We have not written about Vitesse since last August, when we changed our opinion about the company after the announcement of the storage divestiture to Maxim (see “Vitesse Sells Storage“). The company’s valuation has dropped nearly 50% since then and merits closer examination.

Q208 Consumption was $55.5M, a figure that represents the dollar amount of product end-customers took delivery of. They expect to begin reporting a traditional revenue number once audited results are completed, though they expect to continue recognizing customer revenue on a sell-through basis, a move we applaud.

Revenue (Consumption) Breakout

* Nyquist estimates

Historical Revenue (Consumption) Reporting

* Nyquist estimates

The company has completed a remarkable transformation away from using distribution channels for revenue buffering and has sector leading DSO’s on receivables of 27 days- AMCC’s is 36 and PMCS’s is 33.

Vitesse indicated it has stopped offering 2% Net 10 incentives for rapid payment which will consume working capital while raising gross margins 2%. Their distributor, Nu Horizons, can borrow money at near LIBOR rates, and surely took advantage of this profitable arbitrage. Since Vitesse now has sufficient working capital it is a sensible move to end this practice. We see most small/mid cap semiconductor companies moving in the same direction (AMCC and PMCS are doing this) and expect this to impact the P&L of their distribution partners. (See (NUHC) )

Margin guidance was in the range of 55-60%, lower than what we had expected in the past. From the Q&A on the call it appears the company is not yet operating in this range but expect to shortly. The halt of payment incentives may accelerate this.

The company has opex of 23.9M with 3.0M of audit/legal fees included and is operating at it’s stated long term opex target is 21M, which is 38% of current revenue; AMCC is at 59% of revenue, PMCS is 49%. Vitesse’s operational efficiency is notable but what is not yet known is what whether the company can grow or even sustain revenues with this level of investment. We believe that focusing the company on Ethernet only and applying it to the carrier and enterprise markets should give it maximum cost leverage, but we do not believe that Vitesse that much more productive than AMCC and PMCS.

Other notes:

• Networking revenues were up but by our estimates not at the levels they were 9-12 months ago.
• Ethernet over SONET/SDH was highlighted as a growth area with revenues expected to ramp over the next 3-5 years. Vitesse claims 70 wins though this includes companion chips like Ethernet PHY’s and MACs.
• Huawei and IBM were the two 10% customers. China was strong compared with previous quarter
• The company won a GE-PON FTTH design in the ONU, or customer located box with their 6-port Gigabit Ethernet switch. This is almost certainly a design at Huawei or ZTE, which are two of the three leading vendors for the Chinese GE-PON deployment. Chinese GE-PON ONU shipments will not break 250k this year but this is a significant validation of the company continued leadership in low power Ethernet switch/phy technology. The likely silicon vendor alongside Vitesse is PMC-Sierra or Teknovus. More discussion of this topic can be found in the discussion of PMCS Q108 results.
• Vitesse re-iterated their success in the X2 module based EDC (referred to as 1st generation EDC on the call). Our checks conclude that while Vitesse is not yet out of the running for the next-gen SFP+ E
  DC (ultimately, a higher volume opportunity) they are not leading. Cortina and AMCC occupy this position.
• The company emphasized the potential of using EDC technology in backplane applications – referred to as KR. We believe this is the true high volume opportunity. Unfortunately for Vitesse it appears that Broadcom and AMCC have already made significant inroads in this area.

Conclusion 

While the progress made by the company over the last 12 months is admirable, particularly in supply chain management, we believe the company faces risk of a decline in non-core revenues. The cash flow issues of the previous 12 months without a doubt forced the company to accelerate the usage of non-core assets for cash through the use of aggressive end-of-lifes (EOLs). We cannot argue with this decision as it provided vital operating cash. But little detail has been provided on this issue and whether EOLs were conducted on products outside the “non-core” segment. The significant risk of a secular decline in revenues that we identified last August is still present.

