Note: This author has a short position in Impinj (the "Company", "PI") and may buy or sell the Company's securities at any time. Please see full disclosure at the end of this report.

Impinj (NASDAQ:PI ) is a recent IPO overcome by hype. The shares seem to have turned into a vehicle to speculate on the "Internet of Things" theme, with the stock turning over its float ~10x more often than most companies and its share price fluctuating wildly on days with no news. RBC notes that it believes that "the Company is becoming a rare public IoT pure play". We believe Impinj has become a "story" stock with little substance. The shares recently appreciated ~34% on speculation that Impinj technology is included in the Amazon Go (AMZN) deployment. We believe that a sober examination of the evidence suggests that this speculation is largely unwarranted and that there is a very low probability that Impinj has any exposure to this retail concept at all. Further, we believe that due in part to misleading comments made by PI management, the market largely misunderstands the competitive positioning of Impinj and the valuation proposition of its products. With the CEO selling ~$2mm worth of stock and two pre-IPO investors selling out of their entire stakes (~$63mm collectively) in an underwritten offering that required RBC to terminate the IPO lock-up early, we believe we are not alone in our belief that the shares have material downside as insiders are "rushing for the door". As the market begins to appreciate the low probability of Amazon Go exposure, the flaws in Impinj's business model, and as insiders and pre-IPO investors begin to sell their shares, we expect a material correction in the share price. An additional ~11.9mm shares become eligible for sale on January 17th with the expiration of the original IPO lock-up, which will more than double the float. Barcode Handheld Scanner

Impinj makes products that enable communication via ultra-high frequency ("UHF") RFID, also known as RAIN RFID. The largest use case today by far is the tagging of retail goods (~60% of market) - mainly apparel. Given that apparel comprises a very large portion of the market, we will use this implementation to describe the technology. UHF RFID communication occurs between a tag and a reader. Tags are comprised of a tiny silicon chip (the endpoint IC), which is bonded to an antenna, and inserted into a clamshell (often put inside the paper price tag that is attached to the item). This whole assembly (known as an RFID inlay or tag) costs the retailer about $0.06-0.08 per item tagged. It is programmed with certain identifying information (i.e. the color of the shirt, its size, etc.). The store uses RFID readers or gateways to read RFID tags. The retailer can accomplish this by installing a collection of stationary readers throughout the store (i.e. fixed to the ceiling) or by purchasing a few handheld readers that their employees walk around the store with to read item tags. To date, the retail use case for UHF has largely been inventory management.

Impinj makes the endpoint IC (~66% of 2015 revenue) which it mainly sells to the inlay manufacturers. 65% of PI's 2015 sales of endpoint ICs were to top 3 customers Avery Dennison (AVY), Shang Yang, and Smartrac. Impinj also makes readers and gateways - this business represented ~26% of 2015 revenue. These products are largely sold to distributors who sell them to value-added resellers ("VARs") and system integrators who in turn sell them to the retailer as part of a larger solution. 39% of reader and gateways sales in 2015 were to PI's single largest distributor. About 6% of 2015 revenue comprised of sales of reader ICs to third-party reader manufacturers. Reader ICs are the "brains" of the reader. The remaining ~1% of revenue was software, development, service & licensing revenue.

Note that while there are many different forms of RFID (including NFC, low frequency, etc.) Impinj only makes UHF RFID products.

Amazon Go: Likely Misplaced Hype

On December 5, 2016, Amazon posted this video on YouTube unveiling its new retail prototype store called Amazon Go. Amazon Go promises to provide consumers with a checkout-free shopping experience. A customer simply scans their smartphone upon entering the store, picks up whatever items they want to purchase off the shelves, and walks out of the store with the items. The technology automatically updates your virtual "shopping cart" on your smartphone as you pick up items in the store and charges your Amazon account for your purchases as you leave.

On the following day (December 6th), Impinj analyst Mitch Steves at RBC put out a report on Impinj stating that "…we think the Company [Impinj] may be involved in the Amazon Go initiative", although he admitted that he had not received any confirmation of this from Amazon or Impinj. On December 6th and 7th, Impinj's stock went up ~34% relative to the December 5th close price. We believe a careful analysis of the facts suggests that it is highly unlikely that Impinj is involved in the Amazon Go initiative.

