond, continued improvement in underlying demand. We do see that slow and steady increase, and particularly in OEM demand.
We left the first half of the year with a significant customer order backlog, and we expect to be able to address the bulk of that in the second half of the year. That gives us confidence in the second half volumes. We have confidence in the unit economics. We're expecting to see a similar product mix to the first half, and therefore a similar gross profit per unit. We have sufficient DRAM supply, much of it acquired at low historical prices to last us through into 2026.
The consequence is no change to our expectations for the full year. Looking further out, continued improvement in the demand environment, continued progress and the first significant contribution to volumes from the board-to-board initiative should materialize in 2026. Of course, the continued growth of semiconductors, albeit from a low base, as a second contributor to the volumes and to the profitability of the business. Thank you very much. That concludes the formal part of the presentation. I think we can start on the Q&A.
That's great. Thank you very much for your presentation. Ladies and gentlemen, please do continue to submit your questions. You can do so just by using the Q&A tab that's situated on the top right-hand corner of your screens. We have received questions both pre-submitted and throughout today's live presentation. I'll start the Q&A session off with the first one here, which reads as follows, w ill the Raspberry Pi Foundation reduce a 70% stake within the next 12 months- 18 months, and what minimum free float level are you targeting?
I think that this is not an opener for this one. Really all I can say here, I guess the Raspberry Pi Foundation stake in the business is about 47%, not 70%. They are a narrow minority shareholder in the business. Obviously, we are not the Raspberry Pi Foundation. Decisions about the disposition of the foundation's residual stake in the business are a matter for the foundation, not for us.
Thank you very much. Next question here. What steps are you taking to embed dedicated AI acceleration into future Raspberry Pi boards through collaborations with AMD or other AI chip vendors? When do you expect the first AI-enabled Pi to reach the market?
I think we have a fairly consistent story about how we think about AI in the core product, which is, we have a conviction that many AI workloads will run well on CPUs alone on unaccelerated platforms. What we've concentrated on doing over the last few years, is to provision progressively more general- purpose compute capability in both the board-level products, the big Raspberry Pi products, and the microcontroller-based products, to allow them to run the majority of AI workloads. Where we go beyond that capability, our aspiration is to have good relations, is to be the best possible host for third-party accelerators and have good relationships with accelerator vendors. At the moment, our primary AI accelerator vendor relationship is with Hailo. The AI Kit and AI HAT products that I referred to earlier are Hailo-based products.
I think that future AI acceleration on Raspberry Pi is likely to be Hailo-based. We're very happy with, very excited about their silicon roadmap. Would we build compute into an AI acceleration into a core Raspberry Pi product? I think it would be very challenging to build it into the standard SKU, because we have to remember that many of our customers don't require that. If you built it into a standard SKU, you would be taxing every customer who didn't want AI for [crosstalk] do. I don't think it's unimaginable that you might see a product with acceleration, that is something not on the roadmap at present. If we did, it would likely be Hailo as the accelerator vendor.
That's great. Thank you very much, Eben. Next pre-submitted question we have here. What's the current forecast for the microcontroller side of the business? The [RP2040] microcontrollers are used heavily in the enthusiast market. How does it look for gaining a big share of the embedded market, and how do you intend to take share away from the big players?
I I think it is a long road. This is a half in which we did 4.5 million units, maybe a full year in which we might aspire to do between 8 million and 10 million units. This is in a 30 billion unit microcontroller market, so there's certainly plenty of room to grow there. I believe we have a compelling price- performance proposition relative to the incumbent vendors. I think you need a compelling proposition in order to penetrate such a well-developed market. I think what we're likely to see is a slow but steady ramp, probably underpinned by RP2040 in the first instance, in the first few years, but we'll see a slow but steady ramp.
Really, the aspiration is to be material, to reach a point where materiality, say within five years and potentially to reach a point where these are two co-equal, if things go well or these are two co-equal franchises within a decade. It really is powered by the price- performance proposition. These are devices that are bringing to the sub-$1, or in the case of RP2040, the $0.50 range, you're bringing into the sub-$1 range performance points and capability levels, which are traditionally the province of $2, $3, $4, $5 microcontroller [platform].
Understood. Thank you very much. We've got another question here. Sales of semiconductors are growing fast, and this is an important initiative for Pi, though the revenue contribution is still small. On what timeline do you expect sales of semiconductors to become meaningful to the company's financials?
I think I probably answered that in the previous question. Yes, I would say in the five to 10-year timeframe would be the right way to think about this.
That's great. Thank you very much.
We do believe that these initiatives are strategically important to us long before they become important to us from a purely financial perspective.
Perfect. Thank you. What magnitude of the business do you see Raspberry Pi in 10 years' time, and is the payment of a dividend an aspiration?