Author is long AMCC and short NUHC

Summary

PMC-Sierra’s results are no longer the proxy they used to be for the Networking industry. It’s future, like AMCC, is increasingly reliant on the storage and MIPS CPU portions of it’s business. Wireline revenue continues it’s secular decline and all indications are this will not reverse as the company has made few investments outside of FTTH.  We also believe that the positive Q2 guidance is derived from one-time events, primarily EOL revenue and initial shipments to China.

Q108 Revenue of $125.04M was roughly flat with the previous quarter and edged out analyst expectations of $123M. Q208 revenue was projected at $135-$140M, an 8-12% increase from Q108.

PMC reported that this strength was from a significant increase in their Japanese FTTH business as well as MIPS processors for printers. It was a down quarter for wireline (Telecom ex-FTTH) but PMC believes it has reached bottom. This is notable as wireline revenue has been in steady decline since mid-2006 and we believe the company is pursuing an aggressive end-of-life (EOL) program for these products. It isn’t clear to us what would cause the downward trend in wireline revenue to reverse. We believe that, at best, declines in wireline revenue have slowed, and PMC’s comments do not neccesarily mean a return to revenue growth.

Other notable items:

• Bought back $98M in convertible bonds (paid $95M, nearly par) leaving $127M outstanding.
• PMC expects storage will be their #1 business by EOY. This pushes their traditional wireline business into #2 position.
• Strong bookings appeared in late Q1, which we believe account for the strength they expect in Q2.
• Expects to ship several million WiMAX amplifiers in 2008 (Based on the state of WiMax deployments we lack the willing suspension of disbelief required to believe this will happen…)

We believe a significant portion of the Q2 revenue jump at PMC is attributable to one-time EOL revenue, and that this revenue has already shipped in April. The remainder of the growth is coming from initial shipments of FTTH chipsets in China.

FTTH and China

PMC reported increased FTTH shipments in Q1 and expected 2008 to be a big growth year for this product segment. We believe this is being driven by two factors:

1. Increased deployments of OLTs in NTT’s network

The number of FTTH subscribers surpassed the number of DSL subscribers in Japan for the first time this quarter. NTT appears to be exhausting the capacity of existing OLTs as this growth drives increases in subscriber density. Passave saw a large revenue burst early in the NTT deployment as the carrier salted central offices with FTTH capacity, and PMC-Sierra/Passave saw this revenue decline once NTT waited for the capacity to be absorbed. The pendulum is now swinging in the other direction and should remain there for 3-6 months.

2. Initial shipment of GE-PON devices to Huawei and ZTE in China.

Contrary to what G-PON vendors are saying, G-PON continues to lack significant traction in China. G-PON has seen near universal adoption by Western ILECs and their CLEC brethren but has only limited success in Asia. India remains the last battleground but the most exciting broadband growth there is wireless, not wired. GE-PON dominates deployments throughout Asia, including Taiwan and Korea. Point-to-point Ethernet is a greater technology threat to GE-PON than G-PON, as Korea and Singapore are taking this path.

Over 10 vendors are qualified for deployment in China Telecom/China Netcom networks but shipments to-date have been concentrated among Chinese vendors only, primarily Fiberhome.  PMC-Sierra secured wins at both Huawei and ZTE for Chinese FTTH deployments. PMC has not secured Fiberhome, which has shipped the majority of equipment to date. These two new vendors should account for increasing share in 2008 and 2009, and we see little chance of additional outside vendors taking share in China as this appears to be a nationalistic decision, given the dominant vendors of GE-PON equipment are Japanese.

The majority of Chinese FTTH revenue is for OLT shipments. This is not a sustainable trend for the same reasons it was not in Japan and shipments should attenuate as the year goes on. Ultimately, as we have written extensively, the China FTTH opportunity is massive but should not account for more than $5M in revenue in FY2008 for PMC-Sierra.

