Welcome to the Rambus Third Quarter Fiscal Year 2023 Earnings Conference Call. [Operator Instructions] As a reminder, this conference call is being recorded. I would now like to turn the conference over to Desmond Lynch, Chief Financial Officer. You may begin your conference.

Thank you, operator, and welcome to the Rambus third quarter 2023 results conference call. I am Desmond Lynch, Chief Financial Officer at Rambus; and on the call with me today is Luc Seraphin, our CEO. The press release for the results that we will be discussing today has been filed with the SEC on Form 8-K. A replay of this call will be available for the next week at 866-813-9403. In addition, we are simultaneously webcasting this call. And along with the audio, we are webcasting slides that we will reference during portions of today's call. A replay of this call can be accessed on our website beginning today at 5:00 p.m. Pacific Time. Our discussions today will contain forward-looking statements, including our expectations regarding projected financial results, financial prospects, market growth, demand for our solutions, the company's ability to effectively manage supply chain shortages and other market challenges and the effects of ASC 606 on reported revenue amongst other items. These statements are subject to risks and uncertainties that may be discussed during this call and are more fully described in the documents we file with the SEC, including our 8-Ks, 10-Qs and 10-Ks. 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 in 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. We adopted ASC 606 in 2018 using the modified retrospective method, which did not restate prior periods but rather run the cumulative effect of the adoption through retained earnings as a beginning balance sheet adjustment. Any comparison between our results under ASC 606 and prior results under ASC 605 is not an accurate way to track the company's progress. We will continue to provide operational metrics such as licensing billings to give our investors better insight into our operational performance. The order of our call today will be as follows: Luc will start with an overview of the business. I will discuss our financial results, and then we will end with Q&A. I'll now turn the call over to Luc to provide an overview of the quarter. Luc?

Thank you, Des, and good afternoon, everyone. We delivered another strong quarter with revenue and earnings above the midpoint of guidance as we continue to execute on our strategy and successfully navigate the complexities of the industry transition to DDR5 in a challenging macro environment. The company generated $52 million in cash from operations, enabling consistent return of value to our stockholders, most recently with the completion of a $100 million accelerated share repurchase program. We also closed the sale of our IP business strengthening our focus on the development of differentiated chips and digital IP that expand our opportunities in the data center market. Generative AI and other data-intensive workloads continue to drive increasing requirements for memory performance and capacity across the computing landscape. This is a very positive long-term trend for Rambus. Currently, the training of large language models is boosting the demand for AI servers with the most advanced multi-core CPUs provisioned with DDR5 rams to maximize main memory bandwidth alongside server GPUs with dedicated HBM memory. In addition, high-performance general purpose servers enabled with DDR5 are seeing increasing demand to meet the growing computing infrastructure requirements of the AI data pipeline. With increases the opportunity for our expanding family of memory interface chips. As the industry builds out the infrastructure for the broadening adoption of AI, we look forward to continued innovation and growth in server CPUs as well as workload optimized accelerators. This trend also creates opportunities for our silicon IP business. The ongoing specialization of computing systems makes our high-performance CXL, PCIe, HBM and GVR IP controller increasingly critical. In addition, the move to application-specific silicon driven by AI and other advanced workloads, creates increased vulnerabilities to attack as data is distributed across systems. This trend increases the need for advanced security IP an…

Thank you, Luc. I'd like to begin with a summary of our financial results for the third quarter on Slide 5. Once again, we delivered a strong quarter, and we are very pleased with the company's continued execution on our strategic initiatives to drive long-term profitable growth. We delivered strong financial results with both revenue and earnings above our expectations. In Q3, we executed a $100 million accelerated share repurchase program, which retired approximately 1.85 million shares. Our continued strong cash generation allows us to consistently return cash to shareholders. As Luc discussed in Q3, we completed the divestiture of our PHY IP business, which will enable us to redeploy our investments into higher growth areas of products and digital IP. Let me walk you through our non-GAAP income statement on Slide 6. Revenue for the third quarter was $105.3 million, above our expectations, driven by higher product revenue in the quarter. Third quarter revenue included approximately $5 million in revenue by IP business that we divested in early September. Royalty revenue was $28.9 million, while licensing billings was $57.9 million. The difference between licensing billings and royalty revenue mainly relates to timing as we do not always recognize revenue in the same quarter as we bill our customers. Product revenue was $52.2 million, consisting primarily of memory interface chips. Contract and other revenue was $24.2 million, consisting predominantly 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 royalty revenue as well as in licensing billings. Total operating costs, including cost of goods sold for the quarter were $72.9 million. Operating expenses of $52.4 million were in line with our expectations and down $3.5 million versus Q2 as we continue to…

Our first question comes from Gary Mobley with Wells Fargo. [Operator Instructions]

I want to start out by gaining an appreciation of the undercurrents in the product revenue between DDR4 and DDR5. Can you verify if your DDR5 DIMM chipset sales grew sequentially in the third quarter? And how low the R4 related revenue may have trended for the quarter? And contrast that against what you think would be a normalized quarterly shipment number for DDR4 that may eventually layer on top of the growth in DDR5?

