Hello, everyone. Thank you very much for joining us. These are our Interim results for the H1 of 2024. A quick intro to who we are. My name is Eben. Some of you may know me. I'm a silicon engineer by background. I founded the Raspberry Pi Foundation back in 2008 . I've been running Raspberry Pi Limited, and now Raspberry Pi Holdings plc, since we spun it out from the foundation at the tail end of 2012 .
Hi, I'm Richard Boult. I'm CFO. I joined Raspberry Pi about five years ago, so I've been with Eben and the team through five very interesting years.
Can I have the next slide, please? Just a quick recap on who we are. So we are Raspberry Pi. We build compute platforms. I think we used to say that we build computers. But really, we don't just build computers, we build computer hardware, but we also build the software that runs on that hardware, and we build all of the collateral that sits around the hardware and the software and makes it usable, makes it operable, makes it convenient for customers in our various markets.
Those two markets, really, the thing that we call education and enthusiasts, the place that we came from as an organization, and then the world of industrial and embedded computing, which has come to dominate our volume shipments, has come to dominate our revenue over the last five to ten years. We design our products almost exclusively in the U.K. Almost our entire design team is in Cambridge, where we started the company. We manufacture almost every product in the U.K. We manufacture our products with Sony in Pencoed, in South Wales. Since we shipped the first Raspberry Pi 1 in February of 2012 , we've shipped a little over sixty million computers, and over ninety percent of those are exported outside the U.K. Next slide, please.
Some quick highlights. We'll come back to these in more detail in due course. Highlights from the H1. Obviously, the IPO was a watershed moment for us. A $185 million sell down by our still largest shareholder, the Raspberry Pi Foundation, setting them up for a second decade of impact in advancing computer education among young people around the world, and a GBP 40 million primary raise for the company, about GBP 31 million net of fees, that we intend to put towards developing our future technology roadmap and fortifying our supply chain position. Obviously, the former takes more time than the latter.
We've already been able to deploy some of the proceeds of the IPO to improve our supply chain position in the aftermath of the turbulence that followed the COVID pandemic. What's been happening to us this year? I guess H1 was the six-month period in which we finally completed our recovery from the shortages associated with the global semiconductor shortage, with most of our products now being available ex stock, with really one exception, all of our core products are now available ex stock. Like a lot of participants in our industry, while we are, we have recovered from the, while we've seen the back, I believe, of the supply chain shortages, we've not seen the back of supply chain turbulence.
So we've seen a certain amount of overstock, larger than equilibrium, in inventory in our downstream channel partners and some of our downstream OEM customers. We've seen that beginning to normalize over the last six months, and we believe that that normalization will be largely complete by the end of the year. We haven't allowed ourselves to be distracted by the IPO. We've continued to innovate all the way through the supply chain shortages, all the way through the IPO process. The H1 saw us launch our first dedicated AI ML accelerator product, which we launched in the first week of June.
That's the AI Kit that we developed in partnership with Hailo, our first cloud product, Raspberry Pi Connect, and saw the ramp of our second generation microcontroller, RP2350, ahead of the launch of Raspberry Pi Pico in August. We've made solid commercial progress. We executed the transition, at least in our enthusiast customer base. We executed a transition from Raspberry Pi 4- Raspberry Pi 5. We made progress on ASP expansion, and on increasing revenue and gross profit associated with accessory attach, and we were able to return to growth in our approved reseller base.
Thank you. Maybe if you go to the financial highlights. So, overall, I think we had a good H1. Yes, it was against a supply-constrained comparative in the H1 of 2023. It came into 2023 really with the chips that go in, the logic chips in our boards being in very short supply that ran through that period. Whereas in 2024, we had full access to the logic chips we needed to meet demand. And as I said, the back of 2023, we launched the Raspberry Pi 5. So overall, our unit volume was up 31%. It was GBP 3.66 million.
A little bit lower than we hoped for, but against that, I think the profit per board that was a mix of more high-value boards at better margin per unit. So actually, our gross profit per unit was GBP 8.30, up from GBP 7.70, and we do sort of feel we were pleased with that. That was better than expected, which is why I think overall, the profit for the period was very satisfying. The gross profit from that, GBP 34.2 million, up 47%, and as you can see, adjusted EBITDA up 55% to GBP 20.9 million. So our costs didn't rise as fast as our gross profit was rising, so a good position to be there. Adjusted operating profit after depreciation, amortization up 44%.
