Description

Power Integrations (POWI) offers a compelling opportunity for short-side alpha this year. Like a number of our prior posted shorts, POWI’s initial appeal was low earnings quality. In this case, low earnings quality has persisted for at least 5 quarters through what has been repeatedly portrayed as a channel inventory adjustment and recently as one that is nearly complete. Sentiment seems to buy into this, but our analysis leaves us skeptical that the inventory adjustment and the work-from-home normalization fully explain the 37% drop in revenues since 2021. If we are right, POWI’s financial performance in 2024 could lead the investment narrative and debate in the direction of normalized revenues well below recent peaks and the risks of increasing competition from Chinese semiconductor vendors. In combination, these factors can put considerable pressure both on earnings and the forward multiple.

Note: For reference, POWI has been written up once before on the VIC, also as a short by MSLM28 back in August of 2019. Back then it was a relative bargain at 28x LTM EBITDA and 37x LTM FCF!

Business Overview

Power Integrations designs, develops, and markets analog and mixed-signal integrated circuits (ICs) and other electronic components and circuitry used in high-voltage alternating current (AC) to direct current (DC) power conversion. A large percentage of its products are ICs used in AC-DC power supplies, which convert the high-voltage AC from a wall outlet to the low-voltage DC required by most electronic devices (smartphones, appliances, computing and networking equipment, consumer electronics, personal computer, home automation, power strips, and industrial electronics). It also offers high-voltage gate drivers (either standalone ICs or circuit boards containing ICs, electrical isolation components and other circuitry) used to operate high-voltage switches such as insulated-gate bipolar transistors (IGBTs) and silicon-carbide (SiC) MOSFETs. These combinations of switches and drivers are used for power conversion in highpower applications (i.e., power levels ranging from approximately 100 kilowatts up to gigawatts) such as industrial motors, solar- and wind-power systems, EVs and high-voltage DC transmission systems Power Integrations sells its chips to electronics manufacturers and distributors such Avnet (AVT, a 27% customer in 2023, part of the 69-70% of POWI’s product that goes through distribution).

POWI describes its competition as follows:

“Competing alternatives to our high-voltage ICs for the power-supply market include monolithic and hybrid ICs from companies such as STMicroelectronics, Infineon Technologies and Sanken Electric… as well as PWM-controller chips paired with discrete high-voltage silicon or GaN transistors. Such controller chips are produced by a large number of vendors, including those listed above as well as others including NXP Semiconductors, Diodes Inc., On-Bright Electronics, MediaTek Inc., Renesas Electronics and, in recent years, an increasing number of Chinese suppliers such as Southchip Semiconductor, Chipown Microelectronics and Hangzhou Silan Microelectronics Co. Our gate-driver products compete with alternatives from such companies as Broadcom, Infineon, Mitsubishi Electric, Fuji Electric, Semikron and Hangzhou Firstack Technology Co., as well as driver circuits made up of discrete devices. Our motor-driver ICs compete with power modules from such companies as ON Semiconductor, Infineon, STMicroelectronics, Mitsubishi and Sanken as well as discrete designs from a wide range of other suppliers. In general, we expect competition from Chinese semiconductor vendors to intensify over time reflecting China’s stated aim to develop its domestic semiconductor industry.”

POWI’s key product areas and end markets are shown below:

While “Industrial” has grown into the largest part of the revenue mix, that description is a bit misleading in that it includes a number of consumer exposures including battery-powered tools, networked thermostats, power strips, uninterruptible power supplies, and smart home devices.

This is part of the explanation for the Industrial segment (-39% in 2023 as shown below) acting more like the overbought, WFH-hangover Consumer segment (-45%), and underperforming the performance of nearly any industrial company or indicator of industrial activity I have seen.

The table above also serves as a good high-level picture for where POWI is in the cycle, how they got there, and a point of reference for pre-COVID, pre-WFH normalized revenue ($420.7m in CY19 vs $441.3m in CY23, and $448.8m expected in CY24, with a stop at $703m in FY21). There are many moving pieces between those years in the revenue line alone (WFH demand surge, supply-chain-related market share gains in that period and their sustainability, the channel overbuying and then reversing course at various times depending on the end markets, new product areas and design wins, competition changes, pricing, and more). When combined with many moving pieces on the gross margin line and on the addressable market, POWI offers ample opportunity to build a contrarian case against the bull case which seems to rest primarily and simply on a turn in the channel inventory dynamics.

