Good afternoon. We're going to get started. I'm Tim Arcuri. I'm the semiconductor analyst here at UBS, and very pleased to have Impinj, and with Impinj, we have Cary Baker, who is the CFO. So thanks, Cary.
Yeah, thanks for having me.
Great. So let's just start with probably the biggest question that we recently initiated, and the biggest question that we have heard is just sort of the pace. I mean, obviously, I think everyone recognizes that the TAM is huge.
Yeah.
And the challenge is just penetrating that TAM. And can you just talk about sort of what some of the impediments are to penetrating the TAM, maybe in some of the key end markets? I think it's probably easier in some end markets than it is in others.
Yeah. So if you look at any information industry and you go back in time, the industry must first build out the hardware foundation before it can realize its full potential in that. Think about mobile phones as an example. We spent the '80s and '90s building out the hardware foundation so that by the early 2000s, we had a flip phone that was just a phone. You could make calls, and you could struggle with T9 text. Before the industry was sufficiently mature where you could monetize the apps and the data services and the solutions designs and even the AI in the phone foundation, that is not unlike where we are with RAIN. So the RAIN industry, oh, since 2010, has had a unit CAGR of 28%. In 2024, the total volume for the industry was 52.8 billion.
It was up 8 billion from the year prior to that. And that's mostly going to the Western world, and that's mostly on the back of a single vertical, retail apparel. Today, we have now more than one vertical. We have logistics, we have general merchandise, and we're just now announcing item-level food. So it's the situation where Rome's not built in a day, and we're laying that foundation right now. And the unit CAGR that we have delivered is a strong indicator of the potential of this platform once the platform has reached that maturity level.
I mean, I guess it's pretty hard to get these large global organizations to change and to adopt new things, and that's probably the single biggest thing that you kind of work against. So are there some verticals that are more that you would highlight that are sort of out there still that are going to still take a couple of years?
Yeah. I think the verticals that we focus on today, apparel, logistics, general merchandise, and food, those are all beginning to move forward. As I look further into the future, opportunities where I see use cases that exist but the industry is not moving forward yet would be automotive, for example. German-made auto manufacturers have autos with anywhere from 30-50 tags in them. They use it for quality control down the assembly line to make sure a U.S.-bound auto receives a U.S.-spec airbag, or that the color scheme of the seats being put in the car match the color scheme of the dash, and the goal is to identify any challenges before they get too far down the assembly line where they have to scrap the car.
Pharmaceutical, all the regulations necessary to move forward in pharmaceuticals have been approved, but the industry hasn't moved forward because we haven't seen one of the big pharmacies make the decision to move forward. The drug manufacturers would bear the cost of this, but the pharmacies would reap the benefit. Those are two verticals where there are active use cases deployed today, but the industries just haven't moved forward yet. So I see those on the horizon after we tackle some of these bigger ones first.
Got it. How do you think about the share dynamics in the endpoint IC market? Is share tied more to your existing customers or to new programs with Lighthouse customers?
Yeah. So our platform is our competitive advantage, and we put functionality in the ICs that is engaged by functionality in the readers to solve problems that you can't solve with a mix-and-match solution. And if we're successful at that, we will drive share to or drive preference to our endpoint ICs. Largely, the share gains that we've enjoyed so far have been on the back of that platform.
Just for the Lighthouse programs, what creates this long-lasting moat that allows you to retain that share and to gain from there?
It's a solution that you can't solve with a mix-and-match reader and endpoint IC. So it is functionality in the IC that is leveraged by the reading environment that the software and the reader creates. Some examples of that would be in our Gen2X features. So in a normal Gen2 environment, which is the existing radio protocol, the readers or the ICs are constantly responding to the reader, which clutters the environment. With Gen2X features, we've put a capability in the IC that the reader can engage that IC, tell that IC to stop responding because it's already been read so that the reader can focus on reading the rest of the ICs in the market and just kind of go through that process, deselecting the response as it reads. That has a dual benefit.
First, it increases the read speed because the environment's been significantly decluttered, and second, it increases the read range in that environment. That solution, that tag quieting solution, as we call it, is only available with the Impinj M800, which has Gen2X features, and one of our R700 readers or our reader ICs, our E-family reader ICs, or one of our partners that we've licensed the Gen2X features to, like Zebra as an example.
