Hello, everyone, and good morning. Thank you for joining us for the first day of Needham's 28th Annual Growth Conference. My name is Neil Young, and I am a semiconductor analyst at Needham. It is my pleasure to host this fireside chat with Everspin Technologies. Everspin Technologies is the worldwide leader in designing, manufacturing, and commercially shipping discrete and embedded MRAM and spin-transfer torque MRAM into markets and applications where data persistence and integrity, low latency, and security are paramount. Joining me from the company today are?

Sanjeev Agrawal, I'm the CEO for Everspin.

Bill Cooper, I'm the CFO for Everspin.

Thank you for joining us. So to kick things off, can you give us a brief overview of Everspin's story, sort of where the company started and where it stands today, and any key milestones in that journey that's shaped Everspin into the leader in MRAM technology it is today?

Yeah, sure. So Everspin Technologies is a spinoff from Motorola Freescale. in 2008, and then we went public in 2016. We are a specialty memory manufacturer based on magnetic materials. And what's great about Everspin is we've actually successfully commercialized two different technologies, toggle MRAM and spin transfer talk MRAM, over this period of two decades or so. In terms of some of the milestones, I think what I would highlight is we formed a partnership with Global Foundries in 2014 that was our way of scaling densities from less than 32 megabit all the way up to one gigabit using STTM RAM. We successfully commercialized a DDR-like interface with a one gigabit density, and that part is actually shipping into IBM today since 2018. And then more recently, over the last two years, we also commercialized our X-Pi part with densities going from 4 megabit to 256 megabit, also using our STT-MRAM technology. And those are the parts that are actually especially getting traction in the low-Earth orbital satellite markets, the FPGA markets, and the aerospace and defense markets. And then over the last couple of years, we've also brought on TSMC as a foundry using their MRAM technology and our designs. So that brings me to the other point I would like to make, which is we have our own design in-house. So all the products that we are shipping today were designed internally at Everspin in the U.S. We have an 8-inch factory in the U.S. for manufacturing toggle and ramp. And we obviously manufacture our STT at Global Foundries. So what we focus on is innovation and then finding a path to commercialize that innovation and finding the customers that actually want to use that technology and products.

Great. Yeah, so moving in a little bit of the product and portfolio and technology. So Everspin is known for two main MRAM technologies, the classic toggle MRAM and the newer spin transfer torque or the STT MRAM, which you mentioned. Could you please start by explaining your toggle MRAM portfolio? You know, exactly what is toggle MRAM for those less familiar in the audience and what are its typical use cases and end market applications?

So when you think of MRAM, you basically have two magnetic layers. I say one is a fixed layer and one is a free layer. And these two magnetic layers are separated by an insulating layer. So when the fixed layer and the free layer are aligned in the same direction, you basically have a low resistance state or a zero state. And when they're pointing in opposite directions, you have a high resistance state or a one state. So toggle MRAM actually switches the free layer using a magnetic field that is generated internally in the CMOS using a copper lines that we actually line up with a magnetic material to generate that magnetic field. Spin transfer torque actually switches the free layer by passing a current through the insulating layer. So what that does is with toggle MRAM, for us to generate a certain amount of field, you need a certain width of the copper lines and then you need a certain volume. So that actually limits the scaling of the toggle MRAM. The spin transfer torque, because you can switch by passing current through the tunnel barrier, you can actually scale. So toggle MRAM, like I was saying earlier, it goes up to 32 megabit. And then STT MRAM, we have parts out there that go all the way up to 1 gigabit. In terms of applications, that was something you asked, right?

Yeah, typical use cases for toggle MRAM in any end-market applications.

