My name is Charles Shi. I'm the EDA IP analyst at Needham. Welcome to the 28th Annual Needham Growth Conference. Last day, but the best day. We have Arteris IP with us in this session. They're going to do a presentation on the screen. You could just see we have two guests here, two presenters here: Arteris CEO Charlie Janac and Arteris CFO Nick Hawkins. At the end of the presentation, we can conduct a few Q&A. If there are questions coming from the audience, if you are on the Needham portal, you should see a Q&A box. Feel free to type your questions in. I will try to ask your questions at the end. Without further ado, Charlie, Nick, please go ahead. Stage is yours.
Thank you.
Okay. Thank you, Charles. Thank you for allowing us to present Arteris. So standard disclaimer. So Arteris is a semiconductor IP company. And basically, if you look at a semiconductor chip at a high level, it does three things. It processes data, it stores data, and it moves data. We are focused on the part of the chip, IP and software, that essentially facilitates data movement inside the chip, between chiplets, and also potentially between some chips for some applications. So we are essentially the pioneers of something called network on chip. And this is using networking techniques for moving data inside semiconductor chips. So before us, there were wires that were connecting processors to memory and to I/Os and subsystems.
With Arteris, the IP blocks essentially put out their native protocol, and then Arteris converts them to a packet format, and then the data is shipped across the chip with a packet format. We have essentially developed a number of technologies for doing this. There is a cache-coherent product called Ncore, which is essentially used for processor subsystems, which typically are between the processor and the memory. We have non-coherent NoC IP, which is basically used for transporting data for the rest of the chip. And then we have a number of data movement options, such as last-level caches, reorder buffers, and so on and so forth. In 2020, we acquired a company called Magillem. And Magillem does three things. It essentially establishes high-level SoC connectivity. This is connectivity between the blocks before you ever worry about the network-on-chip.
This was previously done with spreadsheets, but there are so many connections now that spreadsheets are impossible to use. There is the register management, hardware-software integration capability, and IP block packaging. I'll also talk about security in a minute. So basically, our business is driven by increasing complexity of chips. When we started, again, it was processor, a couple of memories. Now there's hierarchies. There's 16-bit, 32-bit, 64-bit coherent clusters. There's on-chip and off-chip data storage. Some of the data is kept on-chip, some off-chip, going to DRAM. There's chiplets. So the data movement is becoming one of the most important things that a chip does. And if the chip cannot move data, it's not a chip. So our job is to have the most robust, highest quality, and the best-supported solution for data movement. As I said, the complexity is increasing with more processors, more IP blocks.
AI is increasing complexity in that processing large language model data requires moving billions of weights and trillions of parameters. Many of the data center designs and some automotive designs are essentially being converted to chiplets, where basically different pieces of silicon perform different functions, and they're connected together, and the trick is to make those chiplet connections look like a single programming space of the software. Everything's connected to the internet. Everything's going to have AI features in it, and there's a number of new business models that are being generated based on silicon, such as automated driving, automated vehicles and drones, increasing number of space applications, and so on, so basically, the future of mankind is based on silicon, and our job is to make it easier for our customers to build silicon, so Arteris is a public company. We have over 300 employees now worldwide.
We have been in business in its current iteration of Arteris since 2014, so about now almost 12 years. There have been about almost 4 billion SoCs shipped in electronic systems that contain Arteris IP, and we have about 230 or so active customers. Those customers are some of the largest companies in the world. They are medium-sized companies. They are startups. They're located in all geographies of the world, and we have offices throughout the world to support our customers' activities, and we have a lot of effort built on the to support the ecosystem that's involved in terms of EDA tools, foundries, other types of IP, and industry standards. One of the things that we try to do is have the leading-edge technology based on customer input.
And so our concept is that we have at least two enhancement releases of our existing products and that we deliver one new product per year. What we've delivered in 2025 is something called FlexGen. And FlexGen is a way to generate NoC IPs automatically, starting with the list of connections, the location of the IP block exit ports, and a floor plan. And from that, you specify your performance objectives and get a NoC. It's a big piece of technology. It's something we started shipping in February. And basically, our customers can now build NoC IPs about 10 times faster than they can do it manually with expert users and our manual tools. And they can typically, as long as the chip is complex enough, obtain up to 30% shorter wire length in their designs. So this is a major piece of technology for us.
The other thing that we're focused on is chiplets. For the product for 2026 is a chiplet solution, including cache-coherent heterogeneous chiplets, where you have our Ncore cache-coherent product on one die and cache-coherent Ncore on another, and you can do coherent reads and writes across multiple dies, up to four dies, and then essentially make it look to the programming software as if you're programming against a single memory space or a single processor, so this is chiplets. We have made a strong effort in the AI business. We originally got started with AI with the automotive ADAS chips, which are really AI chips applied to automotive vertical, but AI is much broader than that, and so now about half of our design starts are typically for AI applications from some of the market-leading companies in AI.