How ubiquitous is that, or is that just in its nascency?
It's in its early days right now. We announced Gen2X at the end of last year, so we're just coming up on a year of Gen2X in the market, but we're seeing real traction from that. One example is in overhead reading. So we have a customer that was a historical Impinj customer, but during the supply constraint environment of 2022 and 2023, we were unable to supply them. And what was kind of a dual-source account where they used us and they used NXP went almost fully to NXP. That customer, however, has always had a vision of an overhead reading solution. They want to move away from the handhelds. They want real-time reading available all the time via an overhead solution. But the read range of standard Gen2 was too small to make the economics pencil out.
It required too many overhead readers to make it and made the program cost prohibitive, but with Gen2X and that tag quieting feature that I just talked about, the read range has increased by almost 40%, and now, all of a sudden, with that increased read range, the math starts to pencil on the overhead reading, so this is an opportunity to both deliver a solution that the customer has long wanted but hasn't been able to get, and also be able to drive endpoint IC volumes in that account back to Impinj.
How does the software factor into all of this?
Yeah. So with every reading environment, you need software to control it. So we are introducing software components to our reading to take the RF out of RFID, to make the provisioning easier by putting machine learning at the edge. And what that does is that drives a solution that can be zero-touch provisioning while also an environment that self-monitors and adjusts constantly. So we can make the reading setup much easier for our end customers.
And so is the competing solutions, I mean, in this fully integrated solution, you have the software, you have the reader, and the endpoint, all three?
That's correct.
As we think about these more mature markets like apparel or footwear, for example, how do you think about your market share versus your primary peer going forward? And how do you estimate self-checkout adoption thus far and how that could evolve over time?
The primary use case in retail apparel today is using a handheld reader to count inventory on the store floor. And that is primarily a mix-and-match solution where any reader and any IC can be paired together and deliver sufficient performance in that case. We perform better, but the environment in a mix-and-match works just fine. What we're seeing in the retail market is retailers are moving towards 100% penetrated. Only a handful, probably less than 10, are 100% penetrated at this point, but they're all striving to get to that point. And the reason they do is once you get to 100% tagged, you can then unlock the additional use cases. You can go beyond handheld inventory counting to self-checkout, to loss prevention, to front-store, back-store management, or smart fitting rooms. In those use cases, you no longer rely on a handheld reader.
You're moving into fixed or autonomous reading where there's no human in the loop, and the accuracy threshold increases significantly. The Impinj platform is uniquely positioned in those cases where fixed or autonomous reading is necessary to provide the solution. So we think over time, as those use cases become more prominent, that will help drive share in the retail apparel space to Impinj. Prior to 100% tagging, the example of overhead reading and some of the performance improvements in M800 in hard-to-tag categories will also help drive preference for Impinj ICs.
And so I guess just structurally, until you get to 100%, you can't start to adopt some of these other innovations until everything's tagged.
Yeah. You can't do self-checkout until everything in the store is tagged. You can't permanently replace the hard tags that trigger the alarms at the doors until everything has been tagged.
Can we talk about channels? So in your engagements with end users, you typically supply them through inlay partners and service bureaus, and ultimately, you're exposed to sort of multi-level supply chain risk as customers or partners look to burn or build inventory. So how do you manage the multiple nodes of the supply chain and the inventory that can actually sit in those nodes?
Yeah, very carefully. We try to at least. We get monthly inventory reports from our channel partners. From our Lighthouse accounts, we get forecasts directly from them to track inventory levels. Then we get anecdotal information from the service bureaus, which are two steps removed from us. We also track a subset of public retailers that we think of as a RAIN indicator set so that we can match macro trends that the retailers are seeing with what the ordering patterns are from our direct customers to try to triangulate that. All that being said, it's still a multi-step distribution model. We'll never be 100% immune to channel inventory builds. What we have been able to prove is that we can identify it early on, and then we're able to correct it very quickly.
How much visibility do you typically have?
Typical lead times are six to seven weeks. We turn 50% in a quarter.
And that doesn't change very much?
Not usually. Typically, when there's periods of time where it changes, it's signaling something else is going on. If you go back to the fourth quarter of 2024, we saw our lead time shrink dramatically as tariffs began to be announced, specifically in China and Canada and Mexico, and what was happening at that point was our partners were adjusting their production footprint to optimize for the tariff, so they were moving out of Mexico, as an example, into Southeast Asia, and we saw the lead time shrink as they were doing that, and we saw the lead times go back to normal levels thereafter.
