Operator
the rumble second quarter fiscal 2026 earnings conference call at this time all participants are in a listen-only mode at the conclusion of our prepared remarks we will conduct a question and answer session if you would like to ask a question you may press star 1 on your touch phone at any time if anyone should require assistance during the conference please press star zero at any time as a reminder this conference call is being recorded i would now like to turn the conference over to Sumit Gagneja, Chief Financial Officer. You may begin your conference.
Thank you, Operator, and welcome to the Rambus Second Quarter 2026 Results Conference Call. I am Sumit Gagneja, Chief Financial Officer at Rambus, and on the call today with me is Luke Harafin, our CEO. The press release for the results that we will be discussing today has been filed with SEC on Form 8K. We are webcasting this call along with the slides that we will referenced during portions of today's call. A replay of this call will be available on our website beginning today at 5 p.m. Pacific time. Our discussion today will contain forward-looking statements including our expectations regarding projected financial results, financial prospects, market growth, demand for our solutions, other market factors including reflections of the geopolitical, and macroeconomic environment, amongst other items. These statements are subject to risk and uncertainties that may be discussed during the call and more fully described in the documents we filed with SEC, including our 8Ks, 10Qs, and 10Ks. These forward-looking statements may differ materially from our actual results, and we are under no obligation to update these statements. In an effort to provide greater clarity on the financials, we are using both GAAP and non-GAAP financial presentations in both our press release and on this call. A reconciliation of these non-GAAP financials to the most directly comparable GAAP measures has been included in our press release, in our slide presentation, and on our website at rambus.com on the investor relations page under financial releases. I would like to note a change in how we present our results going forward. Since the adoption of ASC 606, we have disclosed licensing billings, an operational metric that bridges the difference between gap revenue and actual billings to our licenses. This was an important metric in the initial years after ASC 606 adoption, when the delta between royalties revenue and licensing billings was material. As the difference is now minimal and we expect to remain so, we will focus our financial results and guidance on an ASE 606 revenue basis going forward. The order of the call today will be as follows. Luke will start with an overview of the business. I will discuss our financial results, and then we end with Q&As. I will now turn the call over to Luke to provide an overview of the quarter.
Luke? Thank you, Samit. Good afternoon, everyone, and thank you for joining us. Before we begin, I'd like to take a moment to welcome Samit Gagneja to his first earnings call as Rambus Chief Financial Officer. Samit brings more than two decades of leadership experience in the semiconductor industry and a wealth of knowledge in the data center ecosystem. Since joining Rambus, he has quickly become a valued member of the leadership team, and we are very pleased to have him on board. Welcome to me. With that, let's turn to our results. Rambus had an excellent second quarter, delivering a new all-time high in revenue and non-GAAP earnings and beating the high end of our guidance ranges. Fueled by record product revenue and strong contributions from our diversified revenue streams, this quarter marks the first time we have exceeded $200 million in revenue. These results reflect our sustained execution and leadership across our expanding portfolio of chips and IP. We also generated solid cash flow operations, underscoring the strength of our business model and enabling us to continue investing in our product roadmap to drive long-term growth. This combination of record performance, disciplined execution, and sustained investment positions Rambos to capitalize on the exciting market trends in data center and AI. AI continues to drive a fundamental evolution in computing. As inference at a genetic use case in scale, workloads are becoming more diverse, more persistent, and more memory intensive. To support these workloads, AI infrastructure deployments are becoming more complex and heterogeneous, combining a mix of traditional and AI server platforms. This is accelerating demand for CPU-based servers to support orchestration, data management, and real-time execution at scale, while increasing requirements for memory, capacity, bandwidth, and power efficiency. These trends align directly with our strengths and are driving new opportunities for richer chip content and broader adoption of our industry-leading IP. Now, let me turn to our quarterly business results. Starting with chips, product revenue reached a new record of $99 million, up 22% year-over-year, and we expect another quarter of double-digit growth in Q3. This reflects our continued leadership in DDR5 RCDs, strong execution, and