Our gross margin percentage, if you take our revenue and divide it into our gross profit, that was 23.8%. That's down from where we were a year ago. We'll touch on that over the page. Over. Just looking at, in terms of revenue, as you may be aware, we sell Raspberry Pis get out to customers, to users through two principal channels. We sell them direct to our resellers. Within that as well, it's where we sell directly to OEMs, and then we also have-- they're manufactured by Premier Farnell on the same production line at Sony, but to Premier Farnell's order, that licensee, they charge us a royalty for that.
So I think in the year, in six months just gone, the direct to reseller was about GBP 2.4 million, up from about GBP 2.3 million in the previous year, and the licensee, you can see, increased significantly. That was probably a deliberate effect. The Pi 5, it, which is the new product we've released, we sold that really in high price variants initially, and we'd always planned, and we'd always agreed with Farnell that they would take those units initially because they, in terms of managing that flow of inventory, they have a very substantial balance sheet and logistics to handle that. So they saw, I think, a benefit. So you can see a little bit of switch between channels. We'd normally target 20%-30% of our unit sales being through our licensees.
In this particular year, it's, as well as six months, it was 35% compared to 19% in 2023. Revenue by category is the other part of that. Focusing on royalties, so this is the money we get from Farnell for licensing our design to them. That basically tripled, a little bit more than tripled, very similar to the growth in the units. And products, which are Raspberry Pi boards, also accessories, which actually had a very good H1 of this year on the back, I think, Pi 5 launch, which also draws in accessories like power supplies and cases and fans for those products, also grew significantly.
But also some of the other product that we sell as accessories, such as screens, the cost of the components to those came down, so we were making a better margin on those as well. In terms of components, now, components, in the H1, that is, we supplied to Sony to make those Pi boards for Farnell. We supplied memory chips, and in the H1 of this year, we supplied also some of the logic chips that go in there. So partly as a function of Farnell making more boards or by taking more boards from Sony, sale of components went up.
The other factor in there is actually it went up partly because we took a decision to supply more logic chips through that route, because it gave us a better way to manage the launch of the product. We probably wouldn't do that, on an ongoing basis. What's the takeout from this? We don't focus too much on revenue because of, let's say, there are three different streams there, all with quite different economics, and the components are pretty fine margined. The royalty is nearly 100% margin, as you'd expect. There's very little cost of sales related to a license fee. And then the product is somewhere around about, say, 20% on average for a board, 30% for a accessory, so a different margin structure there as well.
So what we really focus on, then coming over to the adjusted income statement on the next page, is really gross profit, which I've mentioned on the first slide. It's up 47%. As you can see on the line underneath, gross margin percentage came down. That's because of the higher volume of components in there, and gross profit growing 47%, you can see then we split our administrative costs. We split them between the engineering costs, the R&D costs that we don't. We do capitalize some, but these are the costs for, say, software engineering, which is something that we expense. They grew by about 35%. Well, number of engineers grew about 36%, and then we also had things like salary increases, so it grew of that order.
Administration costs grew by 34%. There was some headcount increase. There was between 5% and 10% salary inflation there, but the other driver of that was actually we just got additional costs and fees, in part because as the business grows, we're doing more travel, meeting more customers, but also, we're doing some work on systems, so we've had fees there, and then there is the cost of being a listed business. Our audit fees are going up, and some of our other fees have increased on the back of that, so that brought our adjusted EBITDA up 55%. Depreciation up significantly at the back of 2023. We launched the Pi 5.
At that point, we start amortizing all of the design costs, the product development costs that we've capitalized up to that point. So as we launched that, we started, so we had a full six months of amortization this year, nothing last year. That design cost us about GBP 20 million overall. That includes the RP1 I/O chip, so, which manages the sort of Raspberry Pi-like relationship with the world, as well as the overall design of the board. So we started to amortize that. That gave us adjusted operating profit up 44%. That brings us average selling price increased 28%. We were selling, as you can imagine, Pi 5s, which were more valuable, higher priced, so $60, $80, as opposed to some of the other products in 2023, that gross profit per unit rose 8%.
It rose partly because we were selling more highly priced boards, so similar margin percentage lifts that. But it's also against that pushing, we have made less profit per board on a Pi 5 than we do on a Pi 4. The Pi 5 has a initially more expensive $5 more expensive Broadcom chip, but it will continue to be slightly more expensive. So it is a little bit less marginful than the Pi 4. So we've seen some of that headwind as well there. The other standout thing within gross profit was accessories. I think I was touched on earlier. With the launch of a new product, that's been very good for accessories.