I think you've talked about this. You've broken up the ramp into three parts. It's like the first part is hard, and then the middle 60% is easy, and then the last 20% is hard. So where do you stand with respect to food in that life cycle?
Yeah, so we're at the early days of food right now and think of that 0 to 20, 20 to 80, 80 to 100 as a proxy for how new programs ramp, so in the case of food, for the most recent food announcement by Walmart, think of what they've got to do to get that up and running. They've got to put in an infrastructure, which is basically printer encoders in their existing weight scales. They have experience with RAIN RFID in the apparel department, but the food team doesn't have experience, so there's tens of thousands of employees across their 4,600-store network that need to be trained up on doing that. All of that happens at the first part of the ramp, and that contributes to the first 0 to 20% of a ramp going at a relatively slow pace.
Once that foundation is built, they can accelerate the program. And that's why we see typically going from 20%-80% at a much faster pace. Capturing the final 20% typically comes down to solving for the hard-to-tag items or the edge cases that you haven't figured out just yet.
With respect to food, you've called item-level food as the single largest RAIN market. As we understand it, there's, I think, eight global food retailers that are exploring RAIN. Have you sized sort of a more immediately addressable market out of these folks who are looking at being the first adopters?
We haven't sized it specifically yet, but what I would say is that the sell-side estimates that are out there are reasonable in the case. Where we're seeing the grocer start is in the portions of the grocery store where they control the supply chain. So think the four outside walls of a grocery store where they're packaging, labeling, sometimes producing the items directly behind the counter or at their company-owned DC. They're doing this because there's no negotiation required. If they wanted to go to the middle of the store, they would have to negotiate with the CPG companies to put the tag on a box of cereal, as an example. In this case, they skipped that step, and they're able to make the move much faster. They'll eventually get to the middle of the store, but they're starting where they think they can get more momentum earlier.
And sort of as you look into next year, I mean, obviously, as you said, you have very, very short lead times. However, you do have a little more visibility on sort of maybe some of these larger programs and how they're ramping. So how much visibility do you have next year, and what's going to be the main put and take in terms of revenue drivers next year?
Yeah. So we see multiple opportunities for growth next year, kind of in no particular order. So in apparel, we see expansion growth. Most retailers have adopted. Most retailers are not 100% deployed, so they're growing to 100%. We'll also see net new opportunities next year like we did this year, even though there's a high brand penetration rate. Within retail, macro is always the factor that people are focused on these days. And coming out of the impact of tariffs, we saw cautious rebuying behavior from retailers despite a pretty solid back-to-school. Fast forward to today, we've seen very strong early holiday prints from Walmart, from American Eagle, from Inditex, from Macy's even. So that tells us that the consumer is much more resilient than maybe they thought earlier in the year. And if the consumer remains resilient, I think that bodes well for 2026 rebuying behavior.
So that's apparel. In general merchandise, Walmart has announced two phases of general merchandise. We're not complete in tagging those two phases, so there'll be additional penetration in those two phases. We think there's an opportunity for a third phase in general merchandise. Categories have been discussed, but nothing's been announced at this point. In logistics, our second-largest logistics provider achieved 100% domestic penetration in the third quarter. So there's an annualized growth rate that factors into just the domestic, and then there's an opportunity for international beyond. There are other logistics pilots going on right now, and depending on how quickly those can convert to full-fledged programs, that could impact growth in 2026. And then finally, on food, there's additional penetration in the Kroger bakery pilot. There's potential for additional categories with that. And the Walmart food program is just getting started right now.
When you talk about food, so some of the high moisture and produce categories, they seem to have maybe packaging and some constraints that would complicate tagging. What's the sort of binding technical and operational constraint to have full food coverage? And where do you expect the real innovation to come to really unlock these last pockets that have been hard to access?
Yeah. So anything that's liquid or metal interferes with the RF propagation. So you can imagine food between its properties and its packaging introduces a lot of challenges. Those challenges can be solved and likely will be solved by the ecosystem in terms of developing tags. So as an example, one of our partners developed a very clever foam spacer that can get applied to proteins. And basically, once it's applied, the sticker's applied, the spacer expands, and the tag's on top of it, so it creates enough airspace for the RF to propagate. That works great on steak and the water content in steak. Doesn't yet work on poultry because poultry has a higher water content, so we still have to solve that. In some of the most challenging use cases, the end customer or the ecosystem will come to us.