growing traction in new products. And looking ahead, we see increasing customer adoption and remain well-positioned to support the ramp of next-generation platforms as they enter the market. We continue to execute well across our DDR5 roadmap. we expanded our portfolio with complete chipsets for DDR5-9600 clients and server memory modules, further extending our leadership in high-speed memory interface solutions. Our new DDR5-9600 client chipset enables top-of-the-line performance for emerging AI PCs and leverages the same high-speed memory interface expertise we have developed across multiple generations of server platforms as technology requirements increasingly waterfall from the data center into high-performance client systems. For servers, our new DDR5-9600 RDM chipset, built around our sixth-generation RCD and PMIC 5030, supports the next level of memory performance required by advanced CPU-based server platforms. As core accounts, memory channels, and bandwidth requirements increase, solutions like these are essential to enabling higher system throughput and power-efficient performance. Importantly, our server-chip solutions support the expanding range of new and existing processor and system architectures, positioning us to benefit from increasing memory requirements across the industry. Together, these additions expand the breadth of our GDR5 roadmap and demonstrate our continued enablement of higher performance, improved signal integrity, and advanced power management across both data center and client applications. As AI workloads continue to diversify, there's increasing demand for novel memory architectures with application-specific performance, capacity, and power requirements. We are addressing these needs through products like our complete chipset for MR-DIMM and LPDDR5-XO-CAM2 and remain on track to intercept the market as these architectures gain adoption. Supported by active engagements across customers and ecosystem partners, we are expanding our roadmap of differentiated memory subsystem solutions to help shape the next generation of server modules. This reinforces our opportunity for increased chip content and sustained growth in 2027 and beyond. Turning now to Silicon IP, we delivered another strong quarter with increasing customer traction and key design wins across hyperscalers, custom silicon companies, and emerging AI semiconductor developers. As AI infrastructure scales, chip development cycles are accelerating, and performance requirements are pushing beyond industry standard specifications. Customers are building advanced SOCs for high-performance AI systems, driving robust demand for our differentiated IP solutions, spanning advanced memory, connectivity, and security IP. We also have a growing number of deep architectural engagements ahead of standards being finalized to help our customers be first to market with state-of-the-art performance. This includes an exciting design win with a Tier 1 U.S. hyperscaler for next-generation HBM in future AI chips. These engagements are a great testament to the strategic importance of our premium IEP portfolio. The growth of custom silicon for acceleration and connectivity remains an important long-term trend, particularly among hyperscalers and leading AI infrastructure companies. As customers optimize hardware for their own workloads, software stacks, and deployment requirements, they need Rambus Advanced IP to help them deliver performance, power efficiency, and reliability at scale. Secure connectivity is also an increasingly important part of the overall architecture, and Rambo's proven security IP is foundational to enabling trusted, high-performance data movements across distributed AI infrastructure. During the quarter, we also expanded our AI IP solutions with PCIe 7 switch IP, supporting 128 giga transfer per second. this solution is designed to support the next generation of AI scale-up and scale-out architectures where high bandwidth low latency connectivity is critical to overall system performance as AI infrastructure scales Rambus IEP is in great demand enabling faster more efficient and more secure data movement with our strong customer partnerships and deep architectural engagements, we are enabling the future of advanced AI hardware. In summary, Rambus delivered an excellent second quarter with record revenue and earnings. Our results reflect the strength of our product leadership, the depth of our customer relationships, and our ability to execute in markets that continue to present significant opportunities for growth. Looking ahead, we are well-positioned for the major trends reshaping data center and AI infrastructure. As AI scales and agentic workloads drive greater demand for CPU-based servers and memory, Rambus chips and IP are enabling the performance, connectivity, and security customers need to build the next generations of advanced computing systems. We remain confident in our strategy, our roadmap, and our ability to drive strong growth in 2026 and beyond. As always, I want to thank our customers, partners, and employees for their continued trust and support. Now, I'll turn the call over to Sumit to walk us through the financials. Sumit?