We made, I think, about $4.1 million of gross profit on accessories. We made per board, which is perhaps an indication of the sort of activity, GBP 1.10, against GBP 0.8 per board, in 2023. That would be, I think, in a good place for that to continue into the H2. Other areas, tax rate, if you take purely our profit before tax and divide it, into the tax that we are accruing in the accounts, it's 29%. Driver on that, really, we have about GBP 2.1 million of, non-deductible IPO costs within the, reported operating profit that's excluded from these adjusted numbers. So that is the reason for the tax rate being above, I think, our sort of guidance of around about 25%, typically.
And so at an underlying level, it's 25%. Adjusted, looking at EPS, basic EPS, we had a number of exceptionals in the H1. Used to be allowed to call exceptionals, but basically adjusting items, so our EPS came down. If you adjust those out, and I guess adjusted diluted, which allows for the share options that we've granted, that rose to GBP 0.058 from GBP 0.046 a year ago. So I think overall, you know, adjusted earnings as well increased. I'm just touching on the balance sheet over the next page. I'm gonna talk about a couple of three areas there. Intangible assets increased. We've spent about GBP 10 million in the period on further development of boards and semiconductors, and that's been capitalized. So that's increased compared to December 2023.
Our inventory at the end of June was about GBP 146 million, compared to GBP 108 million at the end of 2023. At the end of 2023, we had about GBP 35 million of memory that we had bought in December, which gave us a very good position coming into this year, where despite memory prices, spot memory prices increasing quite significantly, we were locked in at the very attractive rates that were there at the last quarter of 2023. Coming into 2024, where's the growth come from? We've added a little bit more memory.
We've continued to see opportunities to buy memory, tactical sniping, you may say, but we've seen, despite prices rising by about 50%, we've still found on occasion vendors willing to sell to us at a price per unit not dissimilar from the previous year. That's put us into a good position where probably going into memory purchases are probably covered through the H1 of 2025 now, compared to, I think at the time we were giving guidance, we thought we would be covered really till the end of this year, so that's given us some more confidence on margins for next year, so we've spent probably about GBP 5 million there. We've got about GBP 10 million of other components. We have had a deliberate policy of accumulating some logic chips.
We still remember the experience of over 18 months to 2 years ago, of being so badly constrained and really, the distress that caused to key customers that we couldn't supply everything they needed. So we are always accumulating some of those. There's probably about 10 million of those. And then the other part of the increase, 20 million, is finished goods. About 10 of that, I think it's fair to say, is with Pi 5 and Compute Module, is with those products now back in free availability. We are holding probably about the right level of inventory now. We weren't, it was too low at December 2023. I think that's probably about 10 million.
The other ten is where we've seen this overhang of inventory within the channels, therefore, the slowdown in our sales in, say, areas like Pi 3 and Pi 4, and we're probably carrying about GBP 10 million more inventory than we would normally choose to do there. So that's hence the growth there. On the flip side, trade payables are up GBP 71 million, compared to a year ago. We are seeing, interestingly, various opportunities to get quite good payment terms from people, as well as the pricing on memory being attractive, which makes it an interesting market. Just quickly over the page, cash flow, just to sort of narrate through. We came into this year with GBP 42 million of cash. As you can see, we had about GBP 21 million of profit. We grew inventory by about GBP 37 million.
CapEx of circa GBP 11 million here, GBP 10 million was intangibles, and GBP 1 million was mostly fit out of a new warehouse. Had the IPO proceeds, which we talked about, paid a bit of tax and a few other expenses, bringing us to about GBP 40 million of cash at the end of the half. We've got a GBP 40 million RCF, which is currently undrawn. So overall, what's this half been? It's been a good half. Slightly lower units, profit per unit is better than we hoped. So overall, we talked at some length about Broadcom, the first tranche of chips for the Pi 5, being slightly more expensive. We're nearly through that costly inventory, and we've managed really the key part of the Pi 5 launch in terms of the volume ramp and production.
How does that reflect into this slide here in terms of our growth strategy? At the time of the IPO, we very much said to people, "What's our strategy?" It's a simple one. We're gonna grow unit sales. In this half, we've grown those 31%. We were gonna grow unit profit, and we've grown that to $8.30, up from $7.70, and we grow profit participation. So this is what proportion of the profit between the cost of making the product and the price that an end customer's paying, do we retain? I think that's probably gone slightly the other way in this half because of that increase in, licensee, the proportion of licensees as this initial wave of Pi 5s come out.