Gary, I'll take the answer. I think Des -- thanks for the question. We've been very pleased with our execution in 2023. And as Luc mentioned in his prepared remarks, product revenue is up 7% on a year-to-date basis through in a market which was down double digits. So we've successfully navigated a dynamic environment by posting solid results, which will result in year-over-year share gains versus our competitors. Looking at our product mix in Q3, the shipments were predominantly DDR5. And for the second quarter in a row, we shipped out minimal DDR4 products in the quarter. I really comparing Q3 versus Q2, DDR4 shipments were down which was offset by modest growth in the DDR5 shipments from there. Going into Q4, we do expect to see a similar product mix with continued growth in DDR5 in minimal DDR4 shipments in the quarter as customers continue to take a conservative posture towards inventory management at year-end. So overall, we've been very pleased with our execution in this sort of challenging environment, and we're really pleased to see the continuation of growth in DDR5 shipments, which is positive for us.

Okay. A follow-up to the product sales mix. I noticed that the product gross margin was well off of the strong performance you had in the second quarter despite the higher mix of DDR5. Maybe if you could just speak to why the negative variance relative to that long-term view, what, 60%, 65% or relative to the prior quarter? And maybe if you could just give us a sense of the different moving pieces there between DDR4 and DDR5?

Yes, Gary, great question. We are very pleased at how we continue to manage our product gross margins as a company. And if you look at our year-to-date product gross margins again through Q3, our gross margins in the products are around 63%, which is in line with the midpoint of the communicated long-term product gross margins of 60% to 65%. If you look at specifically in Q3, our product gross margins were 63%, which were down from the high of 66% in Q2, mainly driven by ASP erosion as the DDR5 products are now shipping in volume production. The ASP erosion was anticipated and in line with our expectations. And as a company, we'll continue to be disciplined in our ASP management, and will continue to drive product cost savings to offset any ASP erosion from there. Again, with the similar product mix going into Q4, as I mentioned earlier, we would expect product gross margins to be relatively flat at the 63% in Q4. And we're very pleased at how we've been able to execute on the gross margin performance, which will be roughly in line with the midpoint of our long-term gross margin range of 60% to 65%.

Our next question is from Mehdi Hosseini with SIG.

Yes. Can you please provide us an update where we are with companionship. I believe last time we're expecting qualification with a high-volume manufacturing by mid-24, and is that correct? Can you give us an update? And I have a follow-up.

Thanks, Mehdi. Yes, we're pleased with the investments we're making in our companionship rollout. We are actually currently shipping in small volumes, our SPD hub and temperature sensor to the market. And as you say, we believe to have more contribution from these products towards the second half of 2024. We have also sampled our customers with our initial power management chips and the initial feedback from our customers is really, really good. And you should expect some announcement next quarter around these products. And again, I think they will contribute to our revenue starting in the second half of 2024 and into 2025. And finally, although these are not companionships to the data center, we're also working with our customers on a set of client products in the clock space or the power management space. which will contribute to revenue in 2025 and beyond. So we have a rollout of products. We're pleased with the progress. We started to ship the SPD hubs and temperature sensor with ample and we're working on the client rollout of products. And all of these should start contributing substantially more substantially in the second half of 2024.

Great. And 1 follow-up for that. your Q4 guide implies OpEx of $50 million, taking the midpoint. And what I want to better understand is -- how should we think about the scaling of OpEx once the revenue starts to grow once DDR5 RC or buffer chip start shipping. Can you manage the business with the $50 million to $51 million of OpEx as you scale revenue? Or would the OpEx would need to increase?