This is for training, inference in the data center, and edge inference. We're very, very focused on AI. Some of the very recent customers, we were able to announce a large relationship with AMD in the second quarter and with Altera in the third quarter. We have been focused on RISC-V support as well as Arm support. We definitely have an ability to support Arm processors, and we do, but also RISC-V. Some of the exciting applications that recently were announced are in relationship with NanoXplore for satellite SoCs and with 2V systems for a cache-coherent I/O hub, which is a chiplet that is at the center of a chiplet that's basically a traffic cop for chiplets. We're also focused on supporting standards. One of the things that's evidencing this is that we have joined the AMD-led Ultra Accelerator Link Consortium.
We plan to support the UALink protocol that's being championed by that consortium. This is an example of one of the slides. We said that we have a 90% plus customer retention. One evidence of this is our 12-year relationship with Mobileye, which essentially we have been in all of their chips since EyeQ3 all the way through EyeQ6H. This is just an evidence of or a proxy for us maintaining long-term relationships with our large customers across generations of product and generations of technologies. In terms of growth and how we operate the business, again, we are trying to build one new product per year. We have done this now for a number of years. Production of FlexGen was really delivered in February of 2025.
And we will deliver our chiplet solution that we announced in June of 2025 at the end of the first quarter of 2026. And we also have a new product that we're working on already for 2027. We are focused on the most attractive markets. Automotive is a very attractive market. And automotive is really physical AI because it encompasses robotics. It encompasses logistics vehicles, drones, and other automated vehicles that are able to make decisions on their own. RISC-V, as well as Arm, AI, and now an increasing focus on the chiplets and data center. So in addition to the market, we're focused in R&D, we're focused on financial performance. So our goal is to continue funding a robust R&D program. We're one of the few companies in the market that has the scale to undertake large developments with large customers trying to do very complicated things.
So we want to continue funding a robust R&D program. But we have promised Wall Street that we will be free cash flow positive in 2025, and we have done that. We have a strong balance sheet without debt. And we are very, very diversified in terms of customer size, customer applications, and geographies. In addition to this financial performance, we're also adding organic growth, I'm sorry, inorganic growth. And so we made an acquisition of Magillem in 2020, Semifore in 2022. And a couple of days ago, we announced the acquisition of Cycuity. And Cycuity is a hardware security verification company. Very exciting technology, very exciting team that is joining Arteris. And it underlines our commitment to, one, keep Cycuity open to any IP block that needs security hardware verification.
But we're also very interested in securing the data flow through our native network on chip IPs so that we can not only assure our customers of functional safety, but also of secure data movement. So with that, we think that we're in the process of building a major company. And with that, I'd like to turn it over to Nick Hawkins to talk about the financial aspects of our company. Oh, okay. This is just the thing. And so Nick, please.
Sorry. Thanks. I was on mute. Yeah, thanks, Charlie. So yeah, if you could, as a preface to this, this is our because we don't print fourth quarter for another two or three weeks. So this is our third quarter. So it's a little bit historic. But there's some interesting takeaways from this.
The third quarter was characterized by very strong performance, meeting or beating all of our guidance, in some cases on the top line measures above the top end of the range. But I want to focus your attention on the bottom two, interestingly. The top line is great. But there's a divergence you see here between the profitability, which is negative, and the free cash flow, which is positive. And this is common across all quarters. This happens to be the third quarter, but it's common. You'll see it in the guidance as well. And the reason that the profitability trails free cash flow, free cash flow is based on essentially point-in-time deal recognition. That's when people pay us. They tend to pay us upfront, customers. We have a deferred revenue model.
So initially, on signing a deal, the deal value gets placed on the balance sheet, essentially in remaining performance obligations or RPO. So we now have a very significant RPO position, which we'll see shortly. But there's an 18-month to two-year drag on that showing through into profitability because of deferred revenue. And for those of you who have covered deferred revenue model businesses before, you'll have seen this is a common feature. So we're already profitable on a cash flow basis, on a GAAP basis or a non-GAAP basis. That trails. And we get there in the fourth quarter of this year. So it's the first time, I mean, I'll say we're going to be profitable in the quarter this year as opposed to being a future event. So next slide, Charlie, if you wouldn't mind. So again, history here. There's a few things that are callouts here.