There is more to go for there, I think, as we revert back to that sort of 20%-30% that we talk about, and as we do more in terms of either direct where we are or custom boards, where we'll see improvement there. But I think in the H1, it's very much the things that we aim to see are consistent, are congruent with these aims. Overall then, handing over to Eben just to talk a bit more about strategy for the road going forwards for the remainder of the year.
Sure. Thank you, so a little bit more detail on some of those strategic topics. Raspberry Pi 5, obviously, as I said, this was the half year in which we went into stock availability of Raspberry Pi 5, shipped over a million units in the H1. That's a pace of transition that's comparable to historical benchmarks, particularly the transition from Raspberry Pi 3 to Raspberry Pi 4 in the H2 of 2019 and H1 of 2020. Raspberry Pi 5, obviously, a vastly more performant product than its predecessor. Roughly three times the performance of Raspberry Pi 4, roughly a hundred and fifty times the performance of the original Raspberry Pi 1 from 2012.
But it's not just a faster, it's not just a more performant product, it's also one that brought in a number of very significant changes in the production process, particularly in the way that we insert the through-hole components on the board, the way we singulate the boards. Raspberry Pi 5s are manual. Like its predecessor, Raspberry Pi 5 is manufactured in panels of nine, and the way we break those panels out into individual boards is very different from what we've done historically.
Both of those were introduced in order to reduce manufacturing costs, but they also had the side effect of increasing manufacturing quality to the point where we actually see 50% or lower early life failure for Raspberry Pi 5 compared to its predecessors. The expectation is that we will roll those production improvements back, at least to Raspberry Pi 4, possibly not to earlier generation products, as those products continue to taper off into a purely industrial legacy use case. Can I have the next slide, please? I'd say 2024, in addition to being an IPO year, really for me, is a year of new product launches.
People have been doing the first of these, the AI Kit that we launched in the first week of June. Obviously, people have been doing things that you would recognize as artificial intelligence, or probably better branded as machine learning, since the launch of the first Raspberry Pi, the first Raspberry Pi back in 2012, but the AI Kit is the first time we've produced a first-party product, which is targeted at that set of use cases, the first time we've put our logo on an AI-centric product. The kit is the fruit of our partnership with Hailo, bundling the 13 TOPS Hailo-8L accelerator with our M.2 HAT+ that provides the connectivity to the Raspberry Pi 5 PCI Express bus. We've seen very, very robust demand for this product.
When we launch products like this, accessory products like this, it's always a little bit hard for us to tell whether these are gonna be niche products or whether they're gonna be mass-market products. I think our conclusion is this is a mass-market product. Currently, it's a mass-market product, which is targeted at the enthusiasts, like almost all of our products in the first months of its life. It's been adopted by the enthusiast community. We do believe that there are compelling industrial and embedded use cases for this technology, and that this product will transition across as so many of our earlier products have. This product will transition across to that market over time. It's also the first of many. This is not the last AI-focused product that you'll see from Raspberry Pi.
It's not even the last AI-focused Raspberry Pi product that you'll see in 2024. You know, we do believe in this as a use case for our products, and you'll see more of that from us going forward. Another thing that we launched in May, Raspberry Pi Connect. For us, software, Raspberry Pi has always been an enabler. Software has always been something that runs on Raspberry Pi hardware. It's always been a cost center for us. It's never been a profit center for us. It's never something that we have attempted to sell to our users. It's never something that we've attempted to sell separately from the Raspberry Pi hardware. The things that are made of bits are cost. The things that are made of atoms are profit center.
Raspberry Pi Connect is a departure from that for us, but retains a lot of the kind of core philosophical elements of the Raspberry Pi value proposition. It provides a secure and easy to use remote connectivity capability for Raspberry Pi hardware to our enthusiasts and increasingly our industrial and embedded customers. This is our first piece of software that doesn't run on the Raspberry Pi. It's the first piece of software that has a substantial cloud component. It will also, in due course, as we roll out features, it will always be free for our enthusiast customers, but for our industrial and embedded customers, as we roll out features which are specifically tailored to their needs, it will be the first time that we are charging people money.
