EZchip: Customer Concentration And Uneven Revenue Stream Have Forced Risky Shift To The Data Center

| About: Mellanox Technologies, (MLNX)

Earlier this week, we published our first article on EZchip, stating our conviction that EZCH shares are grossly overvalued. Among other things, our initial article discussed the threat of displacement from Broadcom (BRCM) and Marvell (NASDAQ:MRVL) NPUs, customer defections, unrealistic growth expectations, and research analysts' negligent exclusion of stock-based compensation from their EPS and price target calculations.

This article will build on those themes to further make our case that EZCH's current valuation is unjustified. To begin with, EZCH's severe customer concentration makes it susceptible to a sharp correction. A single Cisco (NASDAQ:CSCO) design win by Broadcom would cause an irreversible loss of capital overnight. Next, we believe EZCH's seemingly premature announcement of its research-phase data center product, the "NPS" (network processor for smart networks), was a diversion tactic to shift attention away from the increasingly challenging NPU market. While this measure was successful in temporarily buoying the stock after a disappointing Q2 2012, we believe the NPS's potential is grossly exaggerated. The data center market is much more crowded than network processors, and highly-regarded companies like Netronome and Cavium (NASDAQ:CAVM) already possess best-in-class chips, not to mention larger incumbents like Freescale (NYSE:FSL), Broadcom, LSI (NYSE:LSI), PMC-Sierra (NASDAQ:PMCS), and even Intel (NASDAQ:INTC). And even if EZCH can differentiate itself amongst this deluge of competitors, EZCH has stated that it doesn't expect revenue until 2016, signifying that any value attribution should be heavily discounted for time and risk.

Compounding these concerns is the irrefutable fact that EZchip has a long history of over-promising and under-delivering. While the most speculative of investors will always believe that this time is different, our experience tells us that old habits are hard to break. We continue to believe that EZCH's valuation of 7.4x 2013E revenue, 17.0x 2013E EBITDA, and 32.1x 2013E GAAP P/E is about twice where fair value should be.

EZCH's History of Disappointing Results Place a Highly Volatile Revenue Stream at Risk

Bullish research reports and investor write-ups on EZCH typically argue that EZCH's premium valuation is warranted by a steady revenue stream from past and future design wins. Because EZCH's customers, such as Cisco and ZTE (OTCPK:ZTCOY), generally stick with a single NPU manufacturer for each router model, once a design is won it tends to be sticky for the lifetime of that product cycle. But empirical evidence suggests that EZCH's quarterly revenue is actually highly volatile, quite the opposite of a steady income stream. This is the result of EZCH's over-reliance on telecom carrier expenditures, which is an extraordinarily difficult market to forecast.

Sources: EZCH 6-K and 20-F filings; earnings calls

With such unpredictable end-market demand, it would be hard for any company to have good visibility on near-term results. But this hasn't stopped EZCH from bullishly proselytizing to its investors. As the inconsistency between the quotes below and EZCH's underlying financial results shows, EZCH's growth promises have proved misleading time and time again.

"We continue to believe that ZTE is still early in its ramp up and continue to be excited by the growth potential that ZTE represents with the NP-3 and the NP-4" - Q4 2010 Earnings Call

"But we feel comfortable that NP-4 will grow very significantly starting in the second half of [2011] and next year [2012]." - Q1 2011 Earnings Call

"So, basically, we believe that other than Juniper, we expect all our other customers to grow on a year-over-year basis." - Q2 2011 Earnings Call

"We expect Cisco to grow significantly for us in 2012." - Q4 2011 Earnings Call

"We expect our main growth drivers for 2012 will be the ramp up of our new NP-4 customers and the recovery in… carriers' capital expenditure. We believe this will result in year-over-year growth in 2012." - Q1 2012 Earnings Call

Given management's track record of over-promising and under-delivering, we believe that investors should be highly skeptical of EZCH's commentary going forward.

EZCH's Customer Concentration makes it Susceptible to a Sudden Correction

In the technology sector, the risk of overnight obsolescence is constantly present. This risk becomes compounded when a business is dependent on a single customer for the vast majority of its revenue. EZchip suffers from both risks. Not only does Broadcom's BCM 88030 have the potential to steal NPU market share, but EZCH's over-dependence on Cisco makes its already volatile revenue stream that much more precarious. Should Cisco decide to move to in-house designs, EZCH's primary source of growth will vanish.

