Hello everyone, and welcome to the Everspin Fireside Chats. My name is Robert Blum, Managing Partner at Lytham Partners, and today I'll be moderating a Q&A discussion with Sanjeev Aggarwal, the company's Chief Executive Officer. As a reminder, Everspin trades on the NASDAQ under the ticker symbol MRAM. All right, let's get started. Sanjeev, welcome. Thank you, Robert. Pleasure to be here. All right. For those new to the company, provide an overview on sort of your mission and really how that sort of evolved to address this growing demand for advanced memory solutions. Yeah, Everspin is a semiconductor manufacturing company. We manufacture magnetic-based memories. The claim to fame over here is that we can actually tune our technology to look like different kinds of memories. It can be a data logging memory, it can be a configuration memory, or it can be a data center memory. We have basically shipped about 200 million plus units since our inception in 2008 to about 2,000 + customers worldwide. Most of our applications are where there is mission-critical data involved, where security is a concern, where data logging is important, and those are the applications that we have actually thrived on. What was that moment, you know, the milestone of sorts in this journey that you've been on here that have really shaped the position as this leader now in MRAM technology? Yeah, so when we were developing the technology, we realized that there is no other technology that can actually read and write as fast as MRAM. Basically, just to give you an example, nanoseconds read and write versus a microsecond write with a NOR flash memory. We realized that that is a very critical parameter for many of the applications such as programmable logic computers in the automation industry. The idea is that the robots are actually manufacturing on the floor or manufacturing different things, and when they need to write back to the PLC, constantly informing the PLC as to what they're doing. If there is a power loss in this situation and you are using a traditional memory, like a NOR flash, all the WIP on the floor would actually be scrapped. With our memory, you can actually record within nanoseconds what the robot was doing. In that sense, when the power is lost, we don't actually have to scrap the WIP on the floor. When the power comes on, the robots can start at that very moment, at that very location, or that very action that they were doing. The bottom line is very fast data logging. The other attribute is automotive temperatures, - 40 to 125 degrees C. Very robust, very reliable, and fast data logging. When we realized these three things, we basically understood that it has a wide range of applications. It has an application wherever you're thinking of data logging. For example, battery management systems in EVs, PLCs, like I just said, in automation, casino gaming, or in any medical equipment, for example. That's where we actually took off and started off with our first product targeting the data logging industry. You touched on a couple there, but maybe walk specifically through the product offerings, Everspin's product offerings, and then sort of how they integrate into some of the customer systems that you just mentioned here. Yeah, sure. If you look at Everspin's product portfolio, we actually categorize them into three different categories. The first one is PERSYST, which basically stands for performance memories for systems. The second one is UNISYST, which is unified data and code for systems. The last one is AgILYST, which is basically agile memories. That's basically looking into forward-looking applications like AI and as such. We started off talking about PERSYST, which is basically where you're looking for very robust memories at extreme temperatures, - 40 to 125, for example, or even for aerospace applications. The idea is you can record very quickly as to what's going on, like the battery management system that I was talking about. UNISYST is where you're actually trying to combine data and code. The idea over there is if you think about FPGAs, the configuration of an FPGA is actually stored in a memory which is off the system. On power-up, the configuration is downloaded from the memory into the elements in the FPGA, and then the FPGA executes, or the logic elements execute the action. Now, this NOR flash memory takes a microsecond to write. If you're trying to download a 64-Mb memory, for example, it's going to take several minutes. If you were to replace that with an MRAM, it could only take nanoseconds. For example, if you're familiar with Teslas, what happened in early, I believe in 2024, when they were going through a configuration, all these Teslas were actually parked on the side of the road waiting for this configuration to go through. Many times it failed as well, then it had to be rebooted. If you were using an MRAM instead of that NOR flash memory, it would have been done by the time you hit the power button to actually start the car. If it failed, all you had to do was start it up again, and by that time the configuration would have updated. That's the power of MRAM: very fast data write and read, very fast data logging, and robust temperatures. All right, very good. I want to sort of jump into the business model here a little bit. Describe what