It will be the first time that we have a recurring revenue stream for Raspberry Pi products. It'll be the first time that we charge people money for a product. This is currently in public beta. We've had over 50,000 users sign up for this product since launch in May. We've seen strong interest both in the base product from our enthusiast customers, but we've also started to see, as I say, interest in and willingness to pay for enterprise-specific enhancements to this product, and you'll see those from us over the next year. And finally, of course, we continue to innovate in silicon. Raspberry Pi Pico 2, which we launched in August, is powered by our second generation microcontroller platform, RP2350. That's our second microcontroller.
It's our fourth silicon product after RP2040, the first generation microcontroller, RP1, the I/O controller in Raspberry Pi 5, and RP3A0, which powers the Raspberry Pi Zero 2 product. This offers more performance than its predecessor, but also improved interfacing functionality and some of the security features, which make this the first time a first-class product for a very, very broad range of embedded and industrial use cases, and it does all of those things while retaining almost complete compatibility with its predecessor, RP2040. Next slide. Quick look at our go-to-market priorities in the H2 of this year and into 2025. We continue our gradual shift towards direct to OEM, towards serving some of our largest OEM customers, directly.
This offers improved unit economics, and I think an unexpected benefit of being a public company is that it's been a lot easier for us to engage at a senior level with new prospects in this space, as a public company than it was as a private company. As I said earlier, we've resumed expansion of our reseller network, our approved reseller network. This is the primary way in which we meet our customers. We had to pause the expansion. We paused the expansion of the approved reseller network in 2022, not because we felt that we had approved resellers covering every geography and every vertical, but simply because we had a finite supply of Raspberry Pi product.
Spreading that finite supply ever more thinly across a growing network of approved resellers didn't feel like a respectable thing to do to our existing and very loyal AR partners. With the return to supply, we've been able to return to growth in the AR network. We've recently added an approved reseller, and this is not, this is not purely new markets. This is in markets which you might sometimes consider to be mature for us. We've added resellers in France over the last six months. We've added resellers, we added a large industrial reseller in Poland, and we will continue to do this where there are geographies or verticals which we feel are underserved by the existing network.
And finally, we continue to make those investments in selective investments in increasing sales resource, again, targeting underserved geographies and verticals. And in particular, we have a focus on leveraging some of the capabilities of our major distribution partners. Our major distribution partners have large demand creation capabilities, which we are able to tap into to sell more Raspberry Pis, particularly into the industrial and embedded space. Next slide, please.
What does this look like when we get it right? Well, the Heathrow digital signage deployment is a great example, a great recent example. Flight information display systems are the screens that you see when you're outbound at the airport, typically, these days, tell you how late your plane is going to be. The platforms that were at Heathrow until very recently ran on a platform that was roughly a decade old, and it was a challenge to modernize and improve these platforms with the best possible price performance.
We have a very long-standing and valuable partnership with Sharp NEC, provider of digital signage platforms. And the solution that's been deployed at two terminals, will be deployed across the rest of the estate over the next year or so, is consisting of Sharp NEC display platforms, powered by Raspberry Pi Compute Module 4. Why was Raspberry Pi selected for this opportunity?
We were providing products which were flexible, adaptable, secure, and critically manageable, which had a low acquisition cost, so both a low hardware cost and an open source operating system, which did not run on our license fees, and a low running cost. The contributors to that are both high reliability and therefore a lack of requirement to constantly replace a fully solid-state solution, which is extremely reliable and didn't require replacement over the assumed 10-year lifespan of the platform, and low electricity consumption, so low running cost there. Longevity, this is one of many, many, many examples of a design win, where our commitment to longevity on our platform was central to being able to get a win.
Raspberry Pi Compute Module 4, in this case, is guaranteed to be available until 2034, and you're seeing increasingly 15-20-year availability guarantees becoming a feature of our products at launch. Where's the opportunity here? There's an opportunity, as I said, to complete the rollout of flight information display systems across the entire Heathrow estate. There's a separate opportunity to go and win the baggage information display systems, which tells you where your baggage has ended up. I spent a very enjoyable couple of hours staring at a baggage information display system in terminal four, on
Sunday evening. And this really is a proof point for future wins at other airports. Next slide. People and culture. We talk about our people a lot. Why do we talk about our people? We talk about our people because they are our very, very deepest most. We've continued to grow the team over the last year. We've continued to invest in engineering, in the engineering team. We have, even before the IPO, a high rate of stock ownership, across the employee base. And we're very pleased that since the IPO, we've seen a near 100% retention rate across the organization and 100% retention rate within the engineering team. I think the thing that has probably changed.
