"Technology is not evolving anymore, technology is evolution."
Over the past month or two, there have been a couple interesting bearish pieces written on Mellanox (MLNX) that have caught my attention and got me to dig a little deeper into the story. (I'd been keeping an eye on name ever since it jumped 50% in one day on a significant beat and raise.) Because of my accumulated experience trading niche semi and hardware component plays over the years, Mellanox is a name that is hard for me to ignore. Extreme industry cyclicality, performance leapfrogging and fierce competition are a powerful combination that few if any niche component players can overcome. When a stock starts behaving like that is not the case, it is time to pay attention.
While it takes a good deal of luck to spot the next hot hardware play before it hits a sweet spot, it only takes a little patience to wait around for the same stock to become a headline grabber and performance leader to start shorting it for the inevitable decline. One of the main reasons for this is that over my past fifteen years following this space, I have yet to come across a sell-side model that actually does a decent job of accounting for industry cyclicality. Falling margins due to competition and much shorter-than-anticipated length upgrade cycles are usually the main culprits, but so are issues along the lines of customer concentration (usually one or two big OEMs are driving revenue by utilizing their bells and whistles) and slower than expected adoption rates if a product is really niche focused and performance driven. The net result of all of this is that you tend to get fantastic quarterly volatility and the type of extreme stock swings that attract hedge funds and experienced traders. Basically, this space is just about as bad a place as you can go if you are a buy and hold type. Sigma Designs (SIGM), Qlogic (QLGC), Stec (STEC) recently, Genesis Microchip, Sycamore Networks (SCMR), Portalplayer, Acme Packet (APKT) also recently, Zoran (ZRAN), and countless other names occupy the once hot tech component play graveyard.
Remember sayings like: the fiberoptic boom has just started, our flat panel microcontrollers are way ahead of the competition, LTE rollout assures session border controller demand will be robust for years to come, Broadcom (BRCM) won't replace us in the next iPod, EMC (EMC) is just starting to ramp on Zeus Iops and others will follow in the quarters ahead, they are the Intel (INTC) of DVD chips, and SAN switch demand is only in the 3rd inning. Eternal optimism at its best, and there was a time it even worked on me. But blindly buying into a story can be a good (and often expensive) lesson. Nowadays, there is literally just about nothing you can say to convince me that a component play won't blow up at some point within 12-18 months of the street falling in love with it. So, when I started coming across articles about Mellanox interconnects dominating the HPC space, and how WEB 2.0 and the data center are going to drive InfiniBand demand into the stratosphere; I determined it was time to start digging deeper.
Mellanox: A Short-Sellers Dream Tech Component Short
The company exhibits the classic firing on all cylinders characteristics that typically precede a decline.
1) 56 Gb/sec FDR InfiniBand has zero competition right now as Qlogic is still at 40GB/sec QDR
2) This is leading to record gross margins in the near 70% range as of last quarter
3) Intel Romley Upgrade cycle coincided with this performance lead and has been driving revenue
4) Intel's acquisition of Qlogic's InfiniBand also coincided with the Romley upgrade cycle leaving Mellanox with virtually no direct competition in this transitory integration period
5) All of the above is coinciding with some big WEB 2.0 players' technological infrastructure investments
Usually, when I come across a company like this I'd quickly counter that margins will come under pressure and market share losses will occur as competitors either undercut or leapfrog on performance, that the Romley upgrade cycle will fade much faster than expected, that the core HPC market is still niche, and that Web 2.0 addressable market for their product consists of only a handful of names who are willing to pay for performance at any price. I'd then look at the stock and try to ascertain what is being priced in and start to try and time my short position around catalysts related to the above.
But the Mellanox situation is a little different because while some of the above elements are obviously present, notably the Romley upgrade cycle, the shifting competitive landscape complicates things. The problem centers around competition and margins. When you expect less competition in the near term you can't bet on declining margins, and that obviously makes shorting a name like this at this high valuation rather challenging because the Street is going to continue to shower the stock with praise. But what if the reason there is going to be less competition in the near term is because all current competitors and potential traditional new competitors want nothing to do with the space over the longer term? That's the situation Mellanox is facing.