Thanks for your question. I think as a company, we've done a very nice job in managing our expenses given the softer macroeconomic outlook. In Q3, our operating expenses were around $52 million were down from $58 million in Q1, which shows a discipline and focus on managing our expenses. In Q4, you are right that our OpEx spend will come down roughly to $50 million to $51 million for the quarter. What I would say is that we've highlighted when we divested the PHY IP businesses that we talked about reinvesting some of the R&D back into product programs, which will drive revenue into 2025 and beyond. And Luc talked about some of the client opportunities earlier. Looking at 2024, and what I would say from an R&D perspective, our spend historically has been around 23% to 25% of revenue. And I would expect to be within that sort of range going forward. In SG&A, we'll continue to be disciplined with inflationary type of increases here. And what you will see is some very nice leverage as we continue to grow the sort of, I think overall, we've managed our OpEx very well, which has really struck the right balance of being prudent to our short-term expenses and also balancing the need to invest in the long-term opportunities from there. So that's how we see the OpEx playing out, Mehdi.

Our next question is from Kevin Cassidy with Rosenblatt Securities.

Congratulations on the strong quarter. Just as we talk about the DDR5, and I think you had mentioned sampling or qualifying the next generation. Is there an opportunity, you had mentioned about ASP are with the next generation, does it -- is there a reset on ASP as you go to Gen 2, Gen 3 and so forth?

Kevin, thanks for your question. What we see is that when we look ahead to the next sort of generation of we will see that ASP reset with each generation the generations of DDR5 are coming around quicker. They seem to be on a 12-month cadence just now compared to the sort of 24 months cadence under DDR4 that will offer the opportunity of an ASP reset from there. What I would say is that we've been very disciplined in our approach to sort of pricing and we have a really good track record of producing healthy product gross margins. And as I mentioned earlier, this year, we're going to our product gross margins to be around 62% to 63%, which is in line with the midpoint of our long-term product gross margin range of 60% to 65%.

Okay, great. And just on the HBM3 device -- or that IP, are you recognizing revenue for that? Or did you recognize revenue in the September quarter? Or is it part of your December quarter guidance?

HBM3 announcement we've made is an IP that we've just announced at the top end of speeds in the 9.6-giga tranfers per second. So this is a product to come that will contribute to revenue in the future. It's not a product that has generated revenue this quarter or next quarter.

Okay. I see. I wasn't sure if it was -- if you had a beta customer that already had developed.

Our next question is from Nam Kim with Arete Research.

I have 2 questions. One, I think the market shift to DDR5 seems happening much faster than expected. Can you share your latest view on DDR5 crossover timing? And then second question, also demand for DDR5 Gen 2 is picking up recently. And also you mentioned the PMIC qualification on DDR5 Gen 2. Can you give us some color on how you can compare with the existing big PMIC suppliers such as TI, Samsung, MPS, what's your selling point really here versus others?

Thank you for your questions. So we still see the crossover in the market to happen in the first half of 2024 for servers. We do see a slow burn of DDR4 inventory at our customer side, but it's still a slow burn. Obviously, as far as we are concerned, our shipments of DDR5 compared to DDR4 into the market has passed that crossover point, as you noticed in our Q2 and Q3 results, but the cost of the point in the market is going to happen in the first half of 2024. That's still our view DDR5 Gen 2 is gaining some momentum. We are shipping some volumes to our customers as they buy them. We will go through the same process and cadence as with other generations. People are going to prebuild systems, settle those previous systems to the market. we're going to go through the standard qualification process. And we still believe Gen 2 will start in earnest in the second half of 2024. That's for Gen 2. With respect to your question regarding PMIC, PMIC has been a challenge for the ecosystem in general in that first generation. This is -- this has explained some of the hiccups of the ramp of DDR5. We have built a team for PMIC some months ago. and have worked on a solution. We have sampled the solutions to our customers. And the feedback from our customers is very positive at this point in time. So I think -- the initial feedback is about the quality and robustness of the solution. Now of course, we have to go through the standard process of validating that solution into the market. And as I said earlier in the prepared remarks, we expect to hit the market towards the second half of 2024 with that product, but we're actively working with customers as we speak.

Our next question is from Sidney Ho with Deutsche Bank.

Great. It sounds like the timing of the DDR4 inventory digestion hasn't really changed much from a quarter ago. What gives you that confidence that timing is not moving out? Any tangible data points can you can share with us? And a follow-up to that is if you look at 2024, not looking for guidance here, but what should we think about the split between DDR4 and DDR5 in product revenue for the entire year?