In our approach to the team and our approach to recruitment over the last couple of years, we've got a lot better at recruiting graduates. Historically, for us, recruitment has focused on relatively senior hires, relatively senior mid-career professionals. I think what we've got a lot better at in the last couple of years is running our internship programs and converting those interns into graduate hires. It complements our historical strength in senior hiring. I think one thing we're very good at doing is putting new graduate hires alongside very senior engineers. I think both the senior engineer and the graduate finds that valuable. I think one very useful thing is if you look at the, you know, we recognize that we work in an industry with substantial diversity challenges.
The really encouraging thing is when we look at the diversity characteristics of our graduate stream, we're measuring the future there, and I think the future is bright for a more diverse workforce in our industry. Next one. Finally to outlook. As I said earlier, channel and customer inventory levels continue to normalize. We do expect this. This has been an ongoing process over the course of this year, and we do expect this to be complete by year-end. We expect to see somewhat higher unit volumes in the H2 than we did in the H1, driven by a number of new product launches.
Many of those new products are towards the lower end, particularly. You've already seen us launch Raspberry Pi Pico 2, a $5 product, and the two gigabyte, low density, low cost variant, Raspberry Pi 5. These are towards the low end of our price performance range, and therefore, we expect the unit economics, which were a standout feature of the H1, to normalize somewhat as we go into the H2. We haven't communicated at the time about the IPO. We expected the H2 weighting to profitability in the business, and we are no longer communicating that, but our expectations remain otherwise unchanged. I'm now happy to open the floor to questions.
Thank you. We will now begin the question and answer session. If you would like to submit a question, please do so by clicking the Ask a Question button on the Spark Live Webcasting page. Our first question is: What would you estimate is the mix of enthusiast and industrial customers in H1, and do you think this changes in H2 and 2025?
I think we are still where we were at the time of the IPO, which is we believe that between 70% and 80% of our volume, and 70% and 80% of our revenue is derived from industrial and embedded customers, with the balance from enthusiast and education, with enthusiasts very much dominant in that other small portion.
Great. Thank you. The next question is, what is the roadmap for your own AI products over the medium term?
I think if you look at our general approach to AI at Raspberry Pi, as I say, people have been doing AI applications since Raspberry Pi 1, a platform with one hundred and fiftieth the performance of a Raspberry Pi 5. So people have been doing this, this is not new to us. I think we do see AI, it's a continuous rather than a discontinuous thing. To a great degree, it is just another consumer of compute at the edge, and to the extent that Raspberry Pi produces products which provide large amounts of compute, compute capability at the edge of the network, they are well suited to many AI inference tasks. I think what we try and do is we try to build platforms. There's a Pareto thing going on here, right? There's an 80/20 thing.
We try to produce platforms which are suitable for 80% of the workloads, unaccelerated, can support 80% of the workloads. And for the 20% of workloads which are beyond the capabilities of the unaccelerated platform, we aspire to be the best possible host for other people's embedded accelerators. In the current iteration, that's the Hailo product. We have in the past demonstrated another product, which will in due course become a feature of our roadmap. That's the AI Camera, the Raspberry AI Camera. We demoed this at trade shows earlier in the year. This is based on a product from our strategic partner, Sony, a product called IMX500, that integrates a 12-megapixel camera with an on-die 2 TOPS per second, 2 TOPS accelerator.
A rather lower performance accelerator in terms of headline performance, a rather lower performance accelerator than the Hailo product, would occupy then a distinct tier in our price performance roadmap between the unaccelerated platform and the high-end Hailo-based platform.
Thank you. The next question is, who is the target market for CM 5 when you launch it? Does it open up new verticals?
I think there's a little bit of a mix here. I think we are unlikely to see displacement in existing designs of CM 4 by CM 5. I think we, you know, we are aware that we have CM 4 customers who will want to offer tiered versions of their products, so who want to translate our offer of a new tier of compute module performance into a new tier of performance for their customers.
We provided what we've called forward guidance for people who are doing CM 4 designs to maximize the chance that, and we have had an early sampling program as well, which maximizes the chances that our larger existing CM 4 customers will be able to offer tiered, either, new products based on CM 5 or tiered versions of their existing CM 4 products. There's a high degree of compatibility between CM 4 and CM 5, but there are potential pitfalls, and we've worked very hard over the last year to educate people as to what those benefits are. Does it open up new markets? My belief is that it's-
More of those markets.