Mellanox: The Business Model Short
Mellanox' s business has been booming for the past six months because Intel's Romley Platform provides built in support for PCI Gen3 which provides the bandwidth needed for 56 Gb/sec FDR speeds. Great news right? Well, not exactly, because at the same time Intel is laying out a roadmap to bring the interconnect interface closer to, or onto, the CPU. To quote Intel's Raj Jazra, "We have not worked through all the details yet, and we certainly have not announced if fabric integration will happen in one step or two steps, but be prepared to be surprised."
As far as competitive threats go this is about as bad as things can get in tech component land. The interconnect business is all about improving performance (lower latency), scalability, and energy-efficiency. These guys make a living designing around the processor. So, what happens when the processor maker decides that the next step of evolution is to cut them out all together and bring the interconnect closer to the brain? You don't need much imagination to answer that question.
However, as futuristic talk is common in the tech space, it is not something you want to rest a short thesis on. If you are going to short a stock like this you need something concrete to go on. Like Intel's recent interconnect shopping spree.
In the last year, the chip giant has bought high-end Ethernet maker Fulcrum, Qlogic's InfiniBand business, and supercomputing specialist Cray's (CRAY) custom interconnect unit. Also during this time period, Intel competitor AMD (AMD) purchased SeaMicro which makes the high-end 'Freedom'3D mesh/torus interconnect. What is going on here is pretty cut and dry; the CPU makers have decided to cut out the network interface card middleman. And if that middleman happens to be a public company with a $4-5 billion market capitalization trading at 27x this year's estimated EV/EBITDA (the money minting press at Qualcomm (QCOM) trades at 10x) trailing earnings that is an opportunity you don't want to miss out on if you are an experienced short-seller.
What I Am Focusing On For This Short
1) The interconnect asset sellers are just as interesting as the buyers in this story.
Intel/AMD's acquisitions might be grabbing the headlines, but what has intrigued me more is who has been selling. A lot of bearish chatter around Mellanox usually focuses on the Qlogic InfiniBand sale as that was a direct competitor, and while that exit by the SAN switch maker says something; Cray's move is a lot more interesting. (Qlogic's InfinBand business dates back to their Pathscale and Silverstorm acquisitions in 2006, and as they never really competed as effectively as they would have liked in the space so their decision to sell is no surprise.) Cray's name is synonymous with high performance computing and designing custom interconnects is a critical part of its core business. Consequently, parting with 70 plus interconnect engineers is not an easy decision for this company and tells you something about where they see the space heading. If Intel is going to take over the core interconnect engineering work, how does Cray distinguish itself moving forward in the HPC market? What differentiates them from the likes of a Dell (DELL) or an HP (HPQ) after its Cascade system and the Aries and Gemini interconnect road map has run its course?
Well, according to this interview with Cray's CEO it seems software (system and interconnect) and power packaging and cooling will be the focus of the future. And for those that think Cray is shooting themselves in the foot here remember they made a leading move like this before when they exited the vector processor business. This is the type of red flag that you can't ignore if you are investing in a priced for growth tech component stock.
2) Buyout rumors for this stock make almost zero sense, yet they are always making the rounds.
Hot tech component stocks are always a popular place for the takeover rumor mill machine types to have fun. Sometimes the suitors make sense but the price is just too high, and other times the suitors make no sense even if the price was reasonable. In Mellanox's case, I'd argue that you are facing both problems. Multi-billion dollar technology acquisitions are not done for what a company is selling today, but rather for what they will be able to offer for the next five years. The server rack sellers like HP/Dell are not going to buy this business (despite the market caps neither could digest this transaction right now even if they wanted to because of the challenges facing their business) when they see a name like Cray exiting and their CPU supplier entering. The same goes for IBM (IBM) which already has its own custom interconnect offerings and also is a sizeable ($26 million last quarter) Mellanox customer as services has been their focus for several years. That leaves Oracle (ORCL) which is a significant Mellanox shareholder, and has a CEO who has a history of doing big deals. Personally, I can't see why Oracle would make such a move as the upside to their Exalogic and Exadata business by bringing Mellanox in house is not there. If you are selling engineered systems using Intel processors why would you bother picking up an interconnect business if your processor supplier is moving towards fabric integration? The logic is not there and the revenue numbers ($12 million to Mellanox over the past six months) don't add up as Oracle outright ownership would guarantee that other server competitors end up going elsewhere. Add in the fact that Oracle has had a heck of a time integrating Sun's business and you can basically conclude that hardware deals are off the table for a while (they said as much on their conference call if you want to believe them).