the business model is. You've got revenue from product sales, licensing, partnerships. Provide a little bit of an overview on each of those if you could. Yeah, sure. Our business model basically has three different revenue streams. One is product sales. The second one is licensing and royalty. The third one is, and I basically put it out as separate because it's a U.S. government contract. We are a memory that is mission-critical for various U.S. government applications. There is a revenue stream from there where somebody can come and license our technology, so you get licensing revenue from there. There is a foundry model. Once the technology is being qualified, then the same contractor or the prime would actually use us as a foundry to build those parts. When the part ships out the door, you get a royalty based on the part that uses Everspin's technology. That's one. The second one is basically product sales. We have three different generations of memory over here. One is the field switch or Toggle MRAM. Then we have the perpendicular MRAM or spin transfer torque MRAM that we are actually shipping parts today into the data center, as well as into the configuration memory solutions. Product sales, licensing revenue, and foundry revenue, those are the three categories that we recognize revenue from. Let's talk a little bit more about the customer base, right? How do you balance large companies, smaller customers, specialized sort of situations, more broad situations? Talk about the customer base in general if you could. Sure. We actually have a very diversified customer base. We have 2,000 + customers worldwide. These include customers like Siemens, Schneider, IBM, Juniper Networks, the big Fortune 500 companies as well. The idea is that we serve most of these customers through our distributors worldwide. The idea would be, for example, let's say we have a large customer like IBM. In the early stages, you would actually work with the customer to help them understand our part, work out any issues that are there. Then basically we hand over, and this is the customer's choice, we hand over the sales or the supply chain through a distributor. 90% of our business actually is through distributors worldwide. We have distributors in Asia-Pacific, in Europe, as well as in the Americas. When you sort of put them into different verticals here, right, you talked about some of the medical applications, some of the automotive applications. What are the areas or the focus that you maybe see the greatest growth opportunity from going forward here? Yeah, so to date, most of our revenue comes from our PERSYST family, which is basically industrial automation, the PLCs that I was talking about with the Schneiders and the Siemens of the world, casino gaming. One thing that I forgot to talk about was, you know, today they're using this traditional memory NOR flash in casinos, right? Every action on the casino machine has to be recorded three times or in three different places. If you're using a NOR flash, it takes forever to record that action. That limits the number of times you can actually play on the casino machine. If you replace that with MRAM, you get a reliable memory, you get a faster memory, and now you can get more actions on that same casino machine as opposed to a NOR flash memory. That's another steady revenue or high segment that we have revenue from in the PERSYST family. The third one I would highlight is the aerospace and defense industry. We ship a lot of parts into automotives. For example, I was talking about battery management systems. I was talking about Boeing and Airbus, for example, and then various applications through our primes like Honeywell and Frontgrade Technologies, where we've actually gone into space, you know, on our way to Jupiter on the NASA mission with Lucy. We are also actively collecting data from Mars today with our Toggle MRAM over there. That is our existing revenue. Going forward, where do we see growth? Where we see growth is our NOR flash actually stopped scaling at 40 nm CMOS. Also, the densities basically that you can buy of NOR flash today are basically around 256 Mb as the highest density you can buy. We are now targeting our STT-MRAM product to actually scale, sorry, to enable the roadmap for higher densities but replacing flash with our MRAM. We are actually shipping today anywhere from 64 Mb- 128 Mb MRAM that looks like NOR. In the future, we are actually designing a product that takes that same density from 256 Mb - 2 Gb. We're going to be able to address this market. This has a huge total addressable market of about $4.3 billion. There's no other company that is actually building MRAM today that can address this space. We're very excited and we're actually bringing out our first part with the higher density in the 2026 timeframe. Good. Let's talk about the sort of the competitive environment here. Who are some of your primary competitors? How do you position yourselves against them? Yeah, so if you look at the competitive landscape, there is only one company called NetSol, like Network Solutions. They're based out of South Korea, and they use Samsung as a foundry, just like we use GlobalFoundries as our foundry. They only have one offering. They have a 16 Mb part that they actually sell down to 4 Mb as well, but