The first hint that Amazon Go is unlikely to use UHF RFID is the high degree of communication the platform has with the user's smartphone. As shown in the Amazon teaser video, the user appears to scan a QR code upon entrance to the store:

After entering the store, the user's smartphone appears to know what items are in their "basket" on a real time basis - as the video states "anything you pick up is automatically added to your virtual cart". Likewise, if the customer puts an item back, it is removed from your cart - per the video "if you change your mind about that cupcake, just put it back. Our technology will update your virtual cart automatically." If Amazon Go is using RFID to determine whether or not a specific user has a certain item in their cart, we believe it is highly unlikely that it is utilizing UHF RFID. The reason is simple - UHF RFID readers are not currently built into smartphones. As the IoT Journal states:

"iPhones are incapable of reading RFID tags, and while many Android phones can do so, they can read only Near Field Communication (NFC) high-frequency (HF) tags, not the ultrahigh-frequency (UHF) tags predominantly used for tracking consumer items."

Likewise, the founder of an RFID start-up states "UHF/RAIN RFID currently cannot be built into phones because of both the antenna size requirements and the power requirements". Others echo these comments: "There are no phones that I'm aware of that has UHF RFID reader integrated." Without having a UHF reader integrated into smartphones, we believe that using your smartphone to keep track of what is in your virtual cart would be impossible with UHF unless users were required to plug an external reader into their phone when entering the store. Based on the Amazon video, we believe that Amazon Go does not require any external hardware to be plugged into the user's smartphone. On the other hand, NFC RFID (which Impinj has no exposure to) readers are built into many smartphones. Since the Amazon Go FAQ states the requirement of a "supported smartphone", maybe NFC enables a portion of this technology.

Further, we believe that one of the main reasons that retail apparel is the dominant use case for UHF RFID tags is the relative ease of reading a UHF tag that is attached to a cloth surface. Per Impinj's own website ultra-high frequency RFID has been known to have problems reading tags that are on or near metal or liquid surfaces:

"If an RFID system operates at a lower frequency, it has a shorter read range and slower data read rate, but increased capabilities for reading near or on metal or liquid surfaces. If a system operates at a higher frequency, it generally has faster data transfer rates and longer read ranges than lower frequency systems, but more sensitivity to radio wave interference caused by liquids and metals in the environment."

We believe that tagging metals cans or bottles containing a liquid with UHF RFID would likely result in an unacceptable level of inaccuracy and missed reads. Given the abundance of these types of items in most grocery stores, we believe that the use of UHF RFID in this environment does not make sense.

Disclosure from Amazon itself seems to indicate that UHF is unlikely to be used in this system. An Amazon patent filed in late 2014 appears to describe the exact process utilized in Amazon Go:

"For example, if the materials handling facility is a retail store, users may pick items from within the facility, place the items into a cart, bag, pocket, or otherwise carry the items and the items are automatically identified and associated with the user. When the user exits the facility, the items may be transitioned from the facility to the user and the user charged a fee for the items. This may be done without the user having to undergo the additional step of checking out with a cashier…"

Based on a careful read of the details of this patent, we believe that Amazon is largely relying on cameras and imaging technologies to achieve their check-out free experience rather than any form of RFID. For example, look at the "item removal/placement sub-process" illustrated in Figure 5:

This figure (and the supporting text of the patent) seems to clearly illustrate that Amazon is using imaging technology to determine if an item is removed from the "inventory location" (i.e. shelf) or if it is placed into the "inventory location" (i.e. user decides they no longer want the item).

"As discussed in more detail below, one or more images of the user may be captured. For example, when the user reaches an inventory location and passes their hand into an inventory location within the storage area 130, one or more images may be captured of the user's hand prior to it passing into the inventory location. Again, when the user's hand is removed from the inventory location, one or more images may be captured of the user's hand as it exits the inventory location. Those images may be compared to determine whether a user has picked an item from the inventory location or placed an item in the inventory location."

Further, Figure 6 appears to show that imaging technologies are also used to determine which item a user has picked from the inventory location:

While RFID is briefly mentioned as a potential "other input device" that "may be used to assist in determining the identity of the items picked and/or placed in inventory locations", Amazon also mentions a laundry list of technologies including "pressure sensors, infrared sensors, a scale, load cells, a volume displacement sensor, a light curtain, etc." earlier in the document as technologies that may be used to detect when an item is added or removed from the inventory location.