In the grocery space, early for the bakery pilot, tagging pies was a challenge. Not the pie itself, but the aluminum pie tin that the pies are held by. And we needed to develop a tag to solve that. Another example was tagging a gallon of milk because the milk interferes with the RF, so you can't put a sticker on the side of that. You had to design an inlay that sits on top of the milk cap. We have an in-house inlay design team, and when necessary, we can deploy them to solve those hard-to-tag items. And then once they've solved that, we hand out that inlay reference design to the ecosystem so they can support the end customer.
Yeah. So I was going to ask about that. Where does the ecosystem, the sort of push-pull, how much of these hard-to-access things, how much technology are you building around the tags? How much of it are you pushing into the ecosystem versus them coming to you and saying, "Hey, we would like to tag this. Can you please help us solve this?
Yeah. So we push a lot in the form of IC by making the IC more sensitive to make the IC smaller so it can fit on a milk cap or it can leverage a smaller antenna to get the targeted range. Then they take the IC from there, and they're primarily the ones that develop the inlays for that. It's only when we get to the hard-to-tag, the very hard-to-tag items, that they'll come back to us for help.
Got it. So apparel remains your largest market, but 2025 looks kind of close to flat due to some tagging partner inventory issues and some limited full adopter penetration. What's the forward path recovery in the apparel market, and how many new retailers and deeper deployments would be required to have a big up year in 2026 and 2027 in that space?
Yeah. So I think of apparel as an 80 billion unit opportunity per year that is roughly 40% penetrated by volume based on 2024 data, but is over 90% penetrated by brand or by logo. As I said earlier, most retailers have made the decision to move forward. Most are not 100%. So as I look at the retail opportunity, there's a lot of expansion left in just the existing customers that have already started a program, but there are net new programs. In 2025, we saw Old Navy, we saw Academy Sports + Outdoors, and we saw Aritzia all announce programs for the first time. Those programs will continue to ramp into 2026, and I suspect there'll be new programs launched in 2026 as well, and then layer on top of that, any stabilization or recovery in the macro.
I want to talk about general merchandise for Walmart a little bit. And are phase one and phase two of their general merchandising efforts complete? And do you expect more phases? I guess I ask because the program has been quite lengthy given their vast supply chain.
Yeah. It was lengthy because we were tagging items for the first time, working with suppliers who had never had any exposure to RAIN before. So that took a little while to get it up and running. And since then, the program's been moving quite nicely. There was certainly an impact from tariffs because a lot of the goods were made in China, and it wasn't as easy as it was in apparel to find a new production home for some of the general merchandise items. That being said, we are not fully deployed with phases one and two of general merchandise, so I would expect continued expansion with those existing phases in 2026. There has also been a significant amount of testing and work going on to what categories could be included in phase three, and I expect a phase three to be announced at some point in 2026.
Is that unique to Walmart where they have their suppliers tag the items?
No. Typically, the retailers want the item tagged as close to manufacturer as possible so they get the benefit throughout the chain. They do have the capabilities to tag in store, but we do not see that as a widely used option.
I wanted to ask about logistics. You're nearly done with the U.S. rollout of the second-largest North American supply chain logistics end user, obviously UPS. In the logistics market, where do you see the next leg of growth, and are there any pilots beyond the two end users that you already work with?
Yeah. So in the logistics space, our second-largest logistics provider is well in front of the rest of the competition, and they've been very vocal about the ROI that they're earning from removing manual scans to improving misload rates to taking cost out of their structure. So I think there's a general understanding that many in the industry are well behind that, and there's a sense of urgency to catch up. We see that urgency in our pipeline. We see that urgency in the pilots that are going on right now. It just depends in terms of timing and when those pilots can turn into full-fledged programs and how early in 2026 that might happen to understand the impact of 2026 calendar year.
Do you think that they're, because of what they've done, are they winning business because of what they've done? I mean, is that an example you can go to the other large North American guys and you can say, "Listen, you're losing ground, and here's the tangible evidence"?