I think the aspiration is to grow the electronics business quite substantially, and to grow the semiconductor business probably faster than that. I don't think any thought has been given at the moment to the prospect of a dividend.
That's great. Thank you. Is there much potential for usage in the robotics space?
I think there is. I think we already see some usage of both the electronic products and the semiconductors in the robotics space. I think a thing that we're particularly excited about is the prospect of pairing Raspberry Pi with acceleration, with external PCI Express attached to acceleration hardware in the embodied robotics space. I think that there really is, if you look at the embodied AI space, the opportunity to add really first-rate AI capabilities to robots of all sorts, particularly robots which are mobile within some constrained or unconstrained environment. I think this fits very nicely into our aspiration to be the best possible host. T hese are applications which will require very substantial amounts of communicability, and our aspiration to be the most cost-effective and lowest power host. These high-performance accelerators marries up pretty well with that particular application domain.
That's great. Thank you very much. The final question we have for the moment, do you have multiple DRAM suppliers or are you dependent on just one or two? Can you pass increased DRAM costs onto customers, or do you have to absorb them?
Historically, we've been a big three customer. We've relied on the larger vendors. Since the start of the year, we've diverted a certain amount of attention to bringing up new vendors, particularly at the lower density end of the density curve. In our LPDDR4-based products, which are the ones exposed to these DRAM pricing dynamics, we have quite a wide dynamic range of density. We are the lowest density products to use one gigabyte of memory. Our largest density products use 16 GB of memory. At the lower end, there is more scope of vendor diversity at the low end of that curve than there is at the high end of the curve. That's really where we've been investing [crosstalk]. Right now, we have approximately three or four active DRAM vendor relationships today.
In terms of mitigations, in terms of dealing with only this temporary price excursion on LPDDR4, we have a number of possible approaches available to us. Obviously, as you say, one of them is price increases. There is a point, I think at which it may become necessary to make adjustments to the pricing of the product. We have a number of opportunities short of that. One of them, as I say, is to increase the level of vendor diversity. Another is to move some of our products to LPDDR5. It may be possible to move the Raspberry Pi 5 generation products to LPDDR5 memory. LPDDR5 memory is currently not experiencing this. The underlying cause of this price excursion is the repurposing of LPDDR4 production capacity, to build HBM memory for AI applications. There is effectively a pool of fungible manufacturing capacity between LPDDR4 and LPDDR5 memory.
Therefore, the current inversion in which the older memory technology is more expensive than the new memory technology, isn't necessarily something you could expect to be sustained for a considerable period of time. However, we could easily see it sustaining for one or two more quarters. We are doing some work at the moment to evaluate the feasibility of migrating Raspberry Pi 5 to LPDDR5 memory. We'd expect to have a clearer view of whether that's a path we're going to take at some point in the [next] quarter.
That's great. We have had another question come in. Have you explored setting up a VC fund to invest in third-party innovative ideas for quantum change use of Raspberry Pi offerings? The idea is to raise an investment trust that would take strategic stakes in innovative or developed solutions in smaller companies that require working capital management/marketing skills.
I think it's an idea that comes around every few years. It's an idea that I can remember coming around ever since pretty much the point at which we launched the first product in 2012. I think the reason we've not done it is because we have to be conscious of the limits of our capabilities. We are electronic engineers, software engineers, silicon engineers, industrial designers. We are not generally investment. Most of us are not investment professionals. I think it would be difficult for us to articulate a differentiated value proposition in that space. Never say never. As I say, it's an idea that comes around a lot. We are experiencing some power saving, some inability I think. It is an idea that comes around, but I think it's probably not something that's in our future in the next two or three years' time.
[crosstalk] busy and having too much fun with this business.
Yes. I think we're enjoying what we do, and I think we have a realistic assessment of the value that we'd be able to bring to that.
That's great. Thank you very much for answering the questions from investors. Of course, the company can review all the questions submitted today, and we will publish the responses on the investment company platform. Just before redirecting investors to provide you with their feedback, that's particularly important to you both. Eben, could I just ask you for a few closing comments?
Thank you all for taking the time, for taking an hour to come and talk with us, to listen to us talking and to talk with us about Raspberry Pi. We are now roughly 15 months into our journey as a public company. It remains an honor and a privilege to be involved with Raspberry Pi, to be allowed to spend time leading the Raspberry Pi organization. It is a business which remains as fun to run as it was on the day we sold our first Raspberry Pi back in 2012. Once again, thank you for your time.
That's great. Eben and Richard, thank you once again for updating the investors today. Can I please ask investors not to close the session? As you know, we automatically redirect it to provide your feedback, and all the management team can better understand your views and expectations of the management team of Raspberry Pi Holdings plc. W e'd like to thank you for attending today's presentation, and good afternoon to you all.