"In addition to cameras, other input devices, such as pressure sensors, infrared sensors, a scale, load cells, a volume displacement sensor, a light curtain, etc., may be utilized with the implementations described herein. For example, a pressure sensor and/ or a scale may be used to detect when an item is added and/or picked from inventory locations. Likewise, an infrared sensor may be used to distinguish between a user's hand and inventory items. In another example, load cells may be positioned on a surface of the transition area to detect when a user is passing through the transition area."

Additionally, while the section that references RFID leaves open the possibility of UHF RFID use ("a passive RFID tag"), it also mentions a whole host of RFID technologies that Impinj does not have any exposure to like active RFID tags, NFC tags, and Bluetooth tags.

The Amazon teaser video states that its technology works through "Computer vision, deep learning algorithms, and sensor fusion." Combining this comment with the text of the patent leads us to believe that Amazon is using an imaging technology that utilizes algorithms to track customers and determine what they pick up, which is aided by sensors (i.e. pressure sensors, infrared sensors, volume displacement sensors, etc.).

We struggle to see how UHF could be economic in this type of use case. We believe that historically UHF has been largely used to tag higher value items that cost multiple dollars due to the high cost of UHF tags, which makes tagging low-value items largely uneconomic. For example, market research firm IDTechEx pinned the average cost of a UHF tag in 2015 at ~6.9 cents here ($324mm divided by ~4.7 billion tags):

Given the low selling prices present in a grocery store, we believe that it is largely uneconomic to tag many of the low-cost items that are present in a grocery store. For example, why would a retailer tag an 89-cent can of beans with UHF if the tag alone costs ~7 cents, or close to 8% of the total value of the item?

Investors should ask themselves - if Impinj truly is a large beneficiary of the Amazon Go concept, why did the CEO of Impinj sell $2mm worth of stock and two venture capital backers sell out of their entire stakes (~$63mm collectively) less than a week prior to the Amazon announcement in an underwritten offering that required an early termination of the IPO lock-up?

Even if we are wrong and UHF RFID is built into Amazon Go, given that this is a single store prototype, we believe drawing any conclusions about the potential widespread adoption of UHF for check-out free shopping is premature. In fact, the Wall Street Journal has reported that Amazon has been trailing two other retail concepts besides Amazon Go - a drive-through concept and a "multi-format store" which "combines in-store shopping with curbside pickups". Amazon has also recently denied reports that it plans to roll out a large-scale grocery chain in the near future: "An Amazon spokesperson told CNBC on Wednesday. 'We have no plans to open 2,000 of anything. Not even close. We are still learning.'"

We believe that Impinj has a structural cost disadvantage in the manufacture of endpoint ICs relative to its largest competitor NXP (NXPI). As a reminder, this is a critical product line for Impinj as it represented ~66% of total revenue in 2015.

Impinj operates a fabless model. Per the prospectus:

"We do not own or operate manufacturing facilities. Currently, all our endpoint ICs are manufactured by Taiwan Semiconductor Manufacturing Company Limited, or TSMC, and primarily post-processed by our subcontractor Stars Microelectronics (Thailand) Public Company Limited, or Stars"

In contrast, NXP has its own internal fabrication facilities and manufactures its endpoint ICs in-house. While we don't know what mark-up Impinj pays Taiwan Semiconductor (TSMC) (TSM) and Stars to manufacture and post-process Impinj's endpoint ICs, we do know that TSMC's blended gross margins are approximately 50%. This means that if the profitability of TSMC's business with Impinj is on par with its average client, Impinj is paying TSMC a ~100% mark-up that its top competitor doesn't have to pay!

Further, in October 2016, NXP announced that it was starting to manufacture UHF endpoint ICs using 12-inch wafers, an industry first. NXP commented that this move will yield a "significant increase in production capacity, improved assembly quality and efficiency, and most importantly, a reduction in manufacturing waste and electricity". Given the increased surface area of a 12-inch wafer, you can produce ~100% more endpoint ICs on each wafer than you can on an eight-inch wafer ((12*12)/(8*8) - 1). This meaningfully increases your manufacturing efficiencies and reduces costs since it allows the manufacturer to increase the throughput of their manufacturing process and leverage certain costs like electricity.