there's only one native die, 16 Mb, and that's it. The next part that they're actually designing is all the way up to 1 Gb, which is about three years out. If you compare that to us, we have an offering all the way from 256 Kb - 256 Mb, and they're bringing out a 1 Gb part in the next year or so. Right. A much broader range of parts that you can offer in industry standard packaged parts, BGA, TSOP, etc. I think NetSol is there, but I think they offer only one small density, and therefore we don't see them too much in the space with our customers. Because of the offering that we have, the customers actually like to see a roadmap, and they would rather design us in compared to our competitors. The other one is Avalanche Semiconductor. They're based out of California, and they are actually focused now only on the aerospace and defense industry. They talk about 1 Gb STT-MRAM that they're planning to bring to product over the next year or so, would target it towards U.S. government applications. When you sort of take a step back here and you look at some of the key market trends, I think you've hit on a number of them here, but what are the biggest sort of trends that are shaping the memory industry going forward here, and how are you sort of ideally positioned to capitalize off these? Yeah, good question, right? At least two or three different areas. One we just talked about, right? NOR flash. There isn't another solution because NOR flash stopped scaling at 40 nm CMOS. We are positioning ourselves to look at that market. More recently, we've also realized that this low Earth orbit market, or the LEO market, is actually growing pretty fast. One of the reports actually said that you're going to have 70,000 satellites over the next five years or so. That's greenfield for Everspin. Because MRAM is actually radiation tolerant, it can actually serve the LEO markets very well. Examples of that are you might have seen our press release with Blue Origin, where we are actually designing over there, and then also with Astro Digital. Most of these guys are looking for a reliable memory that is radiation tolerant, and that's where Everspin's MRAM comes into play. Like I said earlier, there is no other player that is actually offering MRAM for these markets today. I think it is another good opportunity for us, and we are making sure that we collect all the data to show radiation tolerance, and we are working with the right people or the right vendors to make sure that we get designed in as these LEO markets evolve over the next few years. If I didn't mention AI, I would be remiss, right? Today, if you look at the data centers, they are heavily focused on very high density for compute applications. In fact, slowly now they're actually turning towards inference. It was all about high density, it was all about compute, and now they're actually looking at inferencing on the edge as well and in the data centers. That's where MRAM comes in. If you're looking at, if you're using SRAM today, for example, for inferencing, what are the challenges of that? The challenge is that it's actually very leaky, and you have to keep refreshing it, and then you also have to have a backup in case you lose power. If you use MRAM over there, you can actually, you're as fast as SRAM, and you can actually compute or execute in place, and then it's non-volatile, so you don't have this large leakage that is associated with SRAM. You can then start replacing all this SRAM or NOR flash that you have on the edge with MRAM for faster AI because you don't have to go to the cloud every time you need to make a decision. You're losing a lot of power going back and forth. You can make the decision on the edge. You can execute on the edge. That's where I think MRAM has a role to play in the coming future. Big topic of conversation here of late has been sort of supply chains, tariff, manufacturing locations. Talk a little bit about the company's operations, any restraints, any issues that you've had of late. In Q2, what we reported, we did not report any impact, or even in Q1, we did not report any impact on tariffs. We haven't seen any yet. It's a changing landscape. Every day there are new rules, so we have to keep in mind of that. As far as the supply chain is concerned, it's pretty diversified. For example, the MRAM that we manufacture in Chandler, Arizona, we actually buy the front-end wafers either from TSMC in Washington State, or we actually get them from Taiwan from Fab 6. We finish the MRAM over here, and then we do our packaging and final test in Taiwan, and from there, we distribute to our customers. For spin transfer torque MRAM, we actually use GlobalFoundries as a single source. By that, I mean as a single turn. They do the front end as well as the back end. The front end is done in Germany, the back end is done in, and by the back end, I mean MRAM that's done in Singapore, and then the parts are shipped again to Taiwan for final assembly and test, and from there they actually ship to our customers. In that sense, you can't point a finger and say, okay, 100% of our products are going to get impacted by this change in tariff policy. I think it's pretty diversified, and we keep looking at opportunities to diversify our supply chain. All right, very good. You just reported, I believe it was