Description

When a management with a reputation for brilliance tackles a business with a reputation for bad economics, it is the reputation of the business that remains intact. – Warren Buffett

Situation Overview

Things have been crazy in publicly traded semiconductor stocks, and I started a discussion topic on the sector in November. Traditionally, downturns in the cyclical semiconductor sector led to cheap valuations and the chance for bargain stock purchases that paid off handsomely in upturns. However, this pattern did not recur in the most recent downturn. In a nutshell, the sector sold off in late 2018 but remained at valuation levels average to high compared to the post-GFC era. An industry downturn materialized in late 2018 and continued throughout 2019 despite predictions of a 2H rebound that never occurred. Amid the enthusiasm for stocks in general and technology stocks in particular during 2019 the SMH ripped 62% driven by an 11.5% decline in 2019 EPS estimates offset by 71% multiple expansion. To highlight the current semiconductor investor complacency, during 2019 out year (2020) revenue estimates for MXL declined from $431M to $319M. Normally a disastrous 26% outlook reduction such as this would cause a stock to blow up and trade at a multiple of 1-2x revenue, instead MXL stock was up “only” 21% and trades at 5x revenue. 

Semiconductor valuation multiples are now at their highest levels this side of the dot com bubble, and several companies that have organically shrunk the last few years even outside of the most recent downturn are trading at multiples previously reserved for high quality growth companies. The sector would seem ripe for short opportunities, but the problem is that at least up until recently the industry has stabilized and appears to be entering an upturn, begging the question of why the stocks would decline now after appreciating on poor fundamentals last year. Just to keep things spicy, coronavirus-related supply chain shutdowns may push the recovery out still further and cause semi companies to miss their more optimistic Q1 2020 guidance.

However, I believe that the elevated multiples have made many semi stocks vulnerable to sharp corrections if forecasts for growth do not materialize over the coming year. The trick is finding companies who may disappoint with sales growth despite the improved industry conditions. I believe that Lattice Semiconductor (LSCC) fits the bill on both counts. LSCC traded at around 2x sales for most of the last five years, and even lower before that. The company hit a revenue peak of $427M in 2016, and has declined organically over the last few years to around a $400M run rate today. After a sales process that resulted in a deal with a Chinese acquirer at less than half of today’s price that was ultimately blocked by the U.S. government, the company brought in a new management team that has slashed costs while doing little for the top line so far. Somehow, enthusiasm for these cost cuts and excitement about future growth combined with the rapid increase in industry multiples has led LSCC’s revenue multiple to nearly quadruple to over 7x. I believe that the loss of Huawei as a significant customer will cause growth to disappoint, and a re-rating to even 3.5x revenue (high by historical standards for this company) could cause the stock to decline almost 50% to $10.00- which would still be above any M&A bid the company received in its sale process on a higher revenue run rate. 

Company Overview and History

Lattice has been the subject of two relatively recent VIC write-ups, a poorly timed short write-up from September 2017 and a long write-up from September 2019. I encourage you to read both of those for background and context, and in this piece I hope to respectfully convey why I disagree with the long recommendation.

Lattice primarily makes small, low-power Field Programmable Gate Arrays (FPGAs). FPGAs are a sub-sector within semiconductors. As the name implies, an FPGA does not have fixed electrical gates in the same manner as application specific integrated circuits (ASICs), microprocessors or microcontrollers. FPGAs can be programmed to change their gate arrangements depending on the needs of the end application. The flexibility of FPGAs means that they are generally larger, slower and more expensive than an ASIC used for the same purpose and engineers only use FPGAs in specific cases when it is impractical or impossible to use an ASIC or a microprocessor/microcontroller + memory.

For years, Xilinx and Alterra (now part of Intel, acquired for a high price) made for a profitable and growing FPGA industry duopoly. Xilinx and Alterra’s flagship products tend to be large, expensive and made on the latest semiconductor manufacturing process nodes. Their chips tend to cost tens or hundreds of dollars and are used in some of the most complex data center computing, communications and defense applications although both companies also sell some cheaper, older parts. Historically, both Xilinx and Alterra were excellent businesses and considered semi industry leaders.