Yeah, it opens up. Yeah, every market is a bell curve. Every market has a distribution of performance requirements, and, you know, within that, within any given market, obviously, CM 5 will let you push further to the right in that bell curve. That's a sort of a quantitative argument. I think the interesting thing about the success of Raspberry Pi 5, and Raspberry Pi 5 has been a great success for us. I think what's been interesting to me, as a technologist who often tends to get quite focused on quantitative stuff, is that the qualitative behavior of the platform, particularly the addition of PCI Express connectivity, has been an enormous driver of enthusiasm for the Pi 5 platform.
The AI accelerated kits is a good example of this, but people have also been connecting network, high-speed networking, and high-speed storage to the platform using that interface. I suspect that the improvements to PCI Express connectivity in the Compute Module 5 platform will be at least as important as the to the adoption of that platform as it has been to the adoption of the Raspberry Pi 5.
Thank you. The next question is: how do you view your current and short-medium growth position is compared to your direct competitors?
That's a fascinating question. So I'll look at you.
I think we've been struck through these last few months how some of our experience with excess inventories in the channel has impacted us. That's probably the electronics industry as a whole, rather than saying a direct competitor. I think we would struggle to point to a direct competitor where we have that degree of granularity to say we are better or worse. But I think that the experience we've had in the last few months, it's resonated with what other people have said for us.
Yeah, I think one of the challenges, I mean, I think we experienced this at the time of the IPO, one of the challenges with Raspberry Pi is to put your finger on a direct competitor. You know, we have... I think you could divide, you know, split the world into inbound and outbound competition, so existing incumbents with whom we compete. I think that the steady increase in the performance of Raspberry Pi products while retaining an attractive price-performance ratio and the availability, the sort of, I would say, the increasing capability of the Raspberry Pi OS platform and the various ancillary services that we provide around it positions us very well.
And I do think we do see increasing industrial and embedded adoption versus incumbents on the basis that we're just bringing a transformative price-performance point to the platform. I think the other thing we think about when we think about competition is we think about insurgents. We think about people who are attempting to compete with us in the way, the same way that we compete with incumbents.
I think the remarkable thing there is to see that is that still, after a decade and a bit, after 12 years of shipping products and evangelizing the opportunity that we are addressing with Raspberry Pi, that we've yet to see the emergence of any competing platform which offers what we offer, both in the way of the price-performance ratio of the hardware, and of the software investments that we've made around the hardware to make the hardware operable. That is a remarkable feature, I think, of the market that we operate in. In some way, you know, it is the reason why we continue to make those large investments, both in hardware and software engineering, in order to maintain and deepen those moats that we have around our platform.
Thank you. Our next question is: Will you need to increase CapEx over the next couple of years as you launch new products?
I think we guided at the time of the listing to about GBP 20 million of CapEx this year, and I think probably a similar number next year. I can't, I think maybe then we said it would move more to be in line with a percentage of gross profit. I can't remember exactly the number there, so I'll need to look that up. So I think at the moment, we're happy with that guidance. I think the aspect that is sometimes difficult to communicate sort of clearly is that we... A lot of these products, particularly the more sophisticated ones, the more expensive ones, are a three-year, four-year program. I mean, I think people have been starting to work on a Pi 5, probably even before the Pi 4 was launched.
The RP1 chip that's in the Pi 5, the first pieces of work on that will go all the way back to 2016. So there has always been a longer tail, and that does go to some scope to sort of manage it as we go through. So, to say it's product launches are driving CapEx, obviously, it is spread over a period of time, so I think there'll be a reasonable stability in that. There's always a risk of a bit of fluctuation. There are some costs as you come to a particular moment on a product that might sit one side or other of a reporting period. So, there will perhaps be some fluctuation, but over the longer timescales, the medium scale, I think, at the moment, the level of guidance is something we're comfortable with, but we'll keep talking.
Great, thank you. And our next question is: what does the pipeline of custom design products look like, and how much do you expect this to contribute to revenue growth?
So I think at the time of the IPO, we communicated a handful of custom opportunities. Those continue to ramp, those are in good shape. I think what's been happening, I think it's a little early, we are still a little close to the IPO to report, like, major changes to that pipeline. I think what I would say is that we are broadening out a little bit our thinking about what custom looks like. And, you know, I think that the examples that we cited at the time of the IPO were extremely custom, were very pure and quite extreme customizations.