Thank you, Luke, and good afternoon, everyone. Before I enter the quarter, I want to say how excited I am to be here and how much I've appreciated the warm welcome from the team having spent the past several weeks meeting with our employees and investors i have come away with a clear conviction we have differentiated technology deep customer relationships and meaningful long-term growth opportunities ahead as cfo my focus is straightforward drive profitable growth through disciplined financial execution allocate capital thoughtfully and provide shareholders with transparent and consistent communication. Now, let me turn to our second quarter financial results. As I noted earlier, because the difference between royalties revenue and licensing billing is now minimal, we will focus our financial results and guidance solely on an ASE 606 revenue basis. We delivered Q2 revenue and non-GAAP earnings per share, exceeding our Q2 guidance, driven by strong contributions across our diversified revenue streams. Revenue for the second quarter was $207.4 million, which is up 20% year-over-year and up 15% sequentially, led by strong performances from our product and royalties revenue. Product revenue was $99.2 million, which is up 22% year-over-year and up 13% sequentially. Royalty's revenue was $84.2 million. Contract and other revenue was $24 million, consisting primarily of Silicon IP. As a reminder, only a portion of our Silicon IP revenue is reflected in contract and other revenue, and the remaining portion is reported in royalties revenue. Total non-GAAP operating costs, including cost of goods sold for the quarter, were $113.7 million. Operating expenses of $73.5 million were up sequentially due to higher SG&A expenses. Entrust and other income for the quarter was $6.8 million. Using an assumed non-GAAP tax rate of 16%, non-GAAP net income for the quarter was $84.4 million, resulting in Q2 non-GAAP earnings per share of $0.77, which is up 24% year-over-year and up 21% sequentially. Now, let me turn to the balance sheet details. We ended the quarter with cash, cash equivalents, and marketable securities totaling $825 million, up $39 million from Q1, with soiled operating cash flow of $61 million, partially offset by $12 million in capital expenditures and $9 million of net equity outflows. Inventory increased by $16 million during the quarter, as we leveraged the strength of our balance sheet to support future product ramps and provide customers with greater supply assurance in the coming quarter. Pre-cash flow in the quarter was $49 million. Let me now turn to our non-GAAP outlook for the third quarter. As a reminder, the forward-looking guidance reflects our best estimates at this time, and our actual results could differ materially from what I'm about to review. We expect revenue in the third quarter to be between $210 and $216 million. We expect product revenue to be between $110 and $116 million, a sequential increase of 14% at the midpoint of guidance. We expect royalties revenue to be between $69 and $75 million, and we expect contract and other revenues to be between $25 and $31 million. We expect Q3 non-GAAP total operating costs, which include cost of sales, to be between $119 and $115 million. We expect Q3 capital expenditures to be approximately $13 million. Non-GAAP operating results for the third quarter are expected to be between a profit of $91 million and $101 million. For non-GAAP interest and other income, we expect $7 million of interest income. Now, assuming non-GAAP tax rate of 16% and Q3 share count of $110 million delivered shares outstanding, we expect Q3 non-GAAP earning per share range between $0.75 and $0.82. In closing, we delivered a strong quarter reflecting the diversification of our business and contributions across the revenue streams. Our third-quarter outlook reflects continued sequential growth in both revenue and earning per share, supported by sustained momentum across the business. We remain firmly focused on driving long-term shareholder value through disciplined execution, heartful capital allocation, and consistent operational performance. Before we open the call to questions, I want to thank our employees for their continued dedication and execution, our customers for their trusted partnership, and our investors for their ongoing support and confidence in Rambus. With that, I'll turn the call back to our operator to begin Q&A. Could we have our first question, please?
Operator
Thank you. Ladies and gentlemen, if you have a question, please press star 1 on your cash fund. We request to limit yourself to one question and one follow-up. Your first question comes from the line of Kevin Cassidy with Rosenblatt Securities. Your line is open.
Yes, congratulations on the great results, and thanks for taking my question. Just to confirm, did you have any capacity issues during the quarter, any orders you weren't able to meet?