PMC remains strategically well positioned to supply silicon for the Chinese GE-PON FTTH deployment with Teknovus providing the only capable threat. Hard numbers are not available but PMC-Sierra is at best matching Teknovus revenue in China and has lagged in recent quarters. Cortina Systems acquired Immenstar, a start-up provider of silicon but has made few inroads in China. Ultimately the largest threat faced by PMC-Sierra and Teknovus are the internal ASIC capabilities of Huawei.

Our internal belief, unsupported by any hard data, is that Broadcom will eventually acquire Teknovus. If this were to take place it would pose a formidable threat to PMC-Sierra because of Broadcom’s deep IP portfolio, and it’s ability to integrate features such as MPEG decoding right into the ONU gateway chip. We don’t think most PMC-Sierra investors take this threat seriously enough.

Conclusion

An enormous valuation gap exists between PMC-Sierra and AMCC and we see few distinctions between the two businesses outside of FTTH. PMC-Sierra sells for an Enterprise Value/Gross Profit nearly 50% higher than AMCC and has a substantially worse balance sheet.

Author holds a position in AMCC and Broadcom.

… and very much in the running at Cisco.

Growth? Is That Really You?

The market for optical modules in Enterprise applications (primarily 1/2/4G SFP’s) suffered through years of secular anemic growth, beginning in 2002 when shipments of copper based GigE ports overtook shipments of fiber based GigE ports. The emergence of copper based GigE effectively capped unit volume growth in the optical business while exposing participants to declining prices as they fought for a share of a fixed pie.

The graph below illustrates the unit volume growth in 1GbE port shipments, and the missed opportunity due to the emergence of copper based GigE connectivity.  Not shown is SFP consumption in fibre channel end markets – about another 1M units/quarter – and not growing at a material rate.

You can understand why optical module vendors dream of an alternate history where the availability of copper based Ethernet was delayed.

However, we believe the secular decline in revenue from enterprise oriented optical modules is coming to an end. While copper GigE arrived at the beginning of the GigE growth cycle we do not think 10GbE will suffer the same fate. It is clear copper 10GBASE-T cannot deliver the densities required for volume applications while SFP+ can deliver 100m reach for less than $100 today. 10GBASE-T shouldn’t reach power density (which is increasingly the critical specification) parity with SFP+ for at least 3-4 years. 10GBASE-CX is an interesting technology and is worth watching but requires yet another cable type.

Therefore, it doesn’t appear that any near term growth in 10GbE will be addressed by copper interconnect. And cost/density requirements will dictate that the growth in unit volume comes from SFP+ – not legacy XENPAK/X2/XFP form factors. All that is needed is a catalyst to drive market demand for 10GbE.

The Emergence of 10G Datacenter Ethernet

Based on public information from Equinix (EQIX), Level3 (LVLT), Akamai (AKAM), IBM (IBM) and other sources it has become clear that the real driver of 10GbE unit volumes is not video but large capital expenditures in datacenters and datacenter interconnect. 10GbE is experiencing a ‘perfect storm’ of sorts as several trends converge:

1. Companies consolidating data centers into fewer locations for various reasons while simultaneously increasing their capacity.
2. The emergence of computing in the cloud (Google Apps, Salesforce.com, corporate applications, etc.) driving the demand for datacenter oriented computing. This in turn drives demand for next generation Bladeservers with multiple 10GbE uplinks. 
3. Carriers adopting 10GbE and L2 trunking and switching as the basis for next generation infrastructure and re-purposing enterprise components to implement it. (see “LightReading Ethernet Conference Notes“)

It remains to be seen if Cisco (CSCO) can extend its dominance of Ethernet switching into the datacenter space. As we discovered and made known before (see Cisco: The Optical Illusion), the 60% market share Cisco holds in the switching business results in an inefficient optical component supply chain market. Cisco has a monopsony on optical components and it extracts the majority of value rather than the suppliers themselves.