The optics of a recovery back to anywhere near prior peak revenue is analytically tempting, and teased by a management team that has discussed revenue capacity of $150m per quarter. But even with all the talk of design wins, product innovation, and sexy new end markets that double the 2022 addressable market by 2027, POWI’s top-line is only modestly higher than it was nearly 5 years ago in 2019. The headwind from channel inventory reduction on trailing revenue levels does not appear to be as large as one might suspect given the amount of discussion the issue gets on earnings calls and conference Q&A (maybe $40m LTM?). Even adding that back to LTM revenues to get to ~$480m suggests an underwhelming amount of real organic annualized growth looking through the cycle spread over 4 years from $421m in CY19.

The main thrust of our short thesis on POWI relates to inventory and related gross margin concerns. POWI management’s spoon-feeding of the channel inventory dynamics has drawn attention away from the issue of inventory on its own balance sheet, which has been building and inflected negatively and materially so in Q423. Gross margins should suffer, perhaps for a year or more as these inventories are worked down.

Using the most recent earnings call as an example of management’s sleight-of-hand and the resulting investor focus, of the 37 boxes in a Bloomberg transcript search with the word inventory, channel inventory won 30-7 against balance sheet inventory. The balance sheet inventory comments were dismissive (‘it’s just wafers’, see below), while the channel-related inventory comments were nearly all presented in a bullish light (either a certain end market’s correction is done, or it is almost done).

Below is the development of POWI’s DSI since Q121 (yellow = bad). Some build would be expected coming out of the tight supply/strong demand conditions of 2021. But the current DSI of 164 days compares to 71 in Q419 and 78 in Q418.

But as mentioned above, POWI management also likes to explain away their own balance sheet inventory as just being ownership of wafer inventory where their purchases protect their position with their strategic foundries during this weaker period of end demand. While what they say about it is factual (“most of it is wafers”, yes, 59% was raw materials in the MRQ), the reality is that there is still a problem brewing in the other 41%. Further, it would not surprise us if fixed or semi-fixed costs related to the procurement of the wafers or other overhead costs are being capitalized to the balance sheet as raw materials inventory, but that is only a hunch based on the unusually large percentage growth. In any case, we will give management the benefit of the doubt and focus in on the other categories of inventory.

In the table above, we are looking at the history of inventory metrics using only Work-In-Process and Finished Goods inventory, to remove the variable of POWI’s strategic wafer purchases. A few points to make:

• Note that traditionally DSI is calculated using COGS. We also like to calculate using sales which can remove some margin noise. It is shown both ways above, but we’ll usually refer to COGS-based DSI the rest of the way.
• The trends are material and weak in FG/WIP, with 3 of the last 4 Q’s showing 22-49% yoy increases. Q423 is 62% higher than the Q4 average since 2019. To ensure we are not being overly influenced by noise around the COVID/WFH demand surge and supply chain shortages, we compare that 142 FG+WIP DSI to the Q418-Q419 average of 82.5, an increase of 72% (this is not shown above).
• Note that we would not have been concerned about the 25-30% yoy increases in the metric in 2022, as they were normalizing towards longer-term DSI levels after a period of having worked down inventories in a demand surge. But starting in ’23, the builds became dramatic and well beyond prior norms.

    It is intuitive that this would happen - as POWI’s distribution channels worked down their own inventories, things would start to pile up at POWI. But a few points help make something that is intuitive, and in some ways obvious, still important in the contrarian/short case:

• This metric is actually larger than the channel inventory, and the swings much larger over the last few quarters. Using a weeks of inventory that grosses up the FG+WIP from a cost to a sales level (using the mid-point of POWI guided gross margin), weeks of “FG-WIP” inventory surged from 30.4 at the end of Q123 to 38.6 at the end of Q423, a period over which management was celebrating a decline in channel inventory of only 1.3 weeks (from 11.8 to 10.5 weeks).
• While understanding the channel inventory has the allure of being predictive of top-line trends and the appeal of having it spoon-fed at every earnings call and conference, this DSI metric we are analyzing does two things. First, it calls into question the insights of the channel inventory, showing that, counter to management’s “mostly wafers” dismissal, there is material on-balance sheet inventory to consider as well. In a way, the 3 weeks of progress working down channel inventories in 2023 is on a base of 39.5 weeks (13.5 weeks of channel inventory and 26 weeks of FG+WIP inventory as at 12/31/22) and on that basis things actually got worse in 2023. Said differently, entering ’23 POWI had channel inventories that were 6.5 weeks above their normal 7 weeks, and FG’s-WIP inventories that were 3-4 weeks above normal, for excess overall inventory of ~10 weeks. Exiting ’23, POWI has the excess channel inventory down to 3.5 weeks, but the FG’s-WIP (grossed up) is now 16 weeks above the pre-COVID average.  Secondly, because of this inventory build, POWI still has an overhead absorption-related gross margin hit to take over 2024, or longer. Note that POWI’s channel inventory decreased by 3 weeks in 2023, while in Q423 the FG’s-WIP Inventory, grossed up to a sales amount, is 16 weeks above 2018/2019 levels and higher by 12.6 weeks vs. Q422. So in terms of overall inventory, they really deteriorated (minus 3 weeks but add 12.6 weeks) while managing to keep investors focused only on the improvement.
• Magnitude and timing of the margin pressure in a situation like this is always difficult. Let’s throw some numbers out there anyway.  Let’s say the right FG’s+WIP DSI is somewhere in the 81-87 range (excepting the first 3 Q’s of ’21 when demand surged and the supply chain had issues, and the first 3 Q’s of ’20 that were COVID impacted, the metric has been in this range for 8 out of the 10 remaining Q’s prior to the large revenue drawdowns starting in Q422). The average is close to 85 in those 8 Q’s. If you decide to believe consensus of $122m in Q324 revenues, then the normalized FG’s+WIP DSI would yield a FG’s+WIP balance of 53.2m, down from 66.7m in Q4. Assuming 50% of the capitalized cost is fixed in nature, you’d have about $6.75m (half of 66.7 minus 53.2m) in costs that did not hit the P&L in ’23 but will in the next 3 Q’s (a swing of $13.5m on a yoy basis). This is not including any price discounting that might have to be done to work down inventories, and also not including the aforementioned potential that there are fixed capitalized costs in the raw materials line that would have the same dynamic. Even excluding those, on the 33.7m in operating income consensus estimates over the next 3 Q’s, $13.5m is a material swing factor on that level of earnings or lower.
• This leads into my next point – the decremental margins implied in POWI’s Q1 guidance and in consensus for the next 3 Q’s (especially the first half of ’24) do not appear punitive enough even before factoring in the accounting-related headwind. Studying the 14 historical quarters since Q317 with revenue growth or declines of 15% or more to limit the noise in the numbers yields a 57.7% average incremental/decremental gross margin. This compares to only 41.0% and 38.2% for Q1 and Q2 2024, where revenues are estimated to be down 15 and 16% respectively.

If you’ve made it this far, you are probably thinking, “I have not yet heard enough about POWI’s inventory.” You’re in luck because there is another angle worth discussing – the answers to the questions of just how badly has the distribution channel inventory held back their revenues, and what could a normalization mean to them?

Not being privy to the exact calculation of the ‘weeks of channel inventory’ disclosure made on each earnings call, we relied on some simplifying assumptions that should get us close enough:

• Current quarter revenue divided by 90, then multiplied by 7 equals a week of revenue
• A week of revenue multiplied by .7 to approximate distribution revenue
• The resulting product multiplied by POWI’s disclosed number of weeks to yield dollars of channel inventory

This yields the following analysis:

The most important takeaway for us is that the channel drew down $40.5m of POWI inventory in 2023 under this set of assumptions, so revenue would have been ~9% higher if not for that. This is smaller than one might think given the disproportionate amount of attention the issue gets. We are not sure if others have built a more robust turnaround scenario, but as we see it, the rebound does not look overly compelling. In fact without a material increase in end market demand, POWI’s channels still need to reduce their weeks of inventory to get from the current 10.5 down to the 6-7 weeks that management has said is normal.