Hi, Kevin. No, we didn't have any capacity issue in the second quarter. We continue to see tightness, you know, in the supply chain. We continue to see lead times increasing, but we didn't have any capacity issue in Q2. We have built strong relationships with our suppliers, and at this point in time, we are able to serve the market demand.
Okay, great. And just because it's topical today, you know, China-based CXMT had a big splash today. Is Rambus involved with CXMT either on the IP side or product side?
Yes, it's great news for CXMT. I think they're going to be a strong player in the market. But, you know, every company that builds, you know, memory has to have a license agreement with us, and they're one of them. So we're very pleased with, you know, with the success, and that's going to be a good thing for us in the long run as well.
Operator
Your next question comes from the line of Sébastien Nagy with William Blair. Your line is open.
Good afternoon. Thank you for taking my questions. Maybe just for the first one, could you update us on your expectation for MR-DIMM, for the ramp of MR-DIMM? AMD is in production with the Vena CPU today. It sounds like servers will start shipping in Q4. Are you starting to get any better visibility into how much of the market will go down the MR-DIMM route versus sticking with more traditional R-DIMMs?
Yeah, thank you. So we do continue to see MRD as a material opportunity, but as you said, the timing is going to be dependent on the platform adoption, when the servers go to market and whether those servers, what percentage of MRD are they going to use as compared to a standard DIM. So we are excited by the opportunity, but at this point in time, we will not over-call the adoption curves before the platform actually ramps and we get feedback from the market. The contribution for Q4 is going to be minimal. We continue to ship to our customers for these early system build-ups, and a more material contribution is going to happen in 2027 when both platforms from the CPU guys rank in the market in earnest.
Got it. That's helpful. And then maybe for my follow-up, one of the concerns that we're hearing more about from investors is just the risk of potentially over-ordering in this very tight memory supply environment. Are you seeing any signs of inventory buildup at your customers, or what kind of signals are you looking at that gives you some confidence this is not happening right now?
We don't see any signs of our customers building inventory, you know, for the concerns that you expressed. This said, however, we are building some inventory on critical products that we believe are going to ramp, you know, in Q3, Q4, and early next year as we do see our lead times lengthening, you know, given the tightness in the supply chain. So no inventory buildup from our customers. We build strategic inventory for the products that we believe are going to contribute to our growth in the next few quarters.
Operator
Your next question comes from the line of Gary Mobley with StoneX. Your line is open.
Hi, guys. Thanks so much for taking my question, and let me extend my congratulations on the, you know, snapback in your product revenue. Now, relating to that, I think you've always stated, at least so far this year, that typically you see seasonal strength in the second half of the year, and that's certainly reflected in your Q3 guidance. And you're obviously building inventory, I assume, in preparation to fill strong demand. And we've also heard from the server processor supply chain that, you know, volumes continue to exceed expectations. So I'm curious to know what kind of visibility you have currently versus, say, last quarter, and what kind of visibility you have looking into the fourth quarter in terms of the continued strength.
Thank you, Gary. I think our confidence is continuing to build. One of the reasons is the use of CPU and agentic AI is certainly a demand driver, and we sense this with our customers. Our guide for Q3 shows another 20% growth year over year compared to last year, so this is a good sign as well. We want to be reasonable, though, in terms of guiding beyond one quarter for two reasons. The same reasons. One is the timing of ramp of the platforms. We hear good things, but, you know, they have to ramp. And the second one is the supply tightness. So we believe our second half is going to be stronger than our first half. We're going to see the same dynamic as we saw in prior years. But we will continue to guide quarter by quarter.
Thanks for that, Colu. As a follow-up, I wanted to ask about the Silicon IP business. If I'm not mistaken, you've been pretty steady in assuming that business is about $130 million, correct me if I'm wrong, but you've consistently and seemingly delivered upside to that number, at least through the first half of the year. So what's your most up-to-date view on the performance of the Silicon IP business, whether it be expressed in growth or dollar terms?
