I’ve been engaged in constant debate with readers and other investors about my position in Vitesse Semiconductor since disclosing it. Based on reader email, it was one of my more unpopular opinions. I thought it would be appropriate to share my investment thesis from early August with a wider audience.

This post is not a recomendation to buy or sell Vitesse. It is the reasoning behind why I personally purchased it. Please draw your own conclusions.
—————————————-

Vitesse Semiconductor Investment Thesis

The breakup value of Vitesse Semiconductor (VTSS) is $2.50/share, net $170M debt and $20M cash this yields a valuation of $1.75/share. This is based on valuation of each of the three Vitesse business units.

There are three future outcomes for Vitesse:

1. Breakup and Sale of all business units. This would likely be catalyzed by an initial private party takeover by Tennenbaum/Chapman/others.
2. Sale of one business unit to generate operating cash and normalize cash burn.
3. Continued operation and turnaround of existing business structure

Scenario One – Breakup and Sale of all Business units

• Telecom – The annuity with a growth component. Generates $100M a year with approximately 65% margins. This revenue should be considered an annuity and would generate cash for several years in the absence of any incremental R&D and SG&A expenses. Products include competitive SONET/SDH and electro-optical components and large legacy revenue streams. Yesterday, AMCC acquired Quake for functionally identical technology at a cost of $69M, a 5x premium over annual revenue. Intel is seeking a sale of it’s Telecom unit at prices much greater than 2x revenue and has inferior products and customer penetration when compared to Vitesse. A valuation of 2x revenue for the Vitesse telecom unit is justifiable. Valuation $200M.
• Storage – The most attractive unit for prospective M&A. Generates $60M a year with 50% margins. PMC-Sierra paid 3.8x revenue for the storage semiconductor unit  of Avago in Fall ’05. Vitesse has products poised for a more significant product ramp than the ones acquired from Avago. An identical valuation of the Vitesse unit is conservative. Valuation $225M.
• Ethernet – The most difficult unit to value. Generates $40M (20% of total) a year in lower margin revenue but is estimated to consume 30% of R&D. Products include Layer 2/3 switch silicon and Gigabit Ethernet PHYs. PHYs are product of Cicada acquisition in January 2004 for $66M in cash and are a relatively unique product. Layer 2/3 switches are comparable from a technology perspective with Marvell (MRVL), Broadcom (BRCM) and Agere (AGR) devices. Likely outcome is acquisition of unit by large semiconductor unit looking to diversify into Ethernet – potential suitors include AMCC (AMCC), Freescale (FSL), TI (TXN), PMCS (PMCS), Atheros (ATHR), or Intel (INTC). Second likely outcome is an incumbent looking to remove competitive pricing pressure, most likely Broadcom. Removal of Vitesse as a competitor could improve Broadcom pricing by $0.25 per port resulting in $8-10M in additional product margins per year Conservative Valuation $80M, could be 25-75% higher.

Total break up value – $505M-$565M, or $1.75-$2.00/share

Scenario Two – Sale of One Business Unit

The Ethernet division is most likely to be sold to generate cash. Further operation and growth of this unit requires substantial cash investment that Vitesse cannot afford (current credit terms are LIBOR +9-11%).

Results from such a transaction at $110M valuation:

• Quarterly revenue goes from $50M to $40M
• COGS goes from $22M to $15M
• Margins go from 56% to 62% (before positive effects of better channel behavior kick in)
• Operating Expenses go from $40M to $27M.
• Tennenbaum note would be repaid ($52M + $20M make whole penalty). $98M in convert debt (1.5%) remains.
• Operating cash balance of around $50M.

This leaves a near break-even company with substantial invested R&D and future prospects in the storage and optical markets, $160M in annual revenue with 65% ($100M) gross margins. Peer company valuations (AMCC) (PMCS) are 3x-4x EV/Gross Margins, which would value the new Vitesse at $1.25 to $1.75/share, which should considered low given future prospects.

Scenario Three – Continued Operation

It is possible, but not likely, that near term growth in the Storage and Ethernet markets leads the company to cash flow break even. The threat of customer migration away from Vitesse has been overstated by conventional sell-side analysts, and analyst cash accounting metrics trend toward pessimistic. Better management of distribution inventory will result in near term margin expansion (1-2%).  Regardless, in the absence of near term operational improvements the company has options to resolve cash flow issues through the sale of business units.