If Oracle is going shopping in the multi-billion dollar price range it will be in software as a service. That leaves the SAN switch Fibre Channel guys whose collective combined market capitalization is less than Mellanox, or a Broadcom/Marvell (MRVL) diversified integrated circuit vendor type both of which are not looking to go into this direction and wouldn't even consider it at this price if they were. If you wanted to get a little creative and crazy, Mellanox buying AMD makes more sense at this point in time than anyone buying them out. At least in such a transaction you could argue that Mellanox would be able to morph into a bigger fish that could try to compete with Intel, but this would be insanely risky as AMD's business is a total disaster. (Note AMD now has a market cap that is less than half of Mellanox) Of course a sleeper in the space could end up being Nvidia (NVDA), which recently snagged a $12 million Exascale research contract from the DOE that centers around their Kepler K20 processors, but at Nvidia's current $7.7 billion market cap Mellanox would also be too much for them to swallow. Now, a licensing deal with Nvidia somewhere down the road makes a lot of sense and you could make the same argument for AMD if they hadn't picked up SeaMicro, but in either case you are talking about backing the upstarts against a titan when Mellanox's current business is thriving precisely because of Intel Romley compatibility driven upgrades. So, as you can see, it's impossible to come up with a Mellanox suitor at these levels. Either AMD and Mellanox get cheap enough to warrant a big server player taking a gamble on pure proprietary component engineered systems, or NVDA and Mellanox or one of the ARM players get together somewhere down the road (again once Mellanox shares are significantly cheaper). Shorting a stock with plenty of buyout rumors when I am very confident, nobody, not even a reckless management team, will buy them at the current market price let alone at a premium to that price is something I like doing.
3) Customer Concentration is high and earnings visibility is low
Mellanox's top two customers accounted for nearly 50% revenue last quarter and management only provides a quarter's worth of guidance. This is a deadly combination if you are long a hardware component stock that is 6-8 months into a strong order cycle for a performance leading premium priced product. An order delay from one customer, even if that delay is just a temporary blip, can still produce the type of stock decline that few investors can stomach. Furthermore, it makes little sense for a management team that has seen its stock explode on the back of two robust quarters to provide annual guidance when the market is happy assigning growth multiples commiserate with their recent rate of revenue growth. But when you have a concentrated customer base and a stock that with each uptick becomes a bigger distraction for stock option owning employees, bad news becomes tough to hide and being early on the short-side a little less risky.
4) The CEO seems to have caught the excessive promotion bug
No less than 24 hours after receiving bearish criticism and experiencing their first greater than 10% down day since this recent 100% run began, the CEO was out talking up the stock. If you are shorting this is usually a good sign as it shows you are dealing with the type of personality that is likely to over-promise and under-deliver down the road. In this case, his statements about future revenue ("we can grow to over a billion in revenue"), and growth rates with respect to a merger partner ("As long as we think we can grow faster than the company we can merge into, the return on investment will be better to our shareholders, we will stay independent") say a lot about his take on the stock. Could be a billion dollar revenue business? Well, considering I wouldn't want to pay more than 2x sales based on where this space is heading how is that a good thing? And growing let's say 3x faster than your acquirer is not a sufficient potential measuring stick if your stock is priced like you will grow at 7x. I also am not a fan of talking up that suitors have come a knocking when you are not sharing anything on price. Considering what Mellanox shares have done over the last six months any real offer was most likely 50% below here and from Intel.
Paying up to $2 billion may have made sense for Chipzilla if they wanted to get a deal done because they would bank the profits on the current Romley cycle, own a component leader with little competition in the interim, and still be able to fulfill their longer-term goal of moving towards fabric integration. My guess is Intel management did what most management teams in this space have done over the years, and concluded that with their road map all other players in the space would be willing to part with their interconnect engineers and IP at minimal cost. Mellanox as the current interconnect king couldn't from a share price perspective afford to acknowledge this, so they traded away longer term viability for short term gains. (There is not much you can do when the likes of an Intel or Apple is calling the shots on where things are heading and they can hire enough engineers to compete with you in a year's time.)