So we continue to see, you know, that business growing 10% to 15% a year. I would say that this is another business where our confidence in that number is continuing to build. Again, with, you know, the inference and agentic AI coming up into the market, we do see a lot of our customers building custom solutions that use our IP, whether it's on the interconnect side, on the security side, or on the memory side. So our confidence is building up, and we're confident in this 10% to 15% growth going forward. We also – it gives us comfort as well in terms of the strategy we're using. We're trying to stay at the bleeding edge of technology on these interconnect memory or security IPs, And that allows us to engage with customers very, very early and gives us a longer-term visibility into that growth. So, again, I would say we still see that business growing 10% to 15% a year, but our confidence in that growth continues to grow. We had a great quarter in Q2, in particular, as you could see.
Operator
Your next question comes from the line of Aaron Rekers with Wells Fargo. your line is open.
Yeah, thanks for taking the questions. I guess my first question is, you know, earlier was asked about MR-DIMS, but there's just a lot of architecture stuff going on in memory or the memory subsystems in general. I'm curious, Luke, as you think about MR-DIMS and you maybe juxtapose that relative to, let's say, CXL, what is the company's views on CXL now Now that we've seen meta endorsed the technology, there's kind of other inklings that other hype.
That's the excitement around them already. Now, when it comes to CXL, we are very supportive of CXL as a very important Internet protocol. It's not a chip, it's an interconnect protocol, excuse me. And I think it's going to play a role, you know, in the AI evolution. in particular in agentic AI, is going to play a role on managing the memory stack or the memory pyramid and moving from cold memory to hot memory. But that's still, I would say, an interconnect protocol and not a product. So it remains very, very relevant to our silicon IP business. Now, at the product chip level, because we actually talk to the people who build those products, We continue to see a fragmented market from a product standpoint with many deployments looking like ASIC-like or customer-specific products. So our positions, you know, with respect to CXL remains the same. We will continue to enable the ecosystem with our IP engagement, and that's why we're building confidence on our IP business. And we'll continue to monitor the traction there. But we will continue also to focus our product investment where we see, I would say, the strongest market opportunity for scaling, and in particular on standard products. So, again, we play a critical role in the deployment in the ecosystem for our IP business, and we're monitoring the product business. But at this point in time, we see this as a custom ASIC business that is fragmented for us. And from the product standpoint, we'd rather invest into standard products at this point in time.
Yep, that makes a lot of sense. I appreciate that. And then as a quick follow-up, I know you referenced it in your prepared remarks that you were engaged with a hyperscaler on some of the IP and some future generation, I'm guessing, XPUs or programs that they have in place. I'm curious, is that changing, like that opportunity set of your business? Is that necessarily a new dynamic or any thoughts on hyperscalers being direct, like real customers and driving some incremental growth for Rambus?
Yes, Aaron, that's a very good question. I think the trend we see is that, you know, hyperscalers are playing a growing role in defining their own architectures, whether they build the products themselves or whether they have ASIC companies or product companies building the products for themselves, because, you know, they want to stay competitive and move fast. So their role in defining the architectures on complex subsystems, like the memory subsystems, is becoming more and more important. So they work very early with us, for example, before even the specifications are complete, to make sure that we can meet the system requirements. And once this is done, they can use that to either build their own products or actually have semiconductors building their own products against those high-end specifications. And the trend that I see here is that these technologies actually proliferate. You know, once a hyperscaler has decided upon a particular implementation of a memory controller, for example, then that proliferates into their own ecosystem. And that, again, is one of the reasons, you know, we feel confident in the growth rate of our IP business. Thank you. Thank you, Aaron.
Operator
Your next question comes from the line of Kevin Garrigan with Jeffries. Your line is open.
Yeah. Hey, team, thanks for taking my questions, and congrats on the results. Hey, I may have missed it, but can you just talk about how much of your revenue was from companionships or new product revenue this quarter?
