Business

EZchip is Overvalued, Inflating EPS, Handicaped: Kerrisdale

EZchip Semiconductor

EZchip Semiconductor Ltd. (NASDAQ:EZCH) is over-valued and is inflating its GAAP earnings according to a report from the famous short biased value hedge fund, Kerrisdale Capital. Started by Sahm Adrangi, Kerrisdale Capital gained its fame shorting Chinese reverse mergers. The hedge fund was up 29% in 2012 (a lackluster year for the firm compared to the 200% return in 2011). Now the hedge fund has been targeting less “wonky” companies. Kerrisdale Capital has just issued a 22 page report detailing the short case. Kerrisdale notes the company faces :immediate Competitive Threats, Customer Defections and Unsustainable Valuation.” Disclosure: Kerrisdale is short shares of EZCH. Please review their full disclosures at the end of our report.

Below is a summary regarding EZchip Semiconductor Ltd. (NASDAQ:EZCH) from Kerrisdale including their “red flags” followed by the full presentation in scribd:

EZchip Semiconductor Ltd. (NASDAQ:EZCH) shares currently trade at $23.97, implying an egregious 7.5x 2013E EV/Revenue and a 32.5x GAAP P/E, surprising multiples for a business that has repeatedly demonstrated its inability to grow. As we frequently see in mispriced businesses, a high-level sector story – in this case the need for greater bandwidth on capacity constrained carrier networks – has steered investors into an overpriced stock with many idiosyncratic risks. We believe that investors are not properly assessing the competitive risks posed to EZchip’s business or the limits of the addressable market, especially in light of the EZCH’s astronomical valuation.

 

Before 2012, EZchip Semiconductor Ltd. (NASDAQ:EZCH) was fortunate enough to have little outside competition in high-speed network processors. But since the start of last year, two well-funded industry leaders have mounted attacks into EZchip’s core Network Processing Unit (NPU) business. To begin with, Marvell Technology Group Ltd. (NASDAQ:MRVL) began to compete for NPU market share following their January 2012 acquisition of EZchip’s then-largest competitor, Sweden-based Xelerated. Given EZCH’s reliance on Marvell Technology Group Ltd. (NASDAQ:MRVL) as a conduit in the foundry relationship with Taiwan Semiconductor (TSM), the new arrangement appears conflicted. The dynamics are akin to Coca-Cola (KO) relying on Pepsi (PEP) for the manufacturing of Coke syrup.

 

But the most substantial competitive risk to EZchip Semiconductor Ltd. (NASDAQ:EZCH), in our opinion, is the entry of Broadcom Corporation (NASDAQ:BRCM) into the high-speed network processor market. Through its acquisition of NetLogic in 2011, high-end semiconductors have become a strategic priority for BRCM. In April 2012, Broadcom introduced a full-duplex 100Gb (i.e. 200Gb/s) NPU, pitting EZchip’s upcoming NP-5 in direct competition with an industry leader. While EZchip touts its 200Gb/s headline figure for the NP-5, the chip is merely a full-duplex (two-way) 100Gb/s device, giving it the exact same bandwidth as Broadcom’s chip. Another critique of BRCM’s NPU is that it lacks an integrated traffic management system (“TCAM”). This argument overlooks the fact that Broadcom Corporation (NASDAQ:BRCM) can add traffic management and other functionality into the Ethernet switch component, a product line that Broadcom Corporation (NASDAQ:BRCM) also manufactures. But most troublingly for EZCH’s investors, Broadcom now produces each of the individual components of the line card (TCAM, NPU, multi-core processor, etc.), allowing them to write software that transcends across individual components. This might allow BRCM to optimize the efficiency of the various hardware inputs and constitute a strong selling point versus EZCH’s NP-5.

 

But even if one believes that the BRCM component is somehow inferior, credible alternate vendors in the marketplace could drive down NPU pricing. This issue is overlooked by many analysts who instead rely on EZCH management for guidance. In their Q4 2012 Presentation, EZCH tells investors that it expects a 40% increase in unit pricing between 2012 and 2016. Not only is this assumption inconsistent with the entry of Broadcom and Marvell Technology Group Ltd. (NASDAQ:MRVL), it’s contradictory to the very nature of Moore’s Law, where more computational power is delivered at ever decreasing costs over time. NPU pricing is also limited by a constant threat of NPU replacement by in-house designs from the likes of Cisco, Huawei, ZTE, and other router manufacturers. Just over three years ago, Juniper decided to replace all of its EZCH chips with in-house designs on future router designs. EZCH shares fell 14% immediately after the 2009 announcement. Huawei may have made a similar decision as they’ve delayed all EZchip Semiconductor Ltd. (NASDAQ:EZCH) orders as of Q4 2012. Management speculates that Huawei is using an in-house solution for at least a portion of their NPU demand, but they admit to having almost no visibility into this key customer account (EZCH Q4 2012 Call). A business that lacks pricing power should earn a commodity-like multiple, not the stratospheric valuation premium that EZCH currently boasts.