Yeah. So yes, we have heard that they've won business because they have RFID solution. They are very good at being vocal about their ROI that they've achieved, about the wins that they're achieving, so I don't even need to go say that. It's in the public domain already.
And is there any way for you to, obviously, if it's that important to them, do you feel like you're getting all the economic value out of that engagement that you could?
I believe so. They're a very key partner. We call them a Lighthouse account. Our engineering teams work closely together. We work together on the solutions to build their solutions and then bring them to life, and it's an opportunity that we're rewarded with high endpoint IC share, which is our prize.
I want to ask about financials and margins. Maybe you can walk through some of the main levers and the puts and takes on gross margin. I think you framed some of the different M800 and M700. There's a big, I think it's a 300 basis points gross margin uplift. Just can you frame some of the puts and takes on gross margin?
Yeah. So we are in the midst of ramping the M800, and that's what's driving the gross margin lift. If you go back in time, prior to the M700, which is our first chip in 65nm node, our corporate average gross margin was roughly 50%. The M700, we delivered two things, or three things. First, it's a more performant IC than the prior generation R6, which was in 152 nm. Second thing we delivered was a lower price point to our inlay partners so they could be more competitive in the market and unlock new opportunities. And third, by shrinking the size of the die, we were able to reduce our cost, and that translated to a 300 basis point gross margin increase. The M800 is our second major chip in 65nm. We get 25% more die per wafer. We get an improvement in sensitivity above the M700.
And what that means to our inlay partners is they can leverage a smaller antenna to get the same read range. So the M800 helps take cost out of their BOM. In addition to that, we priced the M800 at slightly lower than the M700 to drive adoption while still maintaining the cost advantage so that when the M800 is fully deployed, we expect another 300 basis points of gross margin accretion. You're starting to see that in the fourth quarter. The fourth quarter, the M800 will be our volume runner, meaning it crossed over 50%. It won't ever go to 100%, but it will grow above 50% of our IC mix, and we signal that it would deliver over 100 basis points of gross margin accretion in the fourth quarter.
Can you just talk about balance sheet optimization and capital return, how you're kind of thinking about optimizing the balance sheet?
Yeah. The first thing that we're focused on as it relates to the balance sheet is the convertible debt. So about six weeks ago, we refinanced a portion of our convertible debt. So we issued $190 million, 0% convert, and simultaneously repurchased $190 million of our 1 1/8% convert. The goal there was multifaceted. First, we reduced our coupon. Second, we reduced the underlying dilution by a couple hundred thousand shares. And then third, and more importantly, we split up the maturities of our $287 million total convertible debt to sizable chunks where we can leverage the balance sheet when it comes time to retire. This is potentially our path off convertible debt, and splitting up the maturities was a key focus for us so that we could retire them in a dilutive-friendly way.
Can you just go back to your point about gross margin? How much more upside is there to gross margin? Where do you think it can go?
Yeah. It's a good question. So I think there continues to be more innovation we can do. There's a more advanced process node that you should expect us to be working on, even though there's still more iteration within 65nm. To the extent we can take cost out and still deliver a lower-priced IC to our end customers, I think that could drive gross margin. We are also not yet leveraging or seeing the benefit of a cloud service or a SaaS software offering on top of that. That is our ambition, is to leverage our platform and then put a software solution on top of it. We're building out the team to do it right now. We have technology. We don't have a product yet.
That team will be chartered with taking that technology, developing new technology, and productizing all of it in a way that drives a recurring software solution that would have SaaS-like margins.
Maybe to end, I just wanted to ask you about endpoint ICs versus systems. I think endpoint ICs are something like 80% of revenue today, but it sounds like systems are going to start to grow a lot more. Software will begin to grow. So how do you see the mix of revenue evolving over the next few years?
Yeah. Over the next few years, I still see Endpoint IC being the lion's share. I probably always see Endpoint IC being the lion's share. While it's not recurring, it is recurring. So once we tag a pair of jeans, when that pair of jeans sells, we tag the next pair of jeans. So I think that continues to drive Endpoint ICs to be the larger mix. There will always be a lumpiness factor to the systems business as we have a large program coming online that could drive the systems mix up in any given quarter. And we're still a few years away from the software becoming a meaningful portion of our revenue.
Got it. Well, thank you, Cary. We're out of time.
Thank you for having me.