• EZchip has a near monopoly on the merchant market for high-speed ethernet network processors (NPUs). These NPUs are integrated into proprietary systems by OEMs like Cisco and sold to carriers for use mainly in edge routers. As total data traffic across carrier networks grows (up 67% annually through 2016) so does the need for more powerful edge routers to handle all that data efficiently. Although carrier spending can be lumpy in any given quarter, demand for edge routers (and the NPU driven line cards that power them) will be strong as long as mobile data traffic continues to soar. EZchip now has the #1 position in the NPU market with a 32% share, up from only 11% in 2008. More specifically, EZchip has 70% share in the coveted high-speed NPU market.

• In 2016, EZchip brings to market a revolutionary new product line, NPS, which moves the company beyond the edge router end market, to also include switches, data center and network appliances, and software defined network (SDN) functions. This is augmented by the recent acquisition of Tilera that further expands the addressable market to include multi-core CPUs. In total, the new markets addressed by the NPS and Tilera will expand EZchip’s addressable market six-fold.

• Based strictly on the core business of selling NPUs into the fast-growing edge router market, EZCH is valued at 14x 2014 earnings ex-cash. This doesn’t account for the new end markets addressed by the NPS, or the value of Tilera, which greatly expand the addressable market. At $24, we’re buying the core business (growing at 20%) at a reasonable valuation, and getting two new product lines for free.

• Customers like Cisco, ZTE, and Huawei, use EZchip NPUs at the heart of their edge router designs, wrapping proprietary software around the NPU to differentiate their offerings. EZchip introduces a new generation of NPU every couple years and customer retention has been high. This stickiness is driven by EZchip’s leading performance metrics, but also by the large fixed programming costs that go into the initial integration of EZchip NPUs into the customers’ proprietary system. EZchip’s NPUs are backward compatible so customers have a strong economic incentive to continue using the EZchip platform through multiple product cycles.

• Each generation of EZchip NPU has an exceptionally long tail of ongoing revenue driven by fill-in sales of line cards. Carriers typically buy edge routers with numerous open line card slots. This minimizes up-front expense, and allows them to add more line cards (and throughput capacity) when demand builds in the future. Historically, line card sales will grow for 2-3 years after the initial purchase, as the carrier adds capacity to match demand. The current generation, the NP-4, will contribute about 60% of 2014 revenue, with the balance coming from fill-in sales of earlier processors (NP-2, NP-3) and some early sampling of NP-5. We can expect another 1-2 years of strong NP-4 sales before it begins for gradually taper off. Last year, Juniper was still a 19% customer despite announcing in 2009 that it was no longer using EZchip in new designs. These fill-in sales of the NP-2 processor illustrate the long tail of revenue derived from every design win.

• The two drivers of EZCH’s NPU sales are growth in the high-speed edge router market (growing 6%-7% annually through 2016) and market share gains versus in-house NPUs (moving from 30% to 45%). Market share gains are coming from a much stronger buy-in from Huawei and ZTE for the NP-4 (current NPU) and NP-5 (first revenue coming in 2015), than was seen with the NP-3. All NP-4 customers have signed on for the NP-5, and most have already committed to the NPS line (2016 release), which takes the place of the NP-6. The combination of customer retention through successive product generations, and the long-tail of fill-in sales gives us a reasonably good estimate of sales growth over the 4-5 years. With increases in market share, higher ASPs, and a growing overall market, we expect 20% growth in NPU sales for the edge router end market through 2017. For this growth, we’re paying 14x forward earnings; not an unreasonable multiple for a fast growing, predictable earnings stream. Not included in this earning forecast in any contribution from new data center markets addressed by the NPS and Tilera. These will likely be significant contributors to earnings in 2-3 years, but also diversify the revenue and customer mix.

• NPS - The NPS line (commercial production in early 1H’16) dramatically expands the market for EZchip’s processors to include data center appliances. While the NPS offers edge routing customers a substantial performance upgrade over the NP-5, it also offers level 2-7 processing and is C- programmable. This expands the addressable market to include switches and other network and data center appliances. It also opens the door to white-box customers building appliances for software-defined networks. EZCH critics have long pointed to the threat of SDN displacing proprietary networking firms like Cisco (and suppliers like EZCH) with networks built on commoditized network appliances (Google, Facebook, Amazon). These white box appliances still need powerful, flexible NPUs to handle the ever-increasing data flow across networks. With the NPS, EZCH hedges its bet with a product that continues its long line of processors for proprietary edge router designs, while also offering the performance and programability to be an ideal fit for white boxes.  