So if you're looking at MRAM, whether it's Toggle or STT, you're basically looking for an application where data logging speeds are important. So I'll give you an example where, for example, on a manufacturing floor, you have a robot that's actually doing all the manufacturing. It is constantly communicating with the central computer and letting the central computer know what it's doing. So in case there is a power loss, for example, there can be an instant on or instant off type process when you're using our technology. So there is no scrap on the floor. If you're using standard technology like Nordflash, it takes so long to write, almost a couple of minutes compared to a couple of nanoseconds, all the work in progress is actually scrapped. So that's a good example. That's where you would be using it. And several of our industrial automation guys, Siemens, Schneiders, IFM, Mitsubishi, they all use our memory technology preferably over standard memory technology for that reason. Another example is casino gaming, right? With casino gaming, you need to record the events that are happening on the slot machine at least at three different locations. And if it takes forever minutes to write to that memory, then obviously there are only so many operations that you can do in a minute. Replace that with MRAM, and you can actually now certainly double or triple the number of actions you can do per minute. And also the data is stored more reliably. And our chips actually have a tamper detect. So in case somebody tries to turn off the power or bring a magnetic field or some other tampering on the machine, it actually detects and shuts the machine down completely. Like I was saying earlier, good traction with the low-earth orbital satellite markets. And the idea over there is if you are a satellite that is actually communicating with the ground, you have only so much time in which you can update a configuration to that satellite. And if you're using a Nord flash, because it takes so long to write, you're not able to reliably write to that satellite in a particular amount of time that you're given, or the short amount of time that you're given. So that's what MRAM is making inroads. Also, because it's based on magnetic charges, sorry, magnetic spins, and not electric charges, it's radiation immune. So you don't get disturbed when you're actually deployed in the space. And that's the reason why it's being used by Aerospace and Defense for several applications, like low-earth orbital satellites or drones and those kinds of applications.

OK, great. So I think you covered STTM RAM a little bit there. But I just want to ask, how are the different product portfolios manufactured? And what advantages does that bring to Everspun?

GABRIEL SANCHEZ- That's a good one. So when we spun out of Freescale, we actually had our 8-inch manufacturing line in Chandler, Arizona. Actually, that line that we have in Chandler never moved. And the ownership moved from Motorola to Freescale to NXP. But we still own the tools, and we still run the operations independent of the owners. We had a landlord over there. It's been that way now since almost 2000, so about 25 years. And that's where we build all our toggle MRAM. And that is the reason why, actually not the reason why, but we are the only US supplier of MRAM. And that's the reason why the US government is actually specifically interested in partnering with Everspin on many of their programs that are related to aerospace and defense, military applications, contractors like Honeywell, QuickLogic, and FrontRate, for example. When we wanted to go from toggle MRAM to STT MRAM, we developed our first products on the line that we have in Chandler. But it became very clear that if you actually wanted to have a cost-effective solution, you needed to move to 12-inch. And that's when we partnered with Global Foundries. So Global Foundry does our STT manufacturing. We licensed our technology to them so that they can manufacture for us. And also, they got the rights to actually offer embedded MRAM products, which they do on 22 nanometer MRAM, on 22 nanometer CMOS. They basically pay us royalty every time a wafer is shipped for that application. So we have a good relationship with GF. It's ongoing since 2014. And we have product shipping over there on 28 nanometer for Everspint and 22 nanometer for Global Foundries.

OK. So going back to those end markets that you were talking about, I noticed that I wanted you to touch on STT MRAM and how it enables higher densities and is being used in data center memory applications. If you could just talk a little bit about that opportunity.

So like I was saying earlier, it's easier to scale STT compared to Toggle. And the idea is the size of the magnetic cell keeps shrinking as you go to advanced nodes. But because the size of the MRAM is also shrinking, the current required to switch it also decreases. So it actually lends itself favorably to advanced nodes. And that's the reason why all the foundries are looking at 22 and 16 and 12 and 7 nanometer CMOS, because the MRAM cell is shrinking and they can actually use less current to switch it. And that's the reason why those embedded solutions are enabled. So for Everspins STT, we actually built a one gigabit product. If we had done that same one gigabit on toggle, it would have probably been six times the size that our one gigabit is today. That's the scaling question over there. And we've been shipping that part to IBM, and they're using it as a cache memory or a buffer memory. So they have the working memory, then they have Everspins MRAM, and then they have the storage memory, which is NAND flash. So absent our memory, they would have to use batteries to make sure that the data is not lost in flight. But now because they have our MRAM in place, they have removed all the batteries. So actually it turns out that the BOM using our MRAM is actually less than if they were to use storage capacitors or batteries. And the idea is the working memory is downloaded to our STD MRAM, and then over time, they can download it to the NAND flash, which takes forever to write just like NOR flash. So that's the value that we bring to that application over there.

