Netlist Inc

Good afternoon, and welcome to the NETLIST third quarter 2024 earnings conference call and webcast. All participants will be in listen-only mode. Should you need assistance, please signal a conference specialist by pressing the star key followed by zero. Please note, this event is being recorded. I would now like to turn the conference over to Mike Smarjasi, investor relations. Please go ahead.

Thank you, Gary, and good day, everyone. Welcome to Netlist's third quarter 2024 conference call. Leading today's call will be Chuck Hong, Chief Executive Officer of Netlist, and Gail Sasaki, Chief Financial Officer. As a reminder, you can access the earnings release and a replay of today's call on the investor section of the Netlist website at netlist.com. Before we start the call, I would note that today's presentation of Netlist results and the answers to questions may include forward-looking statements, which are based on current expectations. The actual results could differ materially from those projected in the forward-looking statements because of the number of risks and uncertainties that are expressed in the call, annual and current SEC filings, and the cautionary statements contained in today's press release. NETLIS assumes no obligations to update forward-looking statements. I will now turn the call over to Jack.

Thanks, Mike, and hello, everyone. Netlist delivers strong revenue growth in a quarter, driven by the turnaround in the memory market from a year ago period. We continue to work with our partners and leverage our five-year, $600 million supply agreement with SK Hynix to bring new value-added products to our customers. We recently introduced an ultra-low latency memory product line called Lightning for workstations and servers, that delivers double-digit percentage improvement in memory performance without any changes to the system, both AMD and Intel-based. This allows customers with demanding workloads and AI inference, high-frequency trading, and in-memory database to accelerate application performance with minimal additional costs. With a projected 2025 DDR5 server market of greater than $25 billion. The market for ultra-low latency RDEM memory is expected to exceed $250 million starting next year. We're also developing our next generation CXL NVDEM, or non-volatile DEM, that we invented more than a decade ago. The CXL NV is a persistent NAND memory solution that allows server and storage customers to back up critical DRAM data within seconds in the event of a power outage. Although many customers that utilize first-generation NV DEMs shifted to an Intel product called Optane, that product is now end of life and will stop shipping at the end of 2025. So these customers cannot implement rack-level battery backup systems to safeguard the data that must be stored at DRAM speeds. So they therefore are looking to utilize CXL NVDEM and their new platforms for backup. The CXL NV market opportunity is expected to start next year at around $200 million and grow from there. Moving to IP. NETLIS continues to make significant progress. As a reminder, NETLIS has secured nearly three quarters of a billion dollars from two patent infringement jury awards in the past 18 months, an amount that is among the largest in the semiconductor industry during this period and highlights the tremendous value of NETLIS's intellectual property. Each of these awards was granted as a reasonable royalty for infringement for past damages period of around only 18 months. Given that Samsung and Micron continue to infringe these and other netless patents, and there's been explosive production ramp of these covered products, there is significant and growing exposure. We believe these dynamics place us in a strong position to monetize these verdicts. On November 12th, Upcoming, we will commence the jury trial in the Eastern District of Texas for patent infringement in a second case against Samsung, which was first brought in late 2022. In this case, Netlist is asserting that Samsung has infringed three Netlist patents. The 912 and the 417 patents are related to the infringement of large volumes of DDR4 RDEMs and LRDEMs. These are the same patents that were the subject of the May 2024 Micron trial, where the jury handed down $445 million in damages award to Netlist. Netlist is also asserting the 608 patent in this case, which reads on LRDEM technology, but in a different way than the 417. The 608 patent was already found valid in the IPR filed by Micron in the currently state Western District of Texas case. Samsung has filed its own IPR against the 608 patent, and the final written decision from the PTAB on this IPR is due early December. In NETLESS's first Eastern District of Texas case against Samsung, in which the trial took place in April 2023, the court recently upheld the jury's verdict and award of $303 million in damages, confirming that Samsung willfully infringed Netlis' patented technologies and that none of the asserted claims are invalid. With a court order denying post-trial motions and the final judgment entered in August 2023, This case is closed in the district court. As expected, Samsung has filed an appeal of this ruling with the Federal Court of Appeals. Opening briefs are due in December, with final briefs currently scheduled for February 2025. Importantly, nutless patents asserted in this first Samsung case cover high bandwidth memory, or HBM, and DDR5 memory, both critical to generative AI. and continues to grow at an exponential rate. Industry analysts are expecting DDR5 memory to reach nearly $30 billion in annual revenues in 2026 and HBM to hit $80 billion in annual sales in 2026, a nearly fourfold increase from 2024. Now, moving on to appeals. As the district court verdicts and IPR decisions move into the appellate phase of litigation, some of these cases are now in front of the Federal Circuit Court of Appeals. Although the IP enforcement campaign may seem endless, it is a well-defined, finite process and there is an end. We are now at the beginning of the end phase in some of these cases. The front end of the campaign, which takes place in federal district courts and the USPTO, can take up to three years or more to fully wrap up, and they represent the bulk of the time and money spent in the campaign. By contrast, the appeals at the Federal Circuit usually take about half the time and fraction of the cost. We currently have eight IPRs in the pipeline. I'm sorry, we currently have eight IPR appeals in the pipeline involving 11 netless patents. For each appeal, the Federal Circuit process is expected to take approximately 18 months. We expect oral hearings in six of these appellate cases to be conducted in 2025. the appeal of the First Eastern District of Texas jury verdict against Samsung and the companion IPR appeals for the patents that produce that jury verdict will likely be held at the same time at the end of 2025. After the ruling by the Federal Circuit, the parties have the option to appeal to the Supreme Court, but it is unlikely that any of these district court rulings or IPR decisions will be picked up by the Supreme Court. So the Federal Circuit is where the road ends for all practical purposes in these cases. In the event NETLIS is able to reverse any of the negative IPR findings involving patents associated with any of these jury awards and the jury verdicts are upheld on appeal, it would result in a forced collection action likely in the nine figures. Such a collection would be immediate upon a rejection of a certiorari to the Supreme Court and include damages plus accrued interest as specified in the jury verdict. Leading Netlis' effort in these appeals are appellate specialists Molo Lamkin and Stern Kessler. Both firms are located in Washington, D.C., and are among the leading appellate practitioners in the country. Now I'll turn the call over to Gail for the financial review.