• Tilera - With the pending acquisition of Tilera, EZchip gains another foothold in the data center market with a suite of multi-core CPU processors. The deal doubles the size of EZchip’s end markets and bring much needed customer diversity. EZchip will now compete directly with data center players like Cavium in this fast growing space. The deal is likely to close in September and be accretive to earnings in 2015. EZchip paid $50M cash upfront, with a series of earn-outs that could bring the total price to $130M. Tilera is expected to contribute $35M+ in 2015 revenue, with anything over $35M being cash flow positive. The real Tilera contribution will come in 1-2 years when we see the first products combining the strengths of the two companies. For the time being, it doesn’t look like EZchip overpaid (with modest growth the multiple should be 8x-11x, depending on the earn-out tranches) and it doubles the company’s addressable market. We’re willing to give management the benefit of the doubt as they have a strong track record of developing long-term product roadmaps that anticipate changes in the routing and data center markets.  

Balance sheet and cash flow strength

• EZchip has $169M in cash (net of the Tilera acquisition) and no debt. This cash pile has built up due to a unique combination of strong free cash flow, a non-acquisitive management team, and an Israeli tax law that prohibits dividends in exchange for a 0% tax rate. Israeli law locks in the 0% rate for EZchip through 2021 provided the earned income is not distributed. The other factor that increases FCF is the large percentage of share-based compensation (RSUs), running at about 20% of revenue. Like most tech companies, EZchip reports both GAAP and non-GAAP earnings, with stock-comp being the main difference. We use non-GAAP earnings (because they closely track free cash flow), but account for a 3.5% increase in share count annually when looking at forward EPS. The net effect of this policy is substantial cash building up on the balance sheet. Unfortunately, most of this cash is unavailable for dividends, or a buyback, unless management struck a deal with the tax authorities (see CHKP and TEVA) to free up the cash in exchange for a one-time payment. For simplicity, we’ll deduct 25% (Israeli corporate tax rate) from the cash position ($169M cash ex-Tilera x 75% = $127M, or $4.30 per share) when determining enterprise value.

Management

• CEO Eli Fruchter co-founded EZchip in 1990 and owns 2.2% of the company. The remainder of the management team owns an additional 1.1% of the shares. Always careful with his words, and never offering detailed guidance, Fruchter has never been popular with analysts. His refusal to promote the stock, or even to clarify mis-conceptions about the business model, has likely held back the share price through the years. Nonetheless, his track record on product development is excellent, with each generation of NP processor gaining new customers and market share. Early sampling and design wins suggest that the NPS will also be a success across the router and data center markets.  

Cisco and Marvell

• EZchip has a unique sales channel to Cisco (~40% customer) that understates revenue and overstates gross margin. All sales to Cisco first go through Marvell, which adds certain features, and pays EZchip a royalty for the sale of each processor. These royalties come through to EZchip at 100% gross margin, compared to 65%-70% gross margins for other customers. As sales to Cisco have grown from <10% in 2009, to 40% this year, gross margins have increased, but the total sales growth has been muted compared to what it would be if Cisco were a direct customer. 2014 revenue would be roughly 17% higher with Cisco as a direct customer. Obviously, this doesn’t change the operating margins of the business, but it does shine light on the top-line growth that has been masked by the royalty arrangement.

Margins and Valuation

• Non-GAAP net margins are around 45% - 50%, but should come down some with the addition of Tilera. Free cash flow closely tracks non-GAAP net income and cap-ex requirements of the business are minimal.

• Shares trade at 14x our $0.87 estimate of 2014 non-GAAP EPS, net of the adjusted cash position of $4.30/share. As noted earlier, we think this is a reasonable multiple for a business with a highly visible 20% growth rate for the next 4-5 years. Our 2014 estimate doesn’t include any contribution from NPS (2016 introduction) and no earnings from Tilera. A year from now the market will be discounting 2015 earnings, which we project at $1.71. By next summer, we’ll also have a number of NPS design wins to back up the one already announced in the data center market. Getting traction in these new markets should warrant a multiple expansion as EZchip’s top line growth will look more like CAVM (25%+) and less like BRCM (15%) and MRVL (10%). At 18x forward earnings (CAVM trades at 35x) plus cash, the shares should trade at $35, or 45% higher than today.

Variant View

• The anchor holding back EZchip shares is the short thesis that SDN will kill the margins on proprietary networking equipment. Cisco has been the lightning rod for this short thesis, and EZchip has been a derivative of the short Cisco trade. We think this is misguided for several reasons. First, high-speed edge router sales have been on a tear the last two years, especially at Cisco. The ASR9000 has been seen solid double digit growth amongst weak overall router sales. This promises years of line card infills following the initial chassis sale. Second, NPS was designed to support SDN and should allow EZchip to grow sales in the data center, where the company has minimal presence today. If sales are strong in edge routing (which supports the current valuation), and white boxes give the company inroads into a massive new market, why does SDN represent a threat to EZchip? Yes, the company has to prove NPS can take share in the data center, but at this share price its a free call option on that new business.  

• Strong earnings growth in the core business over the next 2-3 years based on predictable growth in line card sales to major OEMs.
• More design wins in the data center market coming late '14/early '15 for the NPS. Grows the overall market, diversifies the customer base, and accerlerates growth.
• Tilera becomes accretive to earnings in early '15 and redefines EZchip as a player in the data center multi core CPU market.