Great. I wanted to shift to IP licensing and your royalty business. So, you know, beyond product sales, you guys also generate revenue from licensing royalties and patent IP. Could you elaborate on this side of the business? You know, sort of where have licensing and royalty revenues traditionally come from? Where do you see them going?

Okay. Yeah, thanks, Neil. I'll take that one. So for MRAM, right, we've built up a very strong IP portfolio over the past 20 years. We've spent over, you know, $100 million developing this IP. We have over 650 patents in that space. And as Sanjeev mentioned, we do license the MRAM technology. We've licensed to Global Foundries, Honeywell, Bosch is a good example. Also, some of our magnetics portfolio with Bosch and with Alps.

So, for the house. And should investors view the licensing business as a growth opportunity or more of an opportunistic, lumpy income stream?

Yes, that opportunity, you know, the licensing, we like to target that space to kind of be in the 10% to 15% of revenue. It does tend to be lumpy because they tend to be specific projects, you know, with a discrete beginning and end, and then some royalty revenue over time as well. So, you know, it's one that is high margin, and we like it for that reason. And, of course, we do look for opportunities to license.

Okay. I want to go back to the low-Earth orbit satellite and space applications you were talking about before. I know you sort of dove in a little bit of why you guys are such a strong fit for LEO satellites. Maybe talk about how you're approaching this market, and if you can, share any design wins or customer engagements in this satellite arena.

Yeah, sure. So in terms of design wins, there are a couple that we can talk about openly. We have made press releases with respect to that. One is Blue Origin, and then the other one is Astro Digital. There we have design wins, and the idea with the Astro Digital is they have flight control systems. And they basically are looking for a reliable, fast boot memory. And that's where Everspins MRAM comes in place. So if you're using NorFlash, the reads are very fast, but the writes are actually pretty slow. And if you're trying to update the configuration system, that's where Everspins MRAM plays a strong role and is a preferred solution compared to NorFlash. In terms of how we are approaching it is, we basically have clear ideas on who is actually planning to launch satellites. So we are actually getting in there, talking about the advantages that we bring to the table, and trying to get design wins from that. Also, we have reps all around the US and Europe that actually go and talk about our technology. So combined with the DISTs, the reps, and our marketing team, including myself, when we visit customers, we talk about the benefits and see if we can actually learn our design win.

Great, and then if you could talk about how large is the content opportunity per satellite in your view. I know, for example, I've seen some estimates suggest that each LEO satellite could use on the order of anywhere from a 4 to 8 MRAM unit. So just any quantitative measures you can put around the opportunity I think would be helpful.

Yeah, so I think there was a report from one of the banks that quoted about 70,000 satellites over the next five years. And that actually includes China as well. So with the relations changing, I don't know how many of that necessarily MRAM can serve. But I think a number on four to eight units per satellite is a reasonably good number. So let's round it up to 10 just for convenience. So 70,000 satellites times 10 units over the next five years is a good enough number with an ASP of anywhere from $50 to $100. So it actually is good revenue to go after that market.

Shifting to competitive landscape, I want to dive in. Who does Everspin typically compete against across your different product lines? In other words, what are the alternative technologies or even rival companies that you might come up against in your target markets?

If I might, I'll step back a little bit and first differentiate between embedded MRAM and discrete MRAM. So all the foundries today do offer embedded MRAM. So Global Foundries, Samsung, TSMC, UMC, they all have an offering of embedded MRAM. Everspin does not compete with them. We don't offer embedded MRAM. We only offer standalone or discrete MRAM. With Global Foundries, we license them so that they could ship embedded MRAM, and we collect royalty on it. But we only ship standalone or discrete MRAM. Over there, if you're looking at traditional memories, I've been talking about NorFlash forever. So clearly, we compete with them on, if you're looking at memories that are used for storing configuration, or when you boot up your phone or whatever it is, that's any device that basically typically boots out of a NorFlash. So we compete with NorFlash over there. If you're looking for data logging that I was talking about earlier with the robots and the computer, that's where we compete with non-volatile SRAM, or we compete with battery-backed SRAM, or we compete with ferroelectric RAM or FRAM. So those are our competitors in the data logging market. And that market is on the order of $700 million or so, the TAM. And with the Flash, the TAM is on the order of $3 to $4 billion. So these are pretty different TAMs. And the NordFlash is obviously a larger TAM that we are targeting. Finally, it's the data center, the one that I talked about where we ship into IBM. There, the TAM is much larger, but we only made one part so far. It's a DDR4-like one gigabit part, and we are serving mainly IBM. And our next product over there in that line would have to be something that can be more universally adopted than IBM. The challenge with our DDR4 part, it's not true DDR4. So you can't remove a DRAM and put an MRAM in there, because all the parameters don't match, unlike NorFlash. And SRAM, we can replace those parts directly with our MRAM product. So the next product that we bring out in that storage line would be socket compatible, so that it's easier to proliferate in the market and get more designments.