Okay, thanks Chuck. Third quarter revenue was $40.2 million compared to $16.7 million in the year-ago period. This was a 140% increase from the third quarter of 2023 and a 9.2% increase on a sequential quarter basis. While we do not formally guide, we currently expect fourth quarter revenue to be in line with the third quarter of 2024. Operating expenses were $10.8 million during the third quarter, compared to $18.3 million in the prior year's quarter, which is a 41% decrease year over year, as well as a 32% decrease consecutively. operating expenses for the nine months ended September 2024 reflected a very active IP enforcement and trial calendar. Legal expenses have been over 60% of operating expenses with the three jury trials we needed to support during this year. We currently expect legal costs to decrease next year with no trials currently on the calendar for 2025. We ended the third quarter with cash and cash equivalents and restricted cash of approximately $30 million compared to approximately $36.4 million at the end of Q2 with minimal debt. We recently strengthened our cash position, raising $15 million through a registered direct offering with the opportunity to raise an additional $33 million from Series A and B warrants connected to the offering. In addition, with a $10 million working capital line of credit with Silicon Valley Bank, we continue to maintain significant financial flexibility and liquidity going forward. And as always, we manage the operational cash cycle very carefully, with days in inventory this quarter improved by 43 days over last year's Q3, and days in sales improved by 22 days year over year. Operator, we are now ready for questions.

We will now begin the question and answer session. The first question comes from Suji De Silva with Roth Capital. Please go ahead.

Hi, Chuck. Hi, Gail. So maybe first question. Chuck, can you talk about the memory demand environment, what you're seeing in terms of 4Q and its 25, the resale environment? Does it continue to remain robust or any signs of pause, any color on what you're seeing trend-wise would be helpful?

Hi, Suji. I think the memory market has been, for the PC and the consumer side, it has been soft. And it is expected to remain relatively soft over the coming quarters. The servers remain fairly strong, AI servers. That side of it is really what is growing and carrying the overall memory market. So the products that we supply tend to be high capacity, high end, that cover the server market. So I think we expect things to continue to grow. the top line will continue to be good for the coming quarters because of the server market.

And then in terms of products, you talked about some products here more than maybe in the past about the CXL NVDIMM, the RDIMS, and the Lightning product. Maybe you could talk about which of those you expect to ramp first out of those, and what the near-term share opportunity is for you there?

For the ultra-low latency or overclock products, that type of memory has been used in gaming for many years. We're doing different tuning and tweaks to that product line And approaching the enterprise market, for example, in high-frequency trading and in AI inference, those require very high-speed memory. So we expect that market to ramp in the coming years, starting next year. at a couple hundred million dollars to start in terms of the market size. So we think that segment of the market will see some good growth for us in that market. For CXLMV, those are design work that's going to happen starting end of this year and into first half of next year, it's going into much larger customers, hyperscalers and some of the major OEMs, where they can no longer use batteries, lithium ion batteries, for power outages. So you have very fast DRAM data, real-time data, If power goes out, that needs to be backed up. And that used to be done through the Intel Optane. And that technology is going away next year. So we're coming in to basically cover that market with the technology we created a long time ago. But that NVDEM will not run in the memory channel. It will run in the CXL channel. But that's a long-term growth market.

Okay. And Chuck, do any of these jobs involve AI training, or is it more traditional computing, high-performance computing?

The ultra-low latency is mostly for AI inference, not training. It's for inference. So all the HBM products, which is in high demand, where Hynix is the leader, where Samsung's running into issues because of their quality problems on HBM, that is for AI training. So that is done at the GPU level. And then at the main memory level, where you require higher speed over clock memory, that is for AI. AI inference.

Okay. And then switching over to litigation, can you contrast the second Samsung case to the Micron case in terms of what's different there that you might want to point out versus a lot of things being similar?

Yes. As I mentioned, two of the patents in this, there are three patents in the Samsung case. Two of the patents are exactly the same as the Micron trial where we were awarded $445 million. That was not in that Micron case, Micron trial in April. that is in the Samsung trial that's upcoming in the next few weeks. And that 608 is a LRDEM patent, but it covers the LRDEM technology in a very different way. Also, that 608 has completed IPR, and the final written decision on the validity of the 608 patent will be rendered by the PTO sometime in the first week of December. So the Samsung trial in November will be similar in scope with the Micron trial that we held in April, but we'll have an additional different patent, a third patent.

And then lastly, Chuck, can you just give us the dates again for the Samsung second trial in November, just the key dates to watch for?

So the trial is set to start on November 12th in Marshall, Texas. So that's a few weeks away. The trial will likely go a week or a week and a couple of days upon which there will be a, usually a jury verdict at the last day of the trial. And then again, that 608 patent, which is a part of that trial, A final written decision from the USPTO is due, I believe, December 3rd, or December 5th is one of those days. Okay.

Thanks, Chuck.

Thanks, Suzy.

Thanks, Suzy.

The conference is now concluded. Thank you for attending today's presentation. You may now disconnect.