We continue, like, in the first quarter, you know, we indicated that, you know, our products, I would say these new products, were in a low double-digit percentage of the product revenue. We continue to be at that type of rate, and we continue to run those products. into the market, and it's going to be, you know, in the double digit by the end of Q4. So we are on that trajectory. And remember, this is on a growing, you know, revenue base for the product side. It actually is growing quite nicely, but it has to go through the qualification process with our customers, with our customers' customers, and ecosystems. So it's never going to be a step function, but we do have momentum there, you know, across the board, and we're happy with the performance of those products.
Yeah, okay, great. That makes a ton of sense. And then so I get a lot of questions about, you know, just LPDDR-based servers, and, you know, you guys now have your SOCAM2 chipset, and I believe just SOCAM in general has lower content overall versus RDIM and MRDIM. But, you know, as the industry kind of shifts or potentially shifts towards more LPDDR-based server modules, I mean, does that kind of cannibalize your RDIM or MRDIM opportunity at all?
That's a great question. I wouldn't say that, you know, the industry is shifting to LPDDR. I think, you know, LPDDR is actually an incremental opportunity, you know, for servers. We believe that DDR will remain dominant where, you know, server, grade scale, capacity, reliability, serviceability are required. So that's going to be continuing to be dominant, you know, in the server space. But LPDDR and SOCAM have a role to play where power efficiency is really, really important. So we see this as complementary. You know, we talk a lot about, you know, the AI market becoming heterogeneous. This is one aspect of that. So our SUCAN 2 gives us, you know, a seat at the table. You know, we have a chipset for the current generation. To the extent that LPDDR is adopted more in the future with future generations, we will continue to develop chipset there. And I think the content is going to continue to increase, you know, as the complexity increase. So we do see this as an opportunity. We said in the last call, you know, the revenue outlook in the short run is modest, but the strategic importance is really high for us. Thank you.
Operator
Your next question comes from the line of Tristan Gare with Baird. Your line is open.
Good afternoon. The 20% year-over-year increase in product revenue guidance that you provided, is that a good reflection of the unit demand that you see for X86 CPU in light of AMD provided or raising their X86 CPU CAGR to a 50% over the next several years? but I understand this includes pricing. So is 20% kind of a good proxy in terms of units that you expect, you know, for CPU? And then on top of that, you're layering additional channel count.
Thank you, Tristan. Yes, it's good that you remind that, you know, the way we look at our business is unit-based, more than dollar-based, as we do not see the same pricing dynamics than the CPU or the memory guys have. That's the nature of a standard product business. But if you look at our business, we grew 20% year over year. Next quarter, we're going to see the same growth type of growth year over year. And in the first quarter, despite the manufacturing issue we had, You know, it was 15% higher than the same quarter earlier. So, you know, we are on that trend. The server market, you know, view in terms of unit has changed positively. You know, I would say last quarter we would say it was mid to high single-digit growth. You know, now it's double-digit growth. You know, Gartner mentioned 12% growth. So, you know, we believe that we're growing faster than that, And it's coming from, you know, a combination of the channel counts, but also the initial contribution of new products. Remember, on the channel counts, we always make the same reflection. It's a great trend for us, but it's not a step function. We had, you know, AMD at 12 channels, then Intel moved from 8 to 12, then the whole market is going to move to 16. So all of that are pointing in the right direction, and the secular trend is really, really good, but that's not a step function. So I would say that we're growing faster than market. We continue to believe we're growing faster than market, and all of these factors come into play.
Okay, great. And then as my follow-up for next year, do you think that we could see an acceleration from that 20% year-over-year growth? you know, given the dynamic that you've mentioned. Is that something that you would be able to get sufficient supply? And then if you could also talk about any potential mixed changes that you're seeing, you know, and anything that could impact, you know, ASPs given the supply constraint in DRAM and the potential this, you know, this has in terms of DRAM content and CPU usage.