If Cisco market share slips significantly during the transition to 10GbE this would break the back of the Cisco monopsony and usher in a much healthier operating environment for optical vendors overall. While such an outcome is far from certain it begs watching given the market impact it would have.

SFP+ Rising

10G SFP+ modules have very quietly started to ship in volume into the 10GbE and 8G FC market. The buyers are storage vendors and private datacenter operators, including Google (see “Google’s Secret 10GbE Switch“). The current volume run rate is approximately 100k units a year.

Other customers with large datacenter requirements may make the same decision as Google and move forward with the SFP+ form factor without support for 10GBASE-LRM and other long reach options. Datacenters are increasingly wiring with OM-3 fiber, which eliminates the benefits EDC based LRM brings to the table.

EDC certainly isn’t going away, but it is no longer the key market enabler. Netlogic (NETL) purchased Aeluros, a maker of 10GbE PHYs who won dominant market share with their 10GbE PHY but failed to make headway with EDC based 10GbE. We expected the Aeluros market share to be undercut by new 10GbE PHY vendors like Cortina, Vitesse (VTSS.pk), and AMCC (AMCC) who offered integrated EDC. It appears this conclusion was wrong, as the importance of EDC is waning. Rather than Netlogic mirroring the error AMCC made when it purchased Quake (see “Quake: Another Failed AMCC Acquisition“) it appears their investment may remain intact. However, Cisco is still taking a different approach and is making EDC a requirement.

O Cisco, Where Art Though?

Using publicly released data from Cisco we estimate it will purchase at most 150k units of SFP+ in 2008, substantially less than half of total end market demand.

Why? Cisco remains 6-9 months away from shipping any equipment with SFP+ enabled hardware. Cisco has placed great emphasis on EDC and will not ship SFP+ until vendors deliver reliable silicon to interface with SFP+ optical modules.

As a result, Cisco will be buying more of the same 10GbE XENPAK/X2 modules for a period longer than the market expects. This is positive for incumbent suppliers such as Intel (INTC) and Opnext (OPXT), and negative for suppliers attempting to take share at Cicso such as Finisar (FNSR). Finisar lost the opportunity to supply the only new 10G XENPAK module (10GBASE-LRM) when a last minute switch from a Japanese laser supplier caused problems with their design. It isn’t 100% clear to us who won this business (we believe Sumitomo and Avago) but we also feel it doesn’t reflect meaningful volume.

Conclusion

• The 10GbE optical component market is hitting a growth knee as multiple trends converge to drive 10GbE port deployment. This is driving SFP+ module growth.
• Best of all, it isn’t Cisco buying the majority of product, which should translate into better gross margins for component suppliers.
• While “This time is different” are the four most expensive words on the planet it does not appear copper interconnect substitution is a near term threat as it was during the transition to 1GbE.
• Legacy form factors will be the dominant shipments into Cisco in 2008, but given the brutal pricing demanded by Cisco it isn’t clear that this business is really desirable.

Our impression is the investment community is overly focused on Finisar’s ability to take share at Cisco with a XENPAK product. Investors have also overlooked the disruptive nature of the AZNA acquisition and the low-cost 10GbE Telecom products it enables.

Our opinion is investors should be focusing on who will supply the SFP+ module market, a market which will be ultimately higher volume and less characterized by the corrosive effects of Cisco’s mo
nopsony. And it is here that Finisar and Avago have large structural advantages over the competition.

Author holds positions in AMCC, Vitesse, Finisar, and Opnext.

I strongly caution readers than some of the below information is single-sourced information that may be ripe with bias. I choose to publish what I see fit based on my opinion of what is likely to be true.