OK, and then so you talked about how you guys are one of the only few, or you're few, if not one of the only, doing the embedded MRAM. Have you seen or heard anything about these larger companies, maybe developing technology that would compete directly with yours? and sort of, if so, what differentiates you guys versus them?

So I think you mean standalone, right? Yes, yes. Sorry, correct. So we are not seeing any of the larger companies actually stepping into the standalone business. But, for example, there are two smaller companies that did copy our business model or that used the same business model. One is Avalanche. They actually licensed their technology to UMC, and they tried to enter the standalone market a few years ago. Unfortunately, I think, well, they're no longer available, at least in Japan and Europe. The other company is NetSol, and they're actually using Samsung as a foundry. They do have a 16-megabit part that can actually be scaled on to 1-megabit, and it offers serial and parallel interfaces as well. But they don't have an entire roadmap that could actually support a customer. So I think perhaps that is one of the challenges that they're facing as they try to get customers or compete with Averspin.

I want to touch on, I know you guys were talking earlier, how you are a uniquely U.S.-based manufacturer of MRAM. So that obviously ties into national security interests, especially with everything going on. So in September 2025, the U.S. Department of Navy put out a request for information seeking industry input on establishing a domestic 300-millimeter back-of-line, back-end-of-line MRAM production capability for national security applications. Obviously, not going to ask any confidential details, but generally, how do you view this initiative for Everspin and your guys' positioning?

Yeah, thanks, Neil. Yes, we did receive the RFI from the government for a 300-millimeter onshore MRAM line. We did submit a response to that RFI on a timely basis back in October after that came out. And given, you know, MRAM's IP portfolio and some of the partners that we submitted with, we feel good about, you know, our chances for getting a positive response from the government.

So would this be something, a standalone effort? Would it be a partnership? Maybe if you could talk about it from that angle.

Yeah, it would definitely be a partnership. We expect that partnership effort. I think, you know, from that perspective, you know, we obviously have a close partnership with Global Foundries. um as well so you know there's there's logic there and of course they have uh also a fab facility here onshore as well and so we would look to partner with those folks to potentially bring that 300 millimeter line onshore as well as other partners potentially and any thoughts on

the potential timeline or scale of this effort to the extent that you can share of course

So, you know, the key thing is, of course, we have the channel line, which has been, you know, here in the U.S. for 20-plus years on 200-millimeter natural progression, right, to move to the current generation technology, 300-millimeter. We submitted the RFI. I will say, you know, the folks on the U.S. government side were still working even across the shutdown. and so we would expect to see some next steps come out hopefully sometime here in the next several months and that next step would be an RFP that we would then, again, with our partners respond to and then we would expect to hopefully by the end of the year have the response to the RFP.

Okay, great. That's really helpful. Last question for me. Finally, is there anything that we haven't covered that you think investors should know about Everspin? I just want to leave the floor to you guys here.

I would say that I think what I've learned in my years at Everspin, I actually joined it before it was spun out of Freescale in 2006. The attention to detail and quality that the team pays to its products speaks volumes. I think we have probably less than a few handful of parts that have been returned to Everspin since its inception, and we have shipped about 200 million plus units. and the depth of the technology team that we have at Everspin. I think we are very well positioned not only to innovate, but then also the experience that we have to actually commercialize it and productize it. So I think we're in a good position. And I think with this flash market and this edge AI market in the storage industry, I think we have a lot to contribute and you will see us over there in the next few years. So thank you for your attention.

Thank you, Aaron. Thank you.