By contrast, Lattice and its poor relation Quicklogic (QUIK) make smaller, lower power FPGAs on trailing process nodes. This has historically not been a great business, for the simple reason that it is much easier and less complex to replace a smaller lower power FPGA with an ASIC or microcontroller + memory. The typical pattern for LSCC or QUIK would be that they would get a design win, but if the design achieved sufficient volume an ASIC would be made or the functionality needed would be designed into a SOC and the FPGA would be designed out. Many times one of these small FPGAs would be used as “glue logic” to tie two protocols together when two more important chips didn’t directly interface for whatever reason. I chuckled a couple of weeks back when Lattice spiked on news that an early-stage AI startup bought by Apple used its chips for demos- “Lattice won Apple!” Leaving aside the fact that the products of this acqui-hire by Apple may not see the light of day for years, and we have no way of knowing what product if any they will be integrated into, investors seem to be forgetting that Lattice has already won Apple and was part of the iPhone 7. In the next generation, Lattice was promptly designed out and the Apple design win did little for their business or stock price.

Due to this inherent problem with the small FPGA business, Lattice for many years treaded water as you can see in a long-term stock price chart. From 2010-2014 the company had uneven revenue growth and profitability, and in a bid to diversify out of its low power FPGA niche Lattice acquired ASIC maker Silicon Image in March of 2015. Silicon Image made ASICs for HDMI and some other applications, although arguably its most valuable asset was the stream of HDMI IP royalty payments. The combined company struggled and continued to experience organic revenue declines, and in November 2016 LSCC announced it would be purchased by Chinese fund Canyon Bridge for $8.30 per share. According to the merger proxy (and as detailed by the VIC short report I linked to) only one other U.S. semi firm submitted an offer for LSCC, well below the Canyon Bridge price. The Trump Administration blocked the deal in September 2017, although it was unclear exactly why since they sold few if any parts to the U.S. military. The stock traded around $6.00 after this.

Recent Developments

While the short report written back in 2017 turned out to be misguided, the author was correct that the company would not meet then-current revenue forecasts. In late 2017 the 2018 consensus revenue estimates were around $426M while the author predicted closer to $370M. As it turned out, Lattice reported $399M in revenue, up slightly from 2017 but down from the peak $427M in 2016. In January of 2018 activist investor Lion Point filed a 13D, which led to a quick settlement that reshuffled the board and led shortly thereafter to CEO Darin Billerbeck stepping down. The company commenced cutting costs and announced Jim Anderson as CEO in August 2018. Jim quickly brought on the rest of a new management team.

Lattice generated 3% revenue growth in 2018, but pro forma EBIT grew from $35M (9.1% margin) to $67M (16.7% margin) largely from cost cuts. The stock started to move up near the end of 2018, but then appreciated an incredible 177% in 2019 as new management slashed costs further and grew EBIT to an estimated $93M in 2019 (23% margin) on the back of 1% expected revenue growth. On May 20th the company held an analyst day that laid out its vision for growing the FPGA business single digits in 2020 and 2021 before accelerating to double digit growth in the out years on the back of edge AI, telecom, compute/server and industrial/auto applications. The company laid out a target model of 62% GMs and 25-30% operating margins, and interestingly enough the company achieved the low end of its operating margin target in Q3 of 2019. I must give Lion Point credit for a great activist campaign and the new management team for making Lattice a much more efficiently run firm, although given the history of the company and the space I am cynical about these growth projections being realized.

The incredible appreciation in 2019 was driven by revenue coming in slightly better than expected (after estimates generally being cut the past few years) and EBIT coming in decently better than expected thanks to the rapid cost cuts. Total non-GAAP opex decreased from a peak of $218M in 2015 to only around $146M this year, so I think cost cuts are tapped out and will perhaps affect forward looking growth. The stock performance was also driven by the expansion of multiples across the semiconductor space. Today LSCC appears to me to be a company trading on peak multiples (7x revenue vs. historical 1-2x, 30x forward earnings) on peak margins (25% is by far the highest non-GAAP operating margin the company has ever reported and within their target range). Amazingly, LSCC is now trading at a higher revenue multiple than its historically-much-more-successful FPGA comp Xilinx which trades at 6x vs. around 5x historically.

The Stock is Expensive, So What?