• Ethernet over SONET/SDH is ramping. According to an undisclosed vendor, Vitesse (VTSS.PK) is shipping in excess of 5k units a quarter of their 10G EoS chip to Lucent as part of the DMX2000 system. This would be around $1M a quarter. Transwitch (TXCC) was running at about $3M a quarter until a recent bump in the road, most likely short term.
• 10G density is now the  focus for all Telecom transport chipset vendors
  • No one I have spoken to is allocating resources to develop any chips for the carrier network targeted at 2.5G speeds or below – except for FTTH – and even that is moving to 10G.
  • Anything designed for 10G supports 10GbE and OC-192 interfaces interchangeably
  • AMCC (AMCC) has a new version of their 10G FEC/SERDES chip that is doing well in the marketplace against Cortina Systems.
  • T/PACK announced a big 10G Framer/Switch combo. It looks great on paper but their FPGA solutions have failed to make inroads in the presence of cheaper hard-silicon alternatives. Perhaps things are different this time.
  • As I have stated before, PMC-Sierra (PMCS) is no longer investing in this area. No one I spoke indicates this has changed.
• The competition is narrowing for the 10GbE EDC PHY socket at Cisco, which should represent 60-70% of the ultimate end market. There are two finalists, as well as a few startups. Vitesse has 100% market share in the first generation of 10G LRM X2 modules.
• During his presentation, Steve Alexander from Ciena (CIEN) referred to Ethernet over SONET/SDH (EoS) as a ‘legacy’ technology. This was the first reference I have ever seen to EoS being over the hill. As readers know, in the Telecom world, technologies don’t start to generate the real revenue until they’re mature. Therefore Mr. Alexander’s reference is both a positive and a negative.
• Level3 (LVLT) disclosed in it’s keynote that they are deploying 10GbE in spades. Lot’s of great data here:
  • Customers are approaching them looking for 10GbE hookups as they consolidate data centers into a single location to reduce costs.
  • Level3 has 500 active 10GbE ports today.
  • 10GbE client connections totally break the backbone, and 100G backbone transport is needed to enable this.
  • Level3 sees 100gbE leapfrogging 40GbE/OC-768.
  • By 2010 they expect edge connections to grow to 60x10GbE and the core to grow to 150 x 10GbE. I guess I can see why these guys like Infinera. (INFN)
  • This is driving a need for very large, cheap 10GbE switches with 64 to 512 10GbE ports. This was echoed by Equinix (EQIX) two days later at the Gilder Telecosm.
  • The presentation is worth a look. Legally, I cannot distribute it, but contact Level3 and I am sure they will make it available.
• RAD displayed an SFP format Ethernet to DS-3 bridge designed for Ethernet over PDH applications. Simply stick this into an access box and presto, it is EoPDH enabled. This is along the same lines at the PON on a stick Teknovus showed earlier in the year. (see “PON Moves Towards the Enterprise“)

I would like to extend a big Thank You to LightReading for putting on a free conference that rivals the quality of most paid conferences. The stories (and my late night photos) from the LightReading Leading Lights party will remain sealed for the professional benefit of all involved.

Author owns positions in AMCC.

They shared the following data with me.

We estimate the total merchant market for PON ICs was about $90 million in 2006, up 20% from 2005. We estimate PMC-Sierra had a share of about 60%, and we rank Teknovus and BroadLight second and third, respectively. A host of other vendors accounted for the rest of the market. In alphabetic order, these suppliers include Centillium, Conexant, Freescale, and Immenstar.

I estimate Teknovus revenue was $10M last year, derived almost entirely outside of Japan. According to Linley, PMC-Sierra (PMCS) did $54 million, which was almost entirely GE-PON. I still follow Centillium (CTLM) though I am no longer invested, and they reported $4M in revenue last year, some % of which was not directly PON silicon. This puts the GE-PON market at about $65M.