Sidney, thanks for your question. With regards to DDR4 inventory at our customers that did come down in the September quarter, which is the second quarter in a row. We are encouraged to see that all customers saw an inventory decline in the September quarter versus the June quarter. from there. And really from our discussions with customers, we do expect the inventory digestion to be substantially complete by year-end. It really is a sort of a fluid situation just now and we're really working with the customers on the timing of the DDR4 reordering pattern, which we expect to take place across Q1 and Q2 next year. But our visibility is limited with regards to this just now, but we do expect that DDR4 will continue to have a long tail of demand from there. I think the second part of your sort of question with regards to product revenue sort of growth and the timing of DDR4, DDR5 for next year. As a company, we only guide sort of 1 quarter at a time, which is prudent, especially given the macro uncertainty just now. We've been very pleased with our execution in 2023 and really assuming the midpoint of our guidance for Q4, our 2023 product revenue relatively flat at $226 million, which really provides us a foundation for us to grow next year. And it is important to put this performance in context because the market has declined double digits this year. And we do expect to grow market share in 2023 versus the competitors. As it relates to 2024, we do anticipate growth, which is in line with all the industry research that's out there just now. And we're excited about our competitive product offerings will be the continuation in growth of DDR5 platforms. We do anticipate the growth of recovery of DDR4. And as Luc mentioned, we expect to see the greater contribution from the companion chips in the second half next year. So overall, our products are very well positioned in the market. We do expect to grow faster than market next year and to see continued share gains in 2024.

Okay, that's helpful. Then my follow-up question is you guys touched upon this in the prepared remarks, but it sounds like you guys are very well positioned to benefit from AI. Can you help us summarize the different ways that you can benefit from that ramp both from a product and licensing standpoint? I understand the memory licensing business is kind of fixed. So I just want to make sure to understand the opportunity.

Yes, so if you look at the 3 pillars of our business and of course, our patent licensing business is not affected by AI. On the product side, let's start with the buffer chips. AI servers do actually some general purpose servers in the AI boxes. And typically, those general purpose servers have a high memory content and typically DDR5. So seeing the impact of AI, especially this year, although the market was a bit depressed in total, the positive impact of AI has been to accelerate the demand for DDR5 modules. And this, I think, explains partially the profile of our DDR4 DDR5 mix between Q1, Q2 and Q3 of this year. With respect to our business. What we see with AI is that we see the emergence of specialized compute nodes, disaggregated architectures. So all of these chips have to communicate between themselves. So CXL and PCIe IP become very, very important for our silicon customers as they build these heterogeneous chips that go into the new architectures. GDDR and HBM, of course, that's why we announced our next-generation HBM at very, very high speed. We want to stay ahead of the curve there. And finally, as we mentioned in the prepared remarks, with all of these specialized chips now in the data center, the vulnerability with respect to attacks on data at rest or actually data in motion between ships is becoming more important. And therefore, our security IP portfolio is becoming more relevant to the market. So as we sold our high business to Cadence, we said we would continue to invest in our IP portfolio because we do see the opportunity brought by with all the semiconductor companies that actually build chips for that market.

Our next question is from Mehdi Rossini with SIG.

Yes. So just a quick clarification. Look, I just want to go back to your prior statement. And this comes up every earnings conference call. Would it be fair to say that your product revenue is mostly driven by the number of temp and not necessarily with the DDR5 number of bits. In other words, your business is units of HBIM driven, not bits of DDR5. Would that be a fair statement?

That's correct. There are several factors going into the capacity equation if you wish. One being the density of the memory itself. But the use of commerce, as you rightfully say, for our buffer shape is really the number of [indiscernible].

Yes. And then there are a number of new service CPUs coming out each with a different number of channels for CPU and number of DIM per channel. And in that context, it will take a little bit of time for the end customer to come up with a set of -- the configuration of what really could meet their demand. And this is why as we go through this transition, that inflection point is going to come down to when the CPU is going to be available and how end customers going to configure the system based on a severe number of DIMs per CPU, fair?

That's correct. The volume of buffer sheet depends on the number of teams. Therefore, it depends on the number of channels and the number of teams per channel, and all of those deployments depend on the end customers. But this being said, the trend is really for capacity expansion. So there's this trend in the market that people are trying to populate as many channels as they can and put as many things per channel as they can because we still have to feel that gap between the compute power from the servers and the memory capacity and memory bandwidth supporting that. So the trend is definitely towards higher volume. The way it's being deployed in terms of when the number of channels are populated and how many teams per channel being used depends on the specific customers.

There are no further questions. That concludes the question-and-answer session. I would now like to turn the conference over to the company.

Thank you. I would like to thank you all for your time and your interest, and we'll talk to you later. Bye, bye.

That concludes the conference call. Thank you for your participation. You may now disconnect your lines.