Sources: EZCH 6-K and 20-F filings; earnings calls

Without a doubt, Cisco has been the savior of EZCH's stock price over the past year. The router vendor has represented between 37% to 49% of EZCH's revenue in each of the past four quarters and an even greater percentage of gross profits, given the royalty arrangement. While investors have largely overlooked the competitive shifts brooding in the NPU market, this would change if EZCH lost a major customer. Should Cisco, or any of EZchip's other major customers, either award an NPU contract to Broadcom or switch future chips to in-house designs in 2013, EZCH shares could experience a fate similar to that of many other overly concentrated hardware suppliers.

A recent example in the technology sector is Audience (NASDAQ:ADNC), a maker of audio processors for smartphones. Like EZchip, Audience was heavily reliant on a single customer, in this case Apple (NASDAQ:AAPL), which represented 55% of Audience's Q2 2012 revenue. After it was announced that Apple would use in-house audio processors in the iPhone 5, Audience shares experienced a one-day fall of over 60%, illustrating the risks involved in heavily concentrated technology suppliers.

Source: CapitalIQ

As if EZchip's dangerous customer concentration and a history of underwhelming results weren't enough, recent data from Gartner indicate that the carrier end-market isn't the robust growth market that many believe.

December Gartner Report Lowered Expectations for the Carrier Capex Market

Gartner, the well-regarded research institution, released a Q4 2012 update to its Carrier Network forecasts that portends a bleak future for the carrier infrastructure market. Gartner now predicts that network infrastructure spending will decline 6.6% in 2012 compared to 2011 due to "weaker demand and aggressive price competition." Making matters worse for EZCH, global telecom players such as AT&T (NYSE:T), Verizon (NYSE:VZ), China Telecom (NYSE:CHA), and Vodafone (NASDAQ:VOD) are increasingly focused on mobile networks and backhaul as opposed to wireline installations. While EZCH does sell into the wireless market, a substantial portion of its business is driven by core wireline routing. Recent spending patterns from AT&T and Verizon should be cause for concern. Wireline capex as a percentage of total spending has collapsed from around 60% in 2009 to around 40% today. EZCH attributes much of its recent weakness to this trend: "There is short-term problem… it depends on… the carrier CapEx and service providers spending especially wireline… and this is the main reason for the decline in expected revenues" (EZCH Q2 2012 Call).

Source: November 9th, 2012 Capstone Investments Research report.

Poor Prospects in Carrier Ethernet have forced a Shift to the Data Center

The telecom infrastructure market is not nearly as exciting as it once was. Between these end-market headwinds and EZCH's inability to drive NPU growth, it's quickly apparent why investors are growing wary of the EZCH story. These fears first materialized on EZCH's Q2 2012 call when management announced Q3 revenue guidance of $8m-10m, a ~40% sequential decline from the previous quarter. Spooked investors sent EZCH shares down 25% after the call. EZCH must have felt compelled to reassure these jittery investors as it rushed to pre-announce plans for a research-stage product just three weeks after the Q2 call. On August 28th, EZchip pre-announced that it was going to unveil a new product line on September 5th. By the time September 5th finally arrived, EZCH shares had already recovered more than half of their original loss. EZchip even hosted an analyst call to announce its new project, dubbed the "NPS" (Network Processor for Smart Networks). We believe that this early announcement was a tacit admission that EZchip foresees continued market weakness and mounting competitive pressures in Carrier Ethernet.

EZchip "does not have customer commitments" for the NPS but said that it expects to "see design wins for the NPS in the first half of next year [2013]" while conceding that "it's too early to talk about target numbers" (EZCH Sept. 5 NPS Call). This project has caused a ramp in R&D spending, up from $4.1m in Q3 2011 to $5.2m in Q3 2012 and Q4 2012. Since EZCH has said that the NPS won't meaningfully contribute to revenue until 2016, this growing R&D expense on an unproven technology should be a headwind on earnings for years to come.