So we, as we said earlier, you know, we don't guide beyond, you know, the current quarter. The data is so dynamic. But I would say there are a few things to take into consideration when we look into 2027. I think the Gen 5, DVR 5 is going to grow in earnest. So, you know, that's the time where the market will have moved to 16 channel per, you know, per CPU. So that's a good thing. This is the trend we were talking about. This is also, as we said earlier, when MRDM is going to start to kick in in the market. So that's another good thing. And we continue to see growing contributions from our companionship and the client space. So from a demand standpoint, the environment is very positive when we look at 2027. You know, this being said, the supply constraints will continue to be there in 2027. You know, when we talk to our suppliers, you know, we work with them, that's the situation that is going to last with us, you know, for some time now. So we have to take this into account when we look at the potential of our business and as well as, you know, the platform timing. You know, by experience, we know that, you know, platform, you know, ramps typically take a little bit longer than what people anticipate. So, you know, when we look and we guide, and again, we cannot guide beyond one quarter, but when we look at the business, we feel very comfortable with the underlying assumptions on the demand side, but we are prudent with respect to, you know, platform timing and supply in particular. You know, at this point in time, because we have standard products, we don't see any opportunity, I would say, for price increases because this is what you were talking about. But, you know, we want to stay competitive and maintain or continue to increase our share in the dim market.
Very useful. Thank you very much.
Operator
Your next question comes from the line of market passes with Evercore ISI. Your line is open.
Hi. Thanks for taking the questions. First question is, you know, I think there's a framework to think about CPUs ramping in data centers along three dimensions. One would be CPU head nose next to the GPU or accelerators. One would be, you know, CPUs kind of standalone agentic AI CPUs, and then the third one would be, you know, CPUs in standard server configuration supporting, you know, legacy workloads like database. Is there a – should we think about a different framework for your silicon content opportunity in either these three categories, or is the MR-DIM opportunity, does it ramp more obviously in one of these versus the other? That's the first question, then I had a follow-up.
Yeah, that's a good framework to look at this. I would say that, you know, every segment that you described have their own, you know, requirements. You know, in head nodes, you know, we see sometimes the emergence of or where people are starting to look at, you know, very high bandwidth, low power. You know, that was one of the driver for the thoughts around SOCAM. You know, in AI servers, you know, we typically see them as a catalyst for the adoption of the fastest technology and the highest, I would say, configuration in terms of capacity. So if we put the questions of platform RAMs and DRAM pricing on the side for a moment, that could be a good candidate for MRD types of solutions, close to the GPUs, HDM, where you need a lot of memory there. That could be an option. And then standard servers, I would say whether they are used for a legacy or agentic AI, would have more standard solutions. In agentic AI, the latency is becoming very, very important. You have to build the key value cache, but then once you have to pull from that key value cache, you need to be very, very fast. The latency becomes very, very important, and we see those servers actually using the maximum number of channels, not necessarily with the highest, I would say, capacity to maintain that, you know, latency smaller or shorter. So one of the strengths we have is, you know, we have a good understanding of those tradeoffs, whether it's through our product business or, you know, our IP business. And if you look at, you know, our roadmap, we're trying to have solutions for each one of those segments. The question we have, as usual, is we have to understand the ramp profile of each one of them, as well as, I keep saying, the supply constraints we're going to have in 26 and 27.
Okay, gotcha. That's a very helpful framework, Luke. And then the design with the hyperscaler for next-gen chip, just to be clear, This is a – is this a product design win for you, or is this IP?
It's an IP design win for a company that designs a product. Gotcha. It's someone building a SOC, if you wish, or providing a spec for SOC for others to build, and we provide critical IP in that SOC. But we see that trend, you know, with the requirements of AI as it moves to agentic AI, the requirements for high speed or the best performance, I would say, are accelerating. And, you know, this is a trend that we're seeing that we're talking more and more directly to the hyperscalers and develop with them the architecture, and then it proliferates into people building the silicon. But this is an IP win.
And is that – would that be a royalty-based opportunity? for you or licensed by like most of our silicon IT business it's a it's a license or multi-license meaning that you know anyone who's going to use that architecture in any product you know well we have an opportunity for a license so it's not volume based and typically the volumes might not be necessarily high you know but it's a license based which is typical with our silicon IT business and the silicon IT Okay, that's very helpful.
And then the last question, you mentioned the PCI Express Gen 7, I believe, IP. What is the timeframe for seeing revenues from that product?