I think where we are perceiving opportunities in the shorter term, where we can see some pipeline already, is in the partial customization of standard products for some of our medium volume customers. If you imagine that customization, I think, is something which takes place at the 50-100 thousand unit a year level and above, and so is intended to sustain us, to sustain our largest volume customers beyond the 50-100 thousand unit a year range. I think what we're starting to perceive is that we can build an organization that can do lighter weight customization a little bit below that, in maybe the 25-50 thousand unit a year levels. So we have, you know, some stuff underway. I suspect we will talk about this a little bit more in our full year results, when we're a little further away from the IPO.
Great, thank you. And the next question is: what is the potential of your own semiconductor units in the medium term?
I think we are very conscious that we are a... We're just getting started as a semiconductor company, rather than a company that builds semiconductors to burnish the capabilities of our traditional core product line. As a pure- play semiconductor company, we are just getting started down this road. You know, it's a show me. Semiconductors from us provides a show me. We are very keen to show people what we can do, and we would rather show people what we can do before we talk too much about what we're gonna do. I think that I am an enormous believer in our ability to do good stuff in the microcontroller space, to do good work in the microcontroller space. I think there is an enormous value in our building microcontrollers, while also using microcontrollers.
That closed- loop world in which, because we are great users of microcontrollers, we do believe that we understand what makes a good microcontroller, and we do believe that the RP2040 and RP2350, we have built microcontrollers which are genuinely differentiating. These are low ASP products. You're talking here, when we're selling microcontrollers, you're talking about selling $0.50 objects rather than $50 objects. And that means that even as, and we believe this will happen either this year or next, the volumes of microcontroller products come to catch up with and potentially overtake the volumes of board- level products.
Still, from a revenue perspective, it's extremely hard to see in the financials. We do believe that there's scope for growth, and we do believe that even at the sales levels we're at the moment, there is enormous value in putting product in the field, which runs our software, has our logo on the front. We believe that there is value to that footprint, which is hard to measure in purely financial terms.
Just to add, I mean, this is a huge market, so any estimate we give has such a large error bars on it. If you have 28 billion units, being wrong by a fragment of a %, you are still a long way out in terms of units.
Yeah.
Great, thank you. Our next question is: Can you talk about the impact on gross margin benefits from the vertical integration strategy, i.e., move to semiconductor?
Let's see. Obviously, it has an impact, obviously it has an impact, and I think we take this two ways. It has an impact, and it needs to have an impact. It has an impact because, of course, the cost structure of a piece of semiconductor, a semiconductor component that you produce yourself will be better than the cost structure of a piece of semiconductor, an equivalent semiconductor that you buy from somebody else. So it will have a positive margin impact. I think the other argument, of course, is it has to have a positive impact, because it will have cost you quite a lot of money to develop, and therefore you do need to recover that expenditure through improved margins.
I think if you look at the microcontroller boards, you look at the Pico products, they are both some of our highest margin products on a percentage basis. Obviously not on a dollar basis, 'cause they have a low ASP, but on a percentage margin basis, they have a substantially better margin structure than the traditional Raspberry, the big Raspberry Pi. So I think there is a benefit there. I think we've said we don't have an aspiration to own all of the silicon. Well, it will never make economic sense for us to own every piece of silicon in every Raspberry Pi product. But I think to the extent that we do turn that dial up a little bit, it does come with an opportunity and an obligation to improve margin.
Thank you. We have no further questions at this time. I will now hand back to management for closing remarks.
Thank you. Well, thank you for taking the time to join us. Thank you for the questions. That's been extremely helpful, and the questions often help in elucidating kind of bits of the story that we haven't done in the presentation before. The IPO is a watershed moment for Raspberry Pi. It was a very important moment in a journey, which for the Raspberry Pi Foundation was fifteen years long, fifteen, sixteen years long, which for Raspberry Pi Limited was twelve years long at the point where we IPO'd. But I think it's important. I think we said on the day, it is an amazing day, but it's also just another day. The best day of...
We both very much enjoyed our participation in the IPO process, but for me, the very best day of the year was the twelfth of June, because it was the day that I went back to the office, and I went back to running Raspberry Pi, so you know, it's an important step for us, but it's just another step, and we think there's a lot more road in front of us than there is behind us. We're looking forward to what's next, so thank you again for your time.