PMC-Sierra revenue grew from $45M in 2005 to $54M in 2006. On an absolute basis, the $300M acquisition looks like a hefty price tag given the growth rate. I thought the acquisition made sense (see “PMC-Sierra Acquires Passave“) and still feel this way (if you read the linked article, I thought PMC paid with overpriced stock). PMC-Sierra needs to use this business to win more silicon content (VoIP, VDSL, Processors) and I still believe they can do this. The numbers do show they are facing a fight outside of Japan as Teknovus appears to be doing well.

As for BPON/GPON, Broadlight revenue was driven almost entirely by BPON sales to Tellabs (TLAB) and Motorola (MOT). The remainder of the merchant silicon market is served by Conexant (CNXT), Freescale (Alcatel (ALU) is their lead customer), and Immenstar (now Cortina). These players account for low single digit share percentages. AMCC (AMCC) appears to have exited the market.

Not quantified in the Linley report are equipment vendors which use in-house designed silicon (ASICs). ASICs play almost no role in GE-PON in Asia, with the exception of OKI which designed their own chip after Centillium failed to deliver working silicon (see “Centillium – Cheap by Any Measure“). While Broadlight supplies the lion’s share of silicon into the BPON marketplace it must navigate the transition of Verizon to GPON. Unlike the GE-PON silicon market, which is supplied almost entirely by two vendors, the GPON silicon market is oversupplied (follow link for details) and faces a much greater ASIC threat.

Contact us with questions or for further consultation.

Full Disclosure: I currently have no position in any of the companies mentioned.

From the Linley Wire:

Over the past three years, Cortina has acquired Azanda, Intel’s optical products, and now Immenstar—a strategy that sets the bulked-up Cortina apart from wimpier single-product-line startups. With the Azanda acquisition, Cortina consolidated its component business for Cisco’s high-end routers. The acquisition from Intel more than quadrupled Cortina’s revenue and gave the startup access to most major communication OEMs. Most of the product lines acquired from Intel, however, were addressing mature markets. Immenstar gives Cortina access to the fast-growing PON market. ….. Given the ease that Cortina raises capital, we would not be surprised to see the company add to its broadband play with the acquisition of small VDSL vendor.

Centillium has a market cap of $76M but an enterprise value of $37M, all liquid (More of my opinion on Centillium here). Centillium also carries a large $21M non-cash liability on their books that I believe may never need to be paid. Removing the liability, the company is currently valued at $16M, less than 2x the most recent (and ugliest) quarterly revenue. It’s also valued at 1/2 the $29M in R&D expenditures made in the last 12 months.

The big negative is Centillium is feeding cash into the flames at a rate of $4.5M a quarter. It needs to be consolidated into a larger, more efficient entity. It makes no sense for $40M/year revenue companies to be public given the fixed costs associated with a Nasdaq listing. I believe that placing the operating entities of Centillium into a more efficient host would make the business cash flow positive immediately.

Next on the list is VDSL market share leader Ikanos (IKAN) at around $120M enterprise value. It too has a large cash stash. That would be a large acquisition for a company the size of Cortina, and the VC investors still holding their stake in Ikanos firmly believe the company has a future as an independent entity. I agree.

At current valuations, any company (including Cortina) could acquire Centillium, milk the revenue stream, and make the acquisition work financially.

I believe a public company, like Conexant (CNXT), PMC-Sierra (PMCS), or even Broadcom (BRCM) are more likely suitors. As public companies, they can issue shares for Centillium and reap the large cash balance. They would acquire a revenue stream, some useful R&D and future products, and a lump sum of cash for stock. Think of it as an indirect secondary offering.

PMC-Sierra is the best match, given it would eliminate a desperate FTTH silicon vendor who is distorting pricing, acquire a DSL product to complete its access portfolio, and pick up an extremely complementary gateway VoIP product.

Even Ikanos would make sense as it would consolidate competition and give them FTTH products, though I would argue they don’t need the distraction now.

One thing is clear- There is no incentive for private investors to bankroll Cortina to acquire mostly cash. There is a better incentive for other public companies to issue shares to capture it.