The RPO that I was just referencing, which is essentially our backlog of future contracted revenue, is now close to or was at the end of the third quarter, close to $105 million. Typically the fourth quarter season uptick in RPO because it's generally a strong bookings quarter. That represents about 1.5x annual revenue run rate at the moment, so GAAP revenue, so that is a very comfortable position to be in from a guidance perspective. I know 85%-90% of the coming quarter's revenue before I start the quarter, so that's good, and it helps take out some of the beta out of your analysis. The interesting thing, so the canary in the coal mine for us in terms of top-line movement is RPO, and RPO was up 34% year over year at the end of the September quarter.
The second lag behind that is annual contract value, ACV, plus trailing royalties. And that was up 24%. And revenue is the laggard that because of the deferred revenue model, it takes time for that to catch up. And that grew at 18%. Our stated target in the short future is for license revenue to be growing at between high teens and low 20s %. This is in line with these numbers. And royalties to grow at roughly 2x that. So sort of high 30s, low 40s % in the short term, and increasing over time, which you'll see in a second. Positive free cash flow in the quarter. And we'll come back to that in the next slide. So if you could just advance one, Charlie, that would be super. In fact, we'll come to it in a future slide.
This is basically our revenue model, which we started putting together because you can see here that there are three different legs to growth. Charlie alluded to these earlier on. We have the base level, which is the bottom piece of the pie, which is our license revenue trajectory. You can see there's a slight uptick in the growth rate in there, but slight. That is always a damped metric because of the ratable, the deferred revenue model. It damps the growth and delays it, as I mentioned. The royalty growth rate you can see is coming from a relatively small base, which is the mid-blue sort of meat in the sandwich here. That is growing much faster than license, 2x, as I mentioned earlier. You can see if you look really closely at this, there's an inflection point around 2027 and then again in 2028.
So in other words, the growth rate of royalties is accelerating. And you'll see more of that in the later slide. And on top of that, you've got the inorganic, the lightest blue piece at the top. And that's got off to a fine start with the addition of Cycuity, which we closed on Wednesday this week. So very excited by that, very excited by the potential. It can, regardless of any synergistic benefits or inputs from our Arteris sales and engineering and application engineering, which is a global thing. And theirs is very much a local U.S. play with a very small team. We have roughly 10x the number of people. And so we think we can take them geographically to places they haven't been to before and into the 230 sort of active customers that we have. And they have a small handful.
So I think there are some benefits to be had from that. And we believe there are other maybe two opportunities in this five-year timeframe that's up here for further M&A accretion. We have a number of irons in the fire there. So Charlie, if you could just turn to the next, which is more the first royalty slide. So the key takeaways on our royalties is they are growing very rapidly and increasingly rapidly over the years. You can see here that the green section here of the pie on the right, for example, is around 60% of all deals by. This is by number of deals, not by value, is our pre-mass production. And that's after having taken quite a sizable haircut to those, assuming that some don't make it all the way to mass production.
Typically around 80% do, 20% fail because the customer removes the program or they get acquired or the landscape changes, but 80% get all the way there, and so if you look in the blue sections, the 20% darkest blue, that is what we call our mature royalty streams. It's dominated by a few major customers, but that number, as you'll see later, is growing, and that is only one-fifth of the total design pipeline, is generating 80% of the royalties right now, and you can see that the ones that are in ramp with the lighter shades of blue are increasingly adding to that, and then the green is not yet to be seen. We haven't seen the checks from that yet, so it's a very robust and exciting model.
If you flip to the next slide briefly, Charlie, so what you see in the background here is this is actually based off our royalty model. It's got every deal that we've ever done that includes a royalty element. So this is only based on existing already won deals at the September 30 date. So no new deals are in there yet. We don't take those into account until they're landed. So there's upside to this, particularly in the outer, the later years. Big position in automotive, it's more or less half of our, a little over half of our total royalty stream comes from automotive. Because automotive is the sweet spot where high volume meets high royalty rate. There are other areas, for example, in data center or some of the space applications where very high royalty rates, but very low volumes.
And so they don't amount to a huge amount of dollar contribution. There are some areas, for example, more consumer where the volumes are huge, but the rates and the same is true for MCU market, for example, which we entered particularly strongly with Infineon in late 2024. And those are very high volume and relatively low rates, but still contribute significantly to dollars. Automotive is in the sweet spot where high volume meets high rate, which is great. The biggest takeaway I would encourage you to look at on this is the small detailed stacked bar chart on the top right. If you roll back to the time we went public and just before we went public in 2020 and 2021, at that point, we were a royalty stream was a one-trick pony. The trick was called, or the pony was called HiSilicon.