As we all know a stock is not a short just because it is expensive, in fact in recent years buying the most expensive stocks would have led to significant outperformance. So why am I risking looking stupid in front of LSCC’s 2/11 Q4 earnings report by writing this up as a short now? I believe revenue growth will disappoint and the company will struggle to deliver management’s targeted single digit revenue growth in 2020, much less the double digit growth they promised beyond that. While management has done an admirable job improving Lattice’s operating margins, I think the cost cuts were at best neutral to Lattice’s growth and may come at a price in terms of innovation and marketing. Fundamentally, I don’t think they can change much about the difficult position lower power FPGA chips have been in historically.

In fact, I think people missed cracks in Lattice’s latest quarterly report. The new management team had delivered beat-and-raises in each quarter since taking over Lattice, but in the Q3 2019 results the company reported $103M in revenue, exactly in line with consensus. Further, they guided Q4 to only $97-$103M vs. $104M consensus- the first guide-down under the new management team. Digging further into the details, the revenue segment disclosure raised my eyebrows. Lattice breaks its business into comms/computing (39% of expected 2019 sales), mobile/consumer (19%), industrial/automotive (37%) and licensing and services (5%). The licensing and services are almost 100% GM and mostly consist of HDMI and MHL royalties acquired from Silicon Image. Analysts had expected about $4M in licensing revenue while the actual came in anomalously high at $6.2M, much higher than the previous two Q3s. This means that chip revenue actually missed, and without this unexplained jump in royalty revenue the company would have missed and lowered, not a great look when your stock is up almost 200% in the last year.

However, this does not mean that Lattice actually collected $6.2M in cash from royalties in the quarter. In fact there appears to currently be a dispute among the HDMI consortium and no royalty sharing agreement has been formally adopted. Perusing the 10-Q you will read that contract assets have grown from $9.1M as of 12/29/2018 to $15.4M as of 9/38/2019, and Lattice has recognized $8.9M of HDMI royalty revenue to date but collected only $2.6M. The company recognized only $5.2M in HDMI royalties during the first half of the year but this bumped up to $3.7M during Q3. There could be an innocent explanation for this and some reason they become more optimistic on eventual royalty collections during Q3, but it smells like earnings management to me.

Huawei Issue

More importantly, I believe that Lattice has a serious Huawei problem. Those of you who follow the semi sector or the news generally know that the U.S. government placed an export restriction on technology sales to Huawei. During 2019 this affected U.S.-based semiconductor companies to varying degrees with some stopping all sales to Huawei and some continuing to ship parts for non-infrastructure applications. This issue continues to develop, and many companies have applied for exemptions to continue shipping to Huawei. Just last week CREE took an inventory write-down on parts it could no longer ship to Huawei. XLNX gave weak calendar Q1 guidance partially because having Huawei out of their customer base has led to declining revenue in their WWG. Lattice and its lawyers seem to have taken the relatively aggressive stance that they would continue shipping every part Huawei wants unabated. From the Q2 transcript:

Before turning the call over to Sherri, I want to comment briefly on Huawei. In mid-May, we stopped shipments to Huawei when the government order was given. However, we resumed shipments to Huawei in late Q2 of those products that we determined to be in compliance with export restrictions.

Later….

Matthew D. Ramsay, Cowen and Company, LLC, Research Division - MD & Senior Technology Analyst [2]

 Congratulations on really strong results and what I imagine was a bit of a turbulent period with Huawei.

 Jim, I wanted to ask, you had mentioned that you guys have suspended shipment to Huawei and then resumed it at the end of the quarter. There's been various companies that have said they've resumed full shipment, there's some that have said they've excluded it from Q3. One of your FPGA competitors, Xilinx, said they're shipping partial product going forward. Maybe you could add a little bit of context. I know it was supposed to be mid-single digits for the year, and I'm just trying to understand how much is in or out and we can gauge, I guess, the relative magnitude of the strength of the rest of the business.

 James Robert Anderson, Lattice Semiconductor Corporation - President, CEO & Director [3]

 Yes, sure. Thanks, Matt. Let me start with Q2 and then I'll give a little color in Q3 as well. So for Q2, first, just to reiterate, so clearly, we stopped shipments in mid-May when we got the government order. And then we worked with our internal legal team and external legal counsel to do a pretty detailed analysis to examine which products we believe were compliant with export restrictions. And so near the end of Q2, we started shipping those products that we determined were compliant with export restrictions. And so -- and after that, we restarted shipments roughly in the last couple of weeks of Q2.