While it may be possible for Impinj to move its process to a 12-inch wafer at some point, since it does not own its own fabrication facilities we believe that it is largely at the mercy of TSMC.

Given these two key cost advantages that we believe NXP possesses over Impinj, we believe NXP can either reap much higher margins on its endpoint IC business than Impinj or use its cost advantage to lower price and win business away from Impinj.

Management's Misleading Market Share Statements

We believe that Impinj has led investors to believe that the competitive position of its business is far superior than it actually is. For example, management likes to tout its very high market share in each of its key UHF product lines:

In this slide, Impinj claims that in 2015 it had 70% market share in UHF "readers and gateways". However, this is caveated by footnote 2 which states that this figure "includes Impinj and Impinj-enabled stationary readers". An "Impinj-enabled" reader is simply a reader that is sold by a competitor that includes an Impinj reader IC inside. The reader IC represents a small fraction of the total cost of the reader - we estimate ~10%. While it is impressive that Impinj's reader ICs are in ~70% of all stationary readers and gateways, since sales of reader ICs to third party reader and gateway manufacturers represents only ~6% of total revenue, this strong market position is not that important to the investment case for PI. Instead, we believe it is much more relevant to examine Impinj's market share in the overall readers and gateways market, since this represents ~26% of total revenue.

We estimate that Impinj-made readers only have ~20-30% market share of the UHF stationary reader market. This calculation can be derived multiple ways:

1. We believe that solid shading in the PI slide above depicts Impinj's stationary reader market share, while the striped line represents incremental market share from reader IC sales to competitors. The solid shading appears to represent ~30% of the total pie.

2. In Impinj's July 2011 prospectus for its canceled IPO, the Company claimed its stationary reader market share was 25%:

3. Market research firm IDTechEx states that the UHF reader market was ~$173mm in 2014 ($68mm "UHF fixed portal" plus $105mm "UHF Embedded and handheld"):

We believe that ~50% of the embedded and handheld market are comprised of handheld readers, a market which Impinj does not have its own product for (more on this below). Using this assumption, we get to a $120.5mm addressable market for Impinj readers in 2014 ($68mm + 50% * $105mm). Comparing this to Impinj's 2015 revenue from readers and gateways (which is likely conservative since the market probably grew in 2015) yields a market share of ~19%.

This 20-30% market share estimate only includes stationary readers in the analysis. Mobile readers make up a substantial portion of the total UHF reader market. For example, Macy's (M) which has rolled out one of the largest and highest profile deployments of UHF RFID, uses handheld readers to track its items:

"By the end of this year or the start of 2017, Connell told attendees, more than 60 percent of all goods at most of Macy's stores will be RFID-tagged and cycle-counted monthly via handheld RFID readers."

Impinj even implies in its prospectus that handheld readers are the "core reading modality in retail stores" today:

"We see a significant opportunity for our gateways to become the core reading modality in retail stores, replacing handheld readers"

Besides the sale of reader ICs to other reader manufacturers, we believe PI has very little economic exposure to the sale of mobile readers. It lists two mobile readers available for sale on its website (here and here) but these are both simply resold versions of readers made by ATID and TSL, with PI likely earning very low margins on these sales.

Management's Misleading Platform Value Statements

Impinj likes to consistently talk about the value of its "platform" as a main differentiating factor. Impinj states that it believes that it is "the only company selling a platform spanning endpoints, connectivity and software". It believes this platform provides PI with a performance advantage over competitors and allows the Company to cross-sell its different product lines to customers. As highlighted in the prospectus:

"Comprehensive Platform. End users who deploy the Impinj Platform gain performance, reliability and ease-of-use we believe is unequaled by "mix-and-match" systems cobbled-together from competitors' components. We believe our success with these end users will drive broader adoption of the Impinj Platform."

"Cross-Sell and Up-Sell Our Platform within our Installed Footprint. We believe the majority of RAIN deployments today use one or more of our products, positioning us for future platform cross-sell and up-sell opportunities."

In fact, the Company used the word "platform" a total of 23 times in the transcript of its first quarterly earnings call.