And that represented at one point 90% of our royalty stream. They ended up being on the wrong end of a U.S. government ban. And they joined the banned entity list. And so that you can see here, this is darkest blue piece on the left-hand side. That evaporated to zero within about a year. Around about the time we went public, ironically, it had got to pretty much zero. And that shows the cost and the risk of lack of diversification. We had high concentration on one customer for royalties. That customer ended up tripping up. And we lost that entirely. It's now been completely replaced. We've now eclipsed where we were when we had this huge HiSilicon royalty stream. And you can see on the far right that we now have five in the third quarter, what we call six-figure reporters, the six-figure club.
The people are sending us six-figure royalty checks every quarter. So instead of one, we've now got five. That number we've already explained to people is growing. So you can expect the fourth quarter to have another one or two added in there. We now have a space customer even who has a six-figure royalty. But note also that the top here, the gray section, which is the sub-100K, the people who haven't made the six-figure club yet, that is growing dramatically. And that's because we're having so many successes in royalty deals. And over time, more of those will shift into the six-figure club, and you'll see them individually identified. We don't actually note the individual customers here for confidentiality reasons, but they're people you would know. One of them is Mobileye, for example. You would know that one. That is public.
So moving swiftly on, Charlie, as I know we're time-constrained a little bit, this I would just take away separately. You can find this presentation on our IR website, for example. This is more about how our business model works. I'm not going to pay too much attention to this because I've caught on most of the key points on this. The one I will just touch on briefly is the OpEx in the bottom left. So we are continuing, Charlie mentioned earlier, the number of new products and even variants of existing products that are released every year. And some of these are quite complex and quite significant. And so we are front-foot forward on R&D. We think that's one of the two engines, the two key engines of growth for the company.
If you stop producing new products, if you cut back on R&D like some people have historically, then growth is compromised. And we don't intend to do that. We think we have a pretty exciting future. So we're keeping up the investment in R&D. Typically, that grows at 50%-60% of the top-line growth rate. So we get a bit of operating leverage out of that. Sales and marketing tends to grow at around 75% of top-line. That is a natural thing for each new company. There's so many deals that each salesperson or application engineer can actually handle. And so there is a little bit of scale benefit to come through automation and AI and so on. But in the majority, there is a certain amount of linearity at 75% level with top-line. G&A, we get a lot of operating leverage from.
We've spent the last three years and have grown 0% in G&A, even taking into account inflation. So actually, we've been net negative about, on average, 4% for the last three years just through efficiency. But we are reaching a point where this year we'll need to have some. We've indicated that in 2026, there may be a small single-digit % increase in G&A, principally because this is the year in which we flip into SOX 404(b). We've grown up now. We're a five-year-old public company. And at that point, that's when 404(b) kicks in. And that's an expensive process. So there's a little bit of a tick up there. So with that, Charlie, I think the last slide, just maybe if you haven't already seen, this is where we see the guidance for the year, or we saw the guidance for the year in third quarter.
We'll update that in two or three weeks in our fourth quarter earnings call, but we're looking at revenue at around $69 million mark, free cash flow at an average of plus $4 million here, and we still feel comfortable with all these numbers, and NGOI, again, to my point earlier, NGOI is a negative $13 million, so it's a P&L loss on a non-GAAP basis of $13 million, but a free cash flow positive $4 million, and that is all to do with the deferred revenue concept and construct of how we recognize the top-line. OpEx, obviously, we recognize as it happens. Revenue is based on the deferred revenue when that amortizes into revenue, so with that, I'm going to hand the baton back to Charles to see if there are any questions that have lined up on your queue.
Yeah, thanks, Nick.
Hopefully, everybody can see there's a Q&A box where you can type your questions. Please use the opportunity and ask management the questions you're looking for some answers for. Okay, let's give it a few more seconds and see if anything's coming up in the queue.
Hopefully, the lack of questions is a reflection of a very clear presentation and strategy as opposed to a lack of interest. It's always, in some ways, a good sign when we don't get too many questions because it does mean that we've got our presentation methodology dialed down.
Yeah, yeah, yeah. Absolutely. Looks like there's no questions in the queue. Charlie, Nick, what about we wrap it up from here? Thank you both for joining our conference. Hope everybody on the line enjoy the rest of your day and have a good weekend.
Thank you, Charles.
Thanks, Charles.
Thanks, everybody. Thanks, everybody. Bye. Thanks. All right. Wow, $100 million RPO. Wow. I didn't know that.
$105, yeah. $105. And you can expect this is not MNPI. We've talked about this many, many times. But you can expect that number to be higher for the fourth quarter because fourth quarter is seasonally always our strongest bookings quarter. And revenue is still ratable. So yeah, we can expect to see some more traction in that.
Yep, yep, yep. Let's definitely find a time to catch up again. I know you probably have something going on. And we'll catch up next time. Okay. All right, Charlie.
Great. Thank you. Thank you. Have a good weekend. Yeah, you too.