 But if I look at kind of where we ended up with Huawei revenue in Q2 versus what we had kind of originally assumed as part of our original Q2 guidance, it was roughly the same. So we kind of ended up where we had expected with respect to Huawei revenue. And then moving forward into Q3, I'd say, our Q3 guidance reflects the current demand outlook that we have from Huawei, again, for those products that we've already deemed compliant with the export restrictions. So that's factored into our Q3 guidance. So hopefully, that's helpful, Matt.

So, apparently alone among U.S.-based semiconductor companies shipping to Huawei, the export ban has had almost no effect on Lattice’s business. I am of the opinion they are violating the spirit if not the letter of the restrictions, and this kind of intransigence is why the Trump administration is seeking to further tighten Huawei export rules as recent news reports have indicated. But whether or not these tighter rules are implemented, from industry discussions I believe Huawei is in the process of completely designing out Lattice products as they widely recognize that sales of silicon from U.S. based suppliers are unreliable. As discussed in the background section, Lattice small low power FPGAs are among the easiest chips to design out and replace with ASICs. Diligence from expert network transcripts have led me to believe that Lattice’s 5G business had more traction with Huawei/ZTE than Nokia/Ericsson. And oh by the way, for the first time in a long while Lattice has new competition in small low power FPGAs from China- Gowin Semiconductor and they may be well-positioned to pick up the pieces.

How much will this impact Lattice? They have disclosed that Huawei is a single digit customer. I believe they are a mid to high single digit customer. At 7-8% Huawei would be a $28M-$32M customer in 2019. From Lattice’s disclosures we know they didn’t have a 10% customer in the first three quarters of 2019, although interestingly enough they disclosed that their top five customers were 24% of sales in the first quarter compared to 16% in Q1 2018, indicating increasing concentration and a decline in revenue from non-top five customers. 2019 revenue growth is only expected to be around 1%, and within this the comms/computing growth was $34M offset by a $23M decline in mobile/consumer, an $11M decline in industrial/auto and a slight gain in licensing/services (so far from greater accruals than cash collections). As I look to 2020, I think mobile/consumer will continue its multi-year decline, industrial/auto will revert to some level of growth and licensing will be flat to down. The hopes for $20M or so in sales growth clearly lie in the $160M comms/computing sector. Losing a $30M customer likely means this sector will be flat to down, although I do think it will be aided by a resumption of growth in the server market. Thus I believe in 2020 Lattice is looking at yet another year of flat revenue.

Note that my timing could be slightly off on this, but I don’t believe it will be off by much. IPHI just reported strong results and guidance, partially thanks to very high demand from Huawei both in Q4 and Q1 as Huawei seems to be building inventory and perhaps making last-time buys as it prepares to transition its supply chain away from U.S.-based silicon. It is unclear if Lattice will follow the same pattern, as I believe they have already benefitted from Huawei strength in 2019 and their parts are easier to replace with ASICs than IPHI’s. However, it is possible they will come in at the high end of lowered Q4 guidance and have a decent Q1 guide if Huawei remains strong for them. But I do believe they will revert to flat to down Y/Y growth by Q2 or Q3 of 2020.

I believe the 7x revenue multiple indicates that the buy side has higher growth expectations and leaves the stock in a vulnerable spot. When investors realize that new management can make the company more efficient but can’t necessarily make it grow, the stock should re-rate to more typical multiples for a low growth company. A $10 stock price puts the company at 3.5x revenue, 12x EBITDA and 17x earnings which seems more appropriate. I believe it could get there in a hurry if they get publicly spanked for their aggressive stance on Huawei and/or that revenue comes out faster than new revenue from a recovering industrial/auto and improving server market comes in. Note that just today, semiconductor maker AOSL disclosed in its quarterly press release that the DOJ recently commenced an investigation into the Company’s compliance with export control regulations relating to certain business transactions with Huawei and guided down due to the DOC requesting they cease shipments to Huawei. Draw your own conclusions on this front.

Disclosure: The fund that I work for is short shares of LSCC. We may buy or sell shares of LSCC at any time without notice.

Catalyst

• 2020 sales growth disappointment
• Huawei designing out Lattice products
• U.S. government action to force U.S. semi companies to comply with Huawei export restrictions