EZchip's NPS Technology Appears Fundamentally Disadvantaged

Whereas EZchip's series of NP-processors are optimized for carrier networks, the NPS will be designed for the data centers. EZchip claims that the advanced 'L2-L7' architecture adds $400m to EZCH's total serviceable market, doubling its overall market opportunity by 2016. Because this opportunity stretches so far into the future and includes a myriad of unknowns, we would warn investors against even considering adding it into EZCH's valuation until definitive customer agreements are signed.

The most direct potential competitor comes in the form of an underfollowed, privately-owned business named Netronome. Netronome is a spin-out from Intel that focuses solely on high-end network processors. Netronome has already designed a 200Gb/s chip (NFP-6xxx) for the data center, putting it well ahead of EZCH's NPS. The Linley Group also believes that the flexibility of the chip makes it a competitor to EZchip's NP-5 for the next-generation network processing market. The Intel spin-out is run by Howard Bubb, a Caltech-educated electrical engineer who previously held senior positions at Intel, and the venture is funded by blue-chip investors such as Tudor, Top Technology Ventures, and DFJ Esprit. Within the next year or so, we think that Netronome could announce a next-generation chip that matches the theoretical bandwidth of EZCH's NPS chip, putting pressure on EZCH's share price. Beside the potential trouble for EZCH's NPS chip, Netronome also aspires to move into EZchip's core L2-L3 Carrier Ethernet market. The Chief Router Architect at ZTE, a major EZCH customer, recently praised Netronome's NFP chip:

"Netronome has once again delivered on its strengths; flow-aware processing and large scale integration of programmable resources. The flexible, ultra high speed interfaces, processing cores and network specific hardware accelerators provided by NFP-6xxx family of products allow it to be positioned in both the line cards as well as service cards of next generation routers. NFP-6xxx along with prepackaged networking software closely aligns with our vision for our router family of products" - Ye Zhining, Chief Architect Router Products, ZTE.

Aside from these glowing reviews, we believe that Netronome holds an important manufacturing edge over EZchip. Netronome's NFP-6xxx chips are built by Intel using a 22-nanometer manufacturing scale. EZchip, on the other hand, is reliant on a TSM foundry that is consistently a generation behind Intel. On the September 5th NPS call, EZCH said it plans to "use 28 nanometer for [the NPS]" (EZCH Sept 5th NPS Call). EZCH's NP-4 currently engages TSM's 55nm scale while the NP-5 will also employ TSM's 28nm manufacturing process. Smaller manufacturing architecture can translate directly into better speed, power use, and efficiency metrics. But by the time of EZchip's potential NPS release, Intel's fabrication facilities will have already moved to 14nm and beyond.

Source: Intel 2012 Investor Meeting, page 37

Beyond the direct threat from Netronome, the data center market is already occupied by heavy hitters such as Cavium (see Cavium Octeon III Sizzles at 100Gbps), Broadcom (NetLogic), Freescale, PMC-Sierra, LSI, and Intel. These well-capitalized competitors will most certainly fight to protect their share of the L4-L7 chip market and will shift resources into higher-speed networking chips should market demand warrant.


While EZchip would like investors to believe that the NPS will be a savior for the business, we believe that the market should pay little attention to this research-stage product. Until EZCH proves it can differentiate itself from a medley of competition, most notably the venture capital-backed Netronome, speculation on the addressable market size holds little weight. Over the near-term, investors should remain wary of EZchip's reliance on a single customer, Cisco, for more than 40% of its revenue. This customer concentration will only grow as Juniper's revenue contribution tapers to zero over the next few years. With the probability of a substantial Huawei contract now remote, the growth headwinds appear even more daunting. And since we know that Alcatel-Lucent uses in-house NPUs, each of the top four Carrier Ethernet vendors by market share have been accounted for. EZCH must now target second and third-tier vendors like Ericsson and Tellabs, signaling that future growth will be hard to come by. Due to these mounting growth concerns, looming competitive threats, and a valuation of 32.1x 2013 GAAP P/E, we aren't sure why anybody would buy EZCH shares at this price. We maintain our conviction that the many downside risks far outweigh the upside potential.

Disclosure: I am short EZCH. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.

Additional disclosure: Please read our disclaimer at kerrisdalecap.com/?page_id=2118.