Author is long Centillium and Ikanos

This acquisition is surprising. Cortina previously focused more on the core of the network and has either built or acquired products designed to enable the next generation of WAN equipment for carrier applications. Immenstar is tightly focused on fiber to the home chipsets, a much more access oriented application. I suspect that Cortina is looking to couple it’s carrier class ethernet solutions with Immenstar’s high density OLT solution. Still, it is not a pretty fit.

Immenstar is a hybrid Chinese-American company with strong ties to UT Starcom, a company that at one point was poised to supply NTT (NTT) with large amounts of FTTH equipment. They stumbled (or were tripped by Passave, now owned by PMC-Sierra (PMCS) ) and Mitsubishi Electric took the majority of the lucrative and prestigious NTT contracts. As a result, Immenstar has not been tremendously successful in the marketplace. I suspect that this acquisition was Immenstar’s only option.

Cortina is one of the more interesting startups out there today, and appear to be pursuing a strategy of strategically integrating underperforming assets acquired from both public and private companies with a goal of consolidation. This is a core Nyquist strategy.

It’s unfortunate that companies like Centillium (CTLM) are not actively engaged in attempting to sell their own FTTH assets. As an owner of Centilluim equity I find this extremely frustrating. It’s clear that Cortina would have been an interested suitor.

Hat Tip: Venturebeat

Author is a Centillium stockholder

We lamented the growing barriers to entry for fabless startups, and talked about what needed to be done differently to make Telecom chip startups a reality. While everyone is avoiding Telecom chips like the plague, investors have not lost their moon-shot mentality of funding 6 businesses all chasing the latest IEEE standard (in this case, Copper 10GE Ethernet transceivers). Two land successfully and folks forget about the four that didn’t.

Other than Cortina, and the FTTH PON guys, there have been virtually no Telecom semi startups in the past 5 years. Zero. Not only that, but the incumbents have been doing little except make duals, quad, and octal port versions of the same designs from 2002.

Pundits proclaim the rebirth of optical yet no money is flowing. Which either means the pundits are wrong or a very big opportunity is lurking out there.

I’m convinced that a massive innovation opportunity exists in this sector that either an incumbent or startup can capitalize on. As is usual for me, I think it exists in the dirtiest and ugliest sector of all – Optical Transport. If you have an idea, I suggest you speak with Alex, because private investments just are not our bag.

During their session at Telecosm 2006, EZ-Chip (EZCH) and Raza Microelectronics read from the same script NPU companies have been using since 1998.

1. The network is filled with changing requirements, and therefore you need a flexible hardware solution.
2. There is this new market coming that will be the turning point (this changes about every 1-2 years, current flavor of the month is Metro Ethernet, last year it was access networks)
3. NPUs put companies on the same footing as a Cisco, allowing networking hardware to be commoditized. When this takes place the NPU will be to networking what the CPU is to computing.

Reasons 1 and 2 are linked. Whenever there is a new market (and quantities are low) there is a high degree of uncertainty in what features are required. If you speak with carriers today about Metro Ethernet requirements it is clear they have no idea what functions are needed. Ask them if they prefer A or B and they will answer “yes”. Ergo, an NPU is a great fit because it provides flexibility.

The problem arises once the market takes shape. The equipment provider now understands what features are needed, which are not, and which are niche. Once these requirements are well understood they can and will be codified in silicon.

When VLANs were first implemented by Xylan in the 90′s it was done in custom ASICs. Eventually, those features were cooked right into Broadcom (BRCM), Marvel (MRVL), Vitesse (VTSS.PK), and Agere (AGR) L2 switches that sell for $1 a Gigabit port today.

If something is required to be flexible and heterogeneous it is hard to see how it is adopted as a widely accepted networking standard. Unlike computing, networking hardware converges to a common set of features due to strict interoperability requirements. In Enterprise Networking, the Venn diagram of flexibility and high volume is a null set.