5) The Stec Experience is Still Fresh in My Mind
Back in 2009 a small little company named Stec exploded onto the scene as the leading vendor of enterprise class solid state drive solutions. Their Zeus Iops product was considered a game changer, and their competitive edge as a OEM supplier unassailable. They had design wins at all the major system OEMs (EMC, IBM, HP, SUN, and Hitachi), and were expected to start seeing orders ramp in the back half of 2009 and on into 2010 and 2011. The future was bright, and Stec was quickly becoming a street favorite. And why not? Here was a company that could claim that Zeus Iops based tiered storage solutions provided for 18% less cost, 60% more disk Iops, consumed 17% less power and cooling, and required 30% few disk drives. They could do no wrong, and over a five-six month stretch their stock reflected that as it rose from a low $4 of at the start of the year to as high as $42 by the middle of the summer. Of course the street quickly fell in love with the name, consensus estimates for 2010 which started the year close to $.30 a share sky rocketed to over $2 a share by the time management announced their big H2 $120 million order from EMC. Things were so good that the company managed to pull off a big 9 million share secondary on Aug 3 at $31.( gross proceeds to insiders on this transaction were over $260 million which is now equal to Stec's total market value) But as is often the case in this space, the party didn't last for long. Demand from EMC hit a wall in Q3 and the other OEM's never managed to really ramp of production.
Then as they moved into 2010/2011 Intel, Samsung, Hitachi, and Seagate started to compete, and considering three of those guys were also NAND producers Stec found itself at a disadvantage. Margins started to fall, revenue declined year over year, they ended up being late to the PCIe party, and street estimates naturally came crashing down. Fast forward to today and Stec is now struggling to get its generation four product qualified at the OEM's, has had to shift its sales channel approach, and seen its CEO resign. Suffice to say the fact that the stock is at $6 and the company is now worth roughly 10% of its $2 billion peak is no surprise. And if you want to visualize this, just look at the way the numbers played out.
Revenue ($ Mil)
Gross Margin (%)
Year End Stock Price ($)
And if numbers are not enough, I always like to go back and dig up headlines to remind myself of how quickly things change.
Stec has some great ones…
Barron's Thinks Stec Could Eventually be Acquired- April, 2009
Industry analysts think STEC's disk-drive performance has about an 18-month lead on competitors- April, 2009
EMC qualifying other SSD suppliers- March, 2010
Stec plunges 37%: Hit by low price competition- July, 2011
Is it Time to Write Off Stec- Aug, 2011
"The SEC complaint is without merit. In the mean time, it will be business as usual at the SEC and I remain as committed as ever to maintaining the same level of outstanding service and top quality products to our customers and to growing our company." Stec CEO, Aug 2012
Stec CEO Resigns- Sep, 2012
Yep, things change awfully fast in this space. But I would like to point out that I actually went long Stec for a stretch after it collapsed in September of 2009, and that despite instincts to the contrary I somewhat bought into the story after grilling my fair share of industry insiders and sell side analysts covering the name. That didn't last long and I was back to shorting the name in early 2010. Now, am I saying Mellanox is the next Stec? No I am not. In the near term, Mellanox is much better off, but over the longer haul its challenges are probably worse than Stec's. Of course with Mellanox trading at over 2x Stec peak market cap being better off over the near term may not matter. Eventually more short-sellers will pick up on the story and start seeing an Stec and other tech component short-like historical opportunity here. As their confidence in the final outcome is likely to be high, getting rid of them and the pressure they will eventually bring to the shares will be no easy task. Getting ahead of that is in my opinion a prudent move.
Sleep on this Short
Put this all together and you should feel quite comfortable shorting this stock at these levels even if you have to endure another pop or two (I think max pain is $125-135, but I expect it to settle comfortably around $65-75 over the next six months). In the short-term Intel announcements regarding Qlogic Truescale and Cray Aries plans should be solid news catalysts as we approach 2013. I also think the end of the Romley upgrade cycle and weakening macro environment in tech space work in your favor here. And eventually, once the street wakes up a bit, there will be pressure to get Mellanox to be more forthcoming about their strategic long-term plans. No matter what the CEO says, ("Mellanox is also not just sitting quietly. We are also executing and I think will come out with solutions that may be better than what Intel is coming out with.") nobody is going to feel comfortable investing in a pure network interface card maker when the processor maker is telling you it plans to marry the two functions under its own roof.
Disclosure: I am short MLNX.