 

Foreseeing the threat in their core NPU market, EZchip Semiconductor Ltd. (NASDAQ:EZCH) has rushed to publically announce a shift towards the data center market. EZCH’s L4-L7 next-generation processor (“NPS”), if successful, won’t generate meaningful revenues until 2016+ (Sept 5th NPS Call). But that hasn’t stopped some investors from glowingly referring to the NPS as a ‘game changer.’ Unfortunately for the EZCH bulls, a privately-held Intel spinout named Netronome already produces a 200Gb/s flow processor with L2-L7 functionality that has won widespread industry praise. And as one of only a handful of outside partners with access to Intel’s semiconductor foundries, Netronome’s chips can be produced at the 22nm scale and below. Compare this to EZchip Semiconductor Ltd. (NASDAQ:EZCH), whose next-generation NP-5 will only be produced at the 28nm scale with Taiwan Semiconductor.

Lastly, we believe that Wall Street research analysts are twisting forward EPS, and by association, overstating their EZCH price targets. Through the use of a non-GAAP technique that excludes stock-based compensation (“SBC”), we believe that the $1.13 2013E consensus EPS is inflated by about 35%. Wall Street analysts have no logical grounding for excluding these costs. Given EZchip’s meager revenue stream in relation to its $690m market capitalization, its history of unfulfilled promises, the rapidly emerging competitive threats from Broadcom and Marvell, and EZCH’s technological disadvantages in the data center market, we believe that EZCH’s share price is poised for a sharp correction. Like many other technology hardware manufacturers dependent on a single customer for much of their revenue, EZCH’s stock is one design loss away from falling by as much as 45%.

 

Summary of Red Flags

We believe that EZchip is significantly overvalued for the following reasons:

  • Broadcom’s 200Gb/s Network Processor Should Pressure EZCH’s NP-5 Market Share and Price Point. EZCH has historically benefited from its cushy position as the leader in a duopoly market with Xelerated (now Marvell). But that dynamic changed drastically this year when Broadcom announced its entry into the market in mid-2012 via the BCM 88030 network processor. According to recent management comments, Broadcom expects to enter production and have customers in 2013, some of which could be poached from EZCH (BRCM 2012 Analyst Day). EZchip has told investors that “we believe that substantially all NP-4 customers will select the NP-5,” but that sounds more like wishful thinking than fact to us (EZCH Q4 2012 Call). While EZchip touts its 200Gb/s headline figure for the NP-5, the chip is actually a full-duplex (two-way) 100Gb/s device. This is precisely the same speed as Broadcom’s BCM 88030, a full-duplex 100Gb/s processor. The confusing nomenclature may be misleading some investors to brush off the risks from Broadcom. Broadcom began its push into more complex processors in late 2011 with its acquisition of NetLogic. This has allowed Broadcom to design all of the individual components of the router line card (NPU, TCAM, multi-core processor, etc.). Broadcom’s CEO explained the benefits of this integrated supplier approach, saying: “We’re able to make the products work better together, so we can optimize our switches to work with the network processors…it’s our goal to design [products] as a platform to bring significant advantage to customers who purchase all of them together” (BRCM Q2 2012 Call). A single publicly announced customer loss to Broadcom Corporation (NASDAQ:BRCM) could spell disaster for EZCH shares.
  • Marvell Technologies, the Sole Supplier for EZCH’s NP-4 and NP-5, has Recently Become a Direct Competitor. In order to guarantee manufacturing capacity, Cisco appears to have demanded that EZCH use a larger company as a conduit between itself and EZCH’s semiconductor foundry, Taiwan Semiconductor Manufacturing. Marvell, as “the sole supplier of our NP-4 and NP-5” (EZCH’s 2011 20-F), serves this function by employing its buying power to move its orders to the front of the queue. But the arrangement has grown increasingly conflicted following Marvell’spurchase of Xelerated, a small NPU startup, in January 2012. Xelerated had launched a 50Gb/s full-duplex processor (100Gb/s) in August 2010 and was EZchip’s chief competitor at the time. We would expect Marvell to support the Xelerated team with much-needed capital to potentially push forward a next-generation chip to compete with EZCH’s NP-5.
  • EZchip’s NPS Chip Appears Fundamentally Handicapped Versus the Privately-Held Netronome. Investors who refer to the NPS chip as a ‘game-changer’ are severely underestimating Netronome, a privately-held Intel spinout run by an accomplished Caltech graduate. Netronome recently made its presence known by directly challenging EZCH at a Linley Tech Conference. A review of the conference illustrates the similarities between the two companies, stating that the NPS “will take EZchip beyond its switch/router roots and into a wider venue that looks more like, well, like Netronome’s business”. Netronome had previously introduced its 200Gb/s L2-L7 chip (called NFP-6xxx) back in June 2012. And notably, as only one of a handful of Intel fabrication customers (others include Archronix and Tabula), Netronome should hold a long-term technological advantage versus EZchip. Intel’s best-in-class foundry has allowed Netronome to create its NFP-6xxx at the 22nm scale, and future designs may be made on Intel’s 14nm scale. On the other hand, EZCH’s foundry provider Taiwan Semiconductor will fabricate the supposedly next-generation NP-5 at only a 28nm scale. A smaller manufacturing scale can increase energy efficiency and add to bandwidth capacity. As a private company, Netronome has been under-covered by equity analysts and we believe that EZCH investors are underestimating the risks imposed by this formidable competitor. The significance of the Intel foundry relationship may be confirmed as Cisco Systems, Inc. (NASDAQ:CSCO) is rumored to be near announcing a billion dollar foundry deal with Intel to produce its own silicon chips. If this were the case, it would prove that Cisco Systems, Inc.(NASDAQ:CSCO) clearly values Intel’s 22nm-scale technology.
  • Consensus EPS Estimates use Non-GAAP Measures that Inflate EPS (and Analyst Price Targets) by Roughly 35%. To boost their lackluster earnings figures, EZchip Semiconductor Ltd. (NASDAQ:EZCH) management guides investors using non-GAAP reporting figures that exclude stock-based compensation (“SBC”). While this tactic is frequently used by technology companies, disciplined investors and analysts shouldn’t be fooled. Since public shareholders of EZCH stock didn’t receive their stock for free, then how can stock be gifted to EZchip’s employees without expense? Of the research reports we’ve read, Felt & Company, Deutsche Bank, Jefferies, Brean Capital, and Oppenheimer all follow management’s non-GAAP EPS to build their price targets. EZchip gifted its employees about $11m in stock options over the last twelve months (EZCH 20F6K), translating to a decrease in EZCH’s non-GAAP EPS by about $0.39/share. This calculation has a very material impact on the price targets for a company which earns as little as EZchip does. For example, Felt & Co reaches a strong buy conclusion and a $35 target price by using a 22.5x P/E multiple on 2014 non-GAAP EPS of $1.55 (February 14th report). If we removed $0.39 of stock-based comp expense from Felt’s figures, their same valuation multiple yields a target price of only $26. Because of this widely perpetuated miscalculation, it is a mistake for the investor community to be guided by forward multiples on Yahoo Finance or other sites that rely on these inflated non-GAAP EPS figures.
  • Given What’s Been Discussed Above, We Believe that Shares are Overvalued by as Much as 45%.  Based on the forthcoming competitive risks from Broadcom and Marvell, EZCH’s extensive track record of overstating growth potential, a relative lack of pricing power, and a weak backdrop in carrier spending, we believe that EZCH stock should now trade at a valuation multiple more in line with the broader semiconductor market. Peers demonstrating year-over-year revenue growth, unlike EZCH, generally trade between 10 and 15x 2013 P/E. After deducting our estimated SBC of $0.39/share from the Street’s non-GAAP EPS estimate of $1.13, we valued EZCH based on our adjusted 2013 EPS estimate of $0.74. With the addition of 25% of EZCH’s cash balance, a generous 2013 P/E multiple range of 15 – 20x leads to a valuation range of $12.50 – $16, or a 32% – 48% discount to the current trading price. We also don’t believe that management’s 3x revenue target by 2016 is realistic, and instead have used Gartner industry projections and our own assumptions to reach a ‘bull-case’ 2016 revenue target of $85m. At a 35% – 40% EBITDA margin and a 6x – 8x multiple, this translates into a price target of $14 – $17, still a 29% – 43% discount to the current price.

EZCH Report March 2013 Kerrisdale Capital by