Valuation/Comps

POWI’s absolute valuation and relative premium is a bit puzzling, especially with the much lower margin profile, and I wish I had a better handle on the bull case. Is there anchoring to the ~230m in average EBITDA 2022-23? Do growth investors believe in the doubling of the addressable market from ’22-’27, and perhaps believe that the channel drawdown is masking the progress there? Is it related to the latent value of their GaN technology, where there has been strategic interest and transactions among other players over the last year or so? Is it irrational ESG investors that like POWI’s ESG story? We are not sure, but we did see some things not typically seen in companies trading at 83x EV/EBIT:

• Persistently weak revenue growth through the cycle - It is not just the 2024 revenue growth where POWI is at the bottom of the comp table, and it is not just because of the hangover from the WFH craze and resultant channel inventory correction. Off of the 2019 base, 2023 revenues represent a 1.4% CAGR. This bumps up to 3.6% organic CAGR if you adjust 2023 up for the $40.5m estimated drag from channel inventory. Still not knocking anyone’s socks off. MPWR from the comp table trades at a similar 42.5x '24 EBITDA multiple, but nearly tripled its top-line from '19-'23, and has digested those gains much better than POWI in late '22 and CY23.
• Indicators of future growth from the financial statements are not exactly screaming “my addressable market is doubling” – Non-GAAP R&D expense in the 2H23 was flat yoy in dollars, while capex is only 8.8m in ’23, down from 22.9m/18m/14.3m in ‘19/’20/’21 and down 21.4% yoy.
• Indicators of strong IP that will protect them from Asian semiconductor competition seem to be lacking – one example is POWI’s stat that 48% of its sales are from product families introduced 2001 and prior. While making the point of the attractiveness of annuity-like revenue streams, it seems that the flip side of this coin is that there are large chunks of the business that are vulnerable as POWI may not be innovating sufficiently to stay a generation or two ahead. Presumably this is concentrated in the original business, where the original product introduction for high-voltage IC’s for switched-mode AC-DC power supply was in 1994.
• Questionable management – I won’t dwell too much here, and I recognize that the distribution layer is a hindrance to business visibility. But I think they should be a bit more careful about what they say given this visibility. Here is a statement from the Q223 earnings call on 8/3/23 (2H23 revenues ended up at 215m, down from 229.5m in the 1H):
  • “we do expect meaningful revenue growth in the second half of the year compared to the first half as inventories continue to improve and new designs go into production". This was also a month into a quarter during which channel inventory increased to 11.6 weeks from 10.1 weeks.

        Another credibility-damaging example was a number of statements made in Q122 and prior praising their own restraint in shipping to distributors at levels:

• "Well, one of the things we have done, actually very well, especially compared to other companies as I've heard is that, we are very careful to shift to real demand and the way we do that is, we monitor the distribution inventory. By the way, 75% of our revenue goes through distribution. So, we don't ship unless they are shipping through" - Sandeep Nayyar, CFO on Q122 call
• "The reason we don't ship to whatever customer wants is because that's just going to cause more problems because you'll end up building inventory at the customer" - CEO Balakrishnan, Q121 call (the following quarter, channel inventory blew out to 11.6 weeks from 7.1 weeks, despite POWI’s sales being very similar at 184m vs. 182m, ie. it was not because demand fell off precipitously)

Price Target

Triangulating the incremental insights from analysis on the channel inventory, the POWI inventory, and historical incremental margins and organic growth leads me to estimates below the Street.

The multiple is always trickier of course, but let’s be generous, and use a range of 20-25x P/E on my ’25 EPS excl. interest income, then add the $5.30 of net cash and marketable securities per share. This yields $31.19 – 37.66, or 46-55% downside.

Key Risks to the short

• GaN - With both Infineon and Renesas making acquisitions in the last year or so of companies with gallium nitride (GaN)-based power conversion solutions, it is conceivable that strategic interest in POWI from a third potential large player would be higher than one would otherwise think. The transactions validate the GaN technology roadmap, but at the same time create two viable competitors. Further, the cost to acquire this capability was more in the $340-830m range, far below the outlay necessary to pay a premium for POWI.
• EV’s – a large part of POWI’s SAM doubling came from ~$1B in EV’s. This is a risk to monitor, but they are a tiny player (“tens of dollars of content” per slide 23 of March 2024 deck, and products in pipeline to expand that to $100), and it is not clear to me that there is enough value-add to unseat incumbents except perhaps in GaN based onboard charging.  
• ESG – they spin a pretty good energy efficiency story, even devoting 10 (out of 50) pages of their deck to their ESG story. So maybe no one will care that they trade at 82x EV/EBIT.

Disclaimer: the author of this posting has a short position in the security discussed. The author reserves the right to transact in any way in the shares or to change his views without any obligation to inform anyone. Please do your own diligence and consult your financial, legal, and/or tax advisors before acting on any advice. While the author has tried to present facts it believes are accurate, the author makes no representation as to the accuracy or completeness of the analysis herein. Reader waives any cause of action against author related to the analysis above.

Catalyst

- Earnings results