Summary

Removal of duplicative R&D and SG&A in the specialty component space through M&A is an unrecognized method of generating shareholder value. Vitesse is either an excellent target for consolidation, or an undervalued head and torso that can be fitted out with the appendages of other companies, such as Intel’s Telecom business unit. The current Vitesse valuation leave investors with a lot of room for error.

At the time of posting, of companies mentioned, I am long VTSS and INTC.

This is not investment advice and reflects my personal opinions. Read the disclaimer.

All photos and component data (click to enlarge) are courtesy of a Japanese web page. I used Google to translate the page – available here. (I don’t think human translators have much to worry about yet). Zoro, whoever you are, thank you.

Zoro was very excited about his GE-PON install. (Translation courtesy of Google)

It is I who use ADSL test service with Mita, but moving, in Musashi boundary, and Yono. Directly, it was the フレッツ ADSL, but with Yono public corporation was converted to DION light and it is the ざ to be, however light it cannot call the VDSL after all, -, with thinking, the everyday life which soaks the pillow with the saliva came continuing… that me to yesterday. It succeeded in coming home finally pulling the optical circuit to the home in Musashi boundary. From today the joke you probably will walk cleanly with feeling!

That’s exactly how I felt when I had Verizon FiOS installed! I too can now “walk cleanly with feeling” and no longer have a wet pillow in the morning.

It appears that Zoro has a complex install at his home, including GE-PON ONU, SOHO router, and some kind of Windows Media Center Extender (MCE). The MCE appears to connect over the PON connection to … something. Not sure if this is a set-top box talking to an IPTV provider or if it is a homebrew network. Regardless, sounds like it isn’t working too well. Anyone who can translate the original and shed some light on this mystery, please do so.

He also has the full gratuitous shot of the Mitsubishi AS-1000GN GE-PON ONU and all of the components inside.

Mmmmmm. Nice.

Try clicking on the photo – I’ve annotated all of the components using Flickr using information provided by Zoro.

We’ve got:

• Passave (PSVE) PAS6001-B as the IEEE 802.3ah MAC.
• Freescale (FSL) Powerquicc MPC852TVR50 … probably doing control, not in datapath.
• Assorted memories… Flash, SRAM.
• Altera (ALTR) FPGAs, sounds like one sits in the datapath. It’s hooked up to two IDT (IDT) 71V546S133PF ZBT RAMs. Looks like the Altera chip is doing some sort of protocol processing that the Passave device couldn’t do.
• Vitesse (VTSS) VSC8201RX Gigabit Ethernet Copper PHY – according to Zoro it only runs 10/100, a limitation of the Altera glue FPGA.
• Texas Instruments (TXN) TNETE2201B as the TBI to serial 1.25G SERDES that connects the Passave MAC to the GE-PON optics (not shown here).

Unfortunately, no shot of the optics or the diplexer. It looks like an SFF footprint is being used based on the solder patterns shown in the photo. An astute reader I shared this with in advance pointed out the lack of SFF alignment holes- which means it is homebrew from discretes, and not an off-the-shelf module. It likely has many Melco components.

According to Passave the PAS6001 has been discontinued in favor of the PAS6201. Looks like this new device has FEC and some classification integrated on board…. likely eliminating some or all of the FPGAs. It also looks like it will support clients at Gigabit Ethernet. The TI and Vitesse chips would stay for now though integration of those is inevitable.

BOM Cost Analysis:

• Passave MAC – $25
• Vitesse Copper PHY – $2
• TI SERDES – $2
• Freescale CPU – $6
• FPGAs – Need help here. Altera EP1C20F400C6 $98 and EP1C12F256C6 $56 on website. This is high and nowhere near the 100K+ pricing Melco would pay. I’m going to ball park $40 for both devices
• Optics and Diplexer- $30
• IDT Memory – 2 x $2 = $4
• Flash Memories – $5
• Assorted power and discretes – $10
• Build and Case – $20
• Market research for free – priceless

Total Cost: $144

NTT must have paid about $250 for this ONU. The FPGA hack is very expensive, and was probably a big barrier to entry for ONU suppliers. We’ve learned of several custom features NTT requested not in the ITU spec – that’s probably how they were implemented.

The next generation Passave device should eliminate the Powerquicc, FPGAs, and the IDT memories used by the datapath FPGA. Assuming the MAC pricing stays the same the total BOM drops to $94, not including potential cost reductions in optics. We’re hearing SFF GE-PON optics are going for around $20 from aggressive suppliers trying to capture this business.

Passave device ASP based on their S-1 filings and NTT install rates. This was the price in the first 1/2 of 2005, it is likely lower by now.