So it's a similar business model as the one we talked about, about, you know, HDM controllers. This is, again, a similar trend where customers are working with us ahead of the specifications being finalized or, you know, as the specifications are being finalized. So it's a license opportunity for us, which we will see, you know, very quickly, I mean, in the coming quarters. Because it's a licensing business, right? It doesn't need to ramp in the market. It's very well ahead of the end products ramping in the market.
So you could get licensed revenues well ahead of the end product shipping from that.
Yes, and that's typical for our IP business. We engage very early. We get the license as we engage. And then, you know, our customers build their chips, and it can take them 12, 18, 24 months before the product actually goes into the market. But we see the revenue much earlier than that. But what we see as well is the trend because we understand what people are building and why they're building it, and that gives us a very good insight as to where the market is going.
All right, very helpful. Thank you for all the insight. We appreciate it. Thank you.
Operator
Your next question comes from the line of Mehdi Hoseni with SIG. Your line is open.
Yes, thanks for taking my question. All the good questions have already been asked. I just have a couple of follow-ups, starting off with Luke. I look at the slide number seven, and it's very exciting that the chipset, especially for memory interface, is diversified. But what I wanted to ask you is, how do you see or what gives the confidence that this combined with additional silicon IP is going to help you with a growth acceleration. We have gone through the DDR5, and you have done a great job of carving out market share in SPD companionship. But as I look into next year, and I think about agentic AI, an arm-based solution, where a number of channels per CPU is not really high priority. At the same time, you have all of these exotic chipset architecture coming to the market. What is it that you see that will give you the confidence that you can actually grow revenue at a high rate and have a follow-up?
Yeah, thank you, Mehdi. I think, as we said earlier, we believe that we have a very strong secular setup for our business. If you look at it and you look into next year, the market will be entirely DDR5 as we move from that transition from DDR4 to DDR5. We continue to see an acceleration of the DDR5 sub-generations, which gives us additional opportunities to grow share on the core business. In the prepared remarks, we talked about introducing Gen 6. Your Gen 5 is not in market yet, and we're introducing Gen 6 for after that. So every generation gives us an opportunity to gain share. You know, on the companionships, we have a great growth opportunity there. You know, we talked about, you know, increasing the percentage of revenue from our companionship, but there's still a lot of room to increase that in 2027. So that's another vector for us. MRDM with four times the silicon content on the module is another vector. And we're starting to see more and more platforms on the client side. So all the seeds that we have planted over the last two years are actually going to grow into something quite solid in 2027. So I'm very confident in the setup from a demand standpoint. Now, if you look at the Silicon IP business, although this is a license-based business, not a volume-based business, we do see this trend with hyperscalers defining their own products with advanced IP, which is also a source of growth for us. So I am confident that we can grow. I know I said it out, but I'll say it again. I think the challenge next year for the industry, not only for us, is going to be the tightness of the supply chain. But we're working with our suppliers to address that as early as we can.
So if part of the strategy is to increase market share, does that mean that your product revenue gross margin is actually going to remain in the low 60%? Because that's what's been a trend despite double-digit product revenue. The gross margin is in the low 60. So is there a trade-off here?
Our model remains 60% to 65%, and we do see fluctuations from quarter to quarter. And, you know, we like to see, you know, the product margin looked at on an annual basis at the end of the year because with, you know, short-term supply constraints, mix, and all of that, it can fluctuate from quarter to quarter.
And if I may add to that, Luke, you covered it, just to reinforce that, you know, on a quarterly basis, you may see that our gross margin may fluctuate based on product mix and other factors. But recently, as you know, we've been operating in the 60% to 63% gross margin, but our long-term model of 60% to 65% remains intact.
Got it. Thanks for the details.
Operator
At this time, there are no further questions. This includes the question and answer session. I would now like to turn the conference back over to the company.
I'd like to thank everyone who has joined us today for your continued time and support, and we look forward to speaking with you again soon. Thanks, everyone.
Operator
This now concludes today's conference.