I see no reason why Metro Ethernet is any different than Enterprise Ethernet. Companies like Transwitch (TXCC), Zarlink (ZL) and Cortina Systems are already building dedicated silicon for Metro Ethernet.

Another barrier to the widespread adoption of NPU’s is that no common instruction set exists among all vendors. Given the frequency that NPU companies have failed there is great hesitation on the part of equipment vendors to put an single-sourced NPU in the critical path of product success. I’ve discussed this effect in depth. (see Wintegra – Following Up on a Comment)

NPU’s play a role, but only in situations where the flexibility they bring to the table matches the cost and risk premium . If dedicated hardware can process 90% of the traffic transiting a switch or router, then the remaining 10% is shunted to the side for processing by the NPU. This radically alters the market size calculations for NPU’s because instead of having one or two per linecard you have two or four per rack. Design win numbers offer no details on the magnitude of future business as a result.

Ultimately, I agree 100% with Gilder’s concept that the router will be hollowed out. This process is already underway with companies such as Vyatta providing a glimpse of the future. I believe Cisco’s (CSCO) only sustainable competitive edge is software and system integration, a la IBM.

But the hollowing out won’t be done with NPU’s. I expect that Broadcom and other ASSP companies will deliver hard silicon that is good enough for the requirements of the majority of applications, and ultimately NPUs will handle exception traffic only.

EZ-Chip is highly respected company and unlike most, has managed to survive, which says a great deal about the folks running the company. The niche market for NPU’s will sustain a small set of players, and perhaps that is Ez-Chip. When it comes to Ez-Chip serving a billion dollar market, historical precedent and market dynamics are on not on their side. I hope for the sake of employees and investors of EZ-Chip that this time is different.

Raza, which appears to me as an opaque promotion company, provided little quantitative information to make a judgement. Right now it looks like all sizzle, no steak.

Full Disclosure: I am long Zarlink and Vitesse, and am short Cisco as part of a hedge. I hold no positions in the other companies mentioned, including Ez-Chip/Lanoptics.

Full EZ-Chip presentation can be found here.

—————————–

Notes Below:

The Future of Networking Silicon: Programmable vs. Hardwire

Panelists:
Eli Fructer, President and Chief Executive Officer, EZchip
Atiq Raza, Chairman and Chief Executive Officer, Raza Microelectronics
Moderator:
George Gilder, Editor in Chief, Gilder Technology Report

Amir Eyal – EZ-Chip VP Business Development

Metro Ethernet is the target market for NPUs and is the make or break market. Discussed various market research projections about how big this market will be. Provided background on EZ-Chip products and competitors.

Atiq Raza – Raza Microelectronics (RMI)

Number and intensity of connections among devices is growing. They have a family of devices that can address these markets. RMI revenue growth is only limited by operational and manufacturing constraints. They refuse to do business with startups. My bullsh*t detector just lit up.

Panel Discussion

Bob Cousins is the moderator. Starts out with the standard NPU love-fest discussion about how everything is moved into software.

My Question: “I’m going to try and create a little controversy here. The Telecosm is littered with the bodies of NPU companies who claimed they would win because they have a technical advantage. It seems to me that the guys who are really benefiting from Metro Ethernet are the Broadcoms and Marvells, as off-the-shelf Ethernet switches are bootstrapped into new applications. How exactly do you fight this trend?”

Atiq Raza – We don’t see Broadcom in our applications. BS Detector buzzing…… The real measure of success is how many chips you ship (I interrupt) ”The real measure of success is whether you make money.” Raza: We do make money, and we will be profitable within 3 quarters. (If this was true I-bankers would be chasing him like a swarm of flies)

Amir Eyal – Requirements are continually changing, soft programmable solutions are required and hard coded Ethernet switches cannot provide the needed features. Same story…. I just don’t believe it. It’s possible I am wrong but I’ll take the risk and say that this time just isn’t different.

——————-