Second Quarter Financial Results here over the last, you know, earlier this month, I guess. Provide some highlights of the quarter, customer wins, milestones, etc. Yeah, sure. We actually reported a solid quarter one more time. We actually came in above our midpoint of guidance and above the consensus of the analysts. We came in at $13.2 million versus $13 million, and once again, our non-GAAP income earnings per share was $0.03. Again, another profitable quarter, which is obviously great for Everspin. The highlights that I would mention from that quarter are about three things. One is we are actually starting to see a depletion of the inventory overbuild at our customers. For the longest time, we actually didn't see many orders coming in because of the inventory overbuild past the pandemic. We started to see the bottom. We are actually starting to see orders grow. That was basically in three different applications. One, for example, in the automation, which is what I talked about, is one of our primary business. The other one was in a data center. The third one is in the FPGA applications. Here I would like to point out that the second highlight was the new parts that we have brought to market over the last two years, the XSPI family. We're actually starting to see significant design wins in this family, and we expect to start generating revenue in the second half of 2025. Last but not the least, in preparing for our growth, we actually have expanded our leadership team. We actually separated out our position of VP of Business Development and Sales. We now have a dedicated VP of Business Development, David Schrenk, and we brought on Sean Dougherty as a VP of Sales. He just joined us from Intel, and he was actually the Chief Revenue Officer for the Altera business over there. Very excited to have him on board and looking forward to our growth over the next few years. All right. This is an Investor Conference, obviously. Talk about what you would see as the key highlights investors should be looking at and why you think this is an attractive opportunity here for shareholders. A few things. One is a solid balance sheet. We are debt-free, and we have about $45 million of cash in the bank. We continue to increase our profit margins, and we continue to run a profitable business. That's number one. Second is we are actually targeting a pretty large TAM of about $4.3 billion by 2029. We actually have products that address the data center market, products that address the data logging market, and also the configuration memory or the NOR flash. I think we are well positioned to address these markets and are well along the way. The other thing is we are the singular domestic supplier of MRAM for mission-critical applications. We are designed into several U.S. government contracts or projects where we actually provide radiation-hard MRAM. The front end is actually hardened by our defense primes. We only do the MRAM portion, which is the same, whether it's commercial or the U.S. government. Because MRAM is naturally radiation immune, it actually helps with the U.S. government applications once the front end has been hardened by our primes. The other thing is we have a pretty diversified customer base, like we were talking about, right, Robert, 2,000 plus customers worldwide. It's not like if one customer had a downturn, it's going to impact our revenue or profitability in a big way. Pretty diverse business, pretty sustainable, profitable business that's growing over the next few years with a large TAM. A strong balance sheet, as you mentioned there. Final question here, as you sort of look ahead, what is your vision for this company over the next decade? Maybe a decade's too long here, but you're going to go as far out as you feel is appropriate. What's sort of that legacy that you hope to leave as CEO here? MRAM is the future of memory. I think that's what I would like to be my legacy once I'm done from this role. The way to get there, or the way to make the mark over there is the journey that we have started. I think the MRAM can be tuned to look like any memory for any application, whether it's a data center, whether it's AI, whether it's NOR flash, whether it's data logging. It's such a versatile memory, and it's so reliable. I think it is just a pleasure to see it get adopted more and more for these different applications. Last but not the least is the environment and the climate, right? There is so much power being consumed, and people are planning for so much power consumption with these AI applications. I think MRAM is a beautiful solution. It can actually reduce the power consumption, increase the speed, and make life easier for the common person as well. Like I said, MRAM is the future of memory, and I think we'll get there. All right, very good. Sanjeev, thank you so much for your time today. I want to thank everybody, obviously, for continuing to watch throughout the day here. If there are any questions for Sanjeev that I can help direct to him or to the investor relations team there, please let me know. Again, that's Blum, B-L-U-M, at Lytham Partners. We have additional presentations and fireside chats coming up here, so please stick around. Again, Sanjeev, thank you so much for your time. Greatly appreciate it. It was a pleasure, Robert. Thanks.