The sell-side seems to have bought into this concept of "platform value" as various analysts note it as a point of differentiation:

"Impinj's end-to-end solution offering, including tag ICs, reader ICs, readers and gateways and software, positions the company to offer full solutions rather than point products, differentiating it from competitors." - Cannacord 11/3/16

"Impinj is the only company with the full platform (end-points, connectivity & software), which enables customers to improve their businesses. Impinj helps lower the total cost of ownership and remains as the number one brand in the industry" - RBC 11/9/16

Based on our research, we believe that there is limited value in the ability to offer the customer a full platform relative to a "mix-and-match" system. These views are reinforced by the following facts:

Endpoint ICs are multi-sourced: We believe that virtually all end customers (i.e. the retailers) multi-source their endpoints to ensure continuity of supply. As a result, all endpoint ICs must be interoperable and able to communicate with readers made by all reader manufacturers. Retailers appear to be more focused on making sure that UHF tags meet certain minimum requirements than forcing their suppliers to pick one specific tag that the retailer believes to be superior.

For example, Macy's has stated that 11 different tag models meet "optimal performance per Macy's requirements". These 11 tag models use six different endpoint ICs made by three different manufacturers:

If one of Macy's vendors wants to use a tag that is not included on this list, they can as long as Macy's approves it prior to shipping. We believe that open policies like Macy's are very common among retailers that use UHF tags. By giving their suppliers a wide range of choices regarding which UHF tag they can use on their products, it results in most retail environments becoming "mix and match" deployments of endpoint ICs.

Different Go to Market: Impinj sells its endpoint ICs to the inlay (tag) manufacturers (Avery Dennison, Smartrac, etc.) who assemble the endpoint IC and other components (like the antenna) into an RFID tag and sell it to the retailer. On the other hand, Impinj sells UHF readers and gateways through distributors to VARs and system integrators who typically sell the retailer a solution that includes readers, associated "middleware", calibration of the system, and other items. We find it hard to believe that Impinj's "platform" has much value when the key products that make up its "platform" are largely sold through separate channels.

We believe that shares of Impinj are priced for perfection. Peeling back the Internet of Things hype reveals a Company with a cost disadvantaged business model, overstated market share, and limited platform value. While the prospect of Amazon Go appears to have increased investor enthusiasm for the shares, we believe a more sober view of the facts suggest this deployment is unlikely to result in a meaningful opportunity for Impinj.

PI management believes that its business will scale to 12-16% adjusted EBITDA margins over the next 3-4 years as revenue grows at ~25% per annum (relative to the full year 2016 adjusted EBITDA margin of ~3.8% implied in the Q4 '16 guidance). We are skeptical of this implied operating leverage given the historical track record of the business: taking the midpoint of management's Q4 '16 guidance, adjusted EBITDA margin in 2016 will be less than half of what is was in 2013, despite a doubling of revenue over that same time period. In fact, adjusted EBITDA margin has declined every year since 2013.

Even if you give Impinj credit for five years of 25% revenue growth and the achievement of the mid-point of management's long-term adjusted EBITDA margin, we estimated that the stock is trading at ~32x 2021 earnings on a fully tax effected basis. Even if investors believe this type of earnings power is achievable, they should at a minimum discount these earnings back to today to reflect the fact that the Company will not generate this level of earnings for another five years. Using an 8% discount rate, the present value of our 2021 EPS estimate is only ~$0.71/share.

While we believe actual business results will materially underperform these projections, we think that the stock is meaningfully overvalued even under these optimistic assumptions.

Given that management and pre-IPO investors sold a material amount of stock in a secondary that yielded them $25.785/share after underwriting discounts and commissions, we believe that those who know the Company and its prospects much better than we do also believe the stock is overvalued at the current price.

This article was written by

Disclosure: I am/we are short PI. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

Android Handheld Barcode Scanner Additional disclosure: Note this author has a short position in Impinj (the “Company”, “PI”) and may buy or sell the Company’s securities at any time. All content in this report represent the opinions of the author. The author has obtained all information within this document from sources they believe to be accurate. However, such information is presented “as is,” without warranty of any kind – whether express or implied. The author makes no representation, express or implied, as to the accuracy, timeliness, or completeness of any information or with regard to the results obtained from its use. All expressions of opinion are subject to change without notice, and the author does not undertake to update this report or any information contained within it. This document is for informational purposes only. The information contained in this document may include forward-looking statements which would include any statements that are not statements of historical fact. Any or all of the author’s forward-looking assumptions or beliefs about future events may turn out to be wrong.