Article Info

Cisco’s Scorched Earth Strategy

Cisco swallowed another chip company this morning, Greenfield Networks. The notable thing about this acquisition is that Cisco rival Huawei/3Com built their high end system around the Greenfield device. I’m willing to bet that Greenfield was a lot more important to Huawei/3Com than it was to Cisco. And I’m willing to bet that’s why Cisco bought them.

By purchasing Greenfield, Cisco continues their scorched earth strategy of buying key suppliers in order to deny these products to competitors. This move shows the lengths Cisco will go to to preserve it’s dominance in the L2/L3 switching space. Wikipedia has a comprehensive list of all acquisitions- it would be neat if they had the purchase prices as well.

I profiled their reliance on reselling optical modules as a source of earnings growth, but Bill Koss nails the reason why they can do this, and why they make so much money today – market share.

When Cisco announced their quarterly results yesterday, the fist thought that came to my mind was how nice it must be to own 80% market share for routers and switches. I know the Cisco team talked about ‘firing on all cylinders’ and a ‘validation of their business model’, but the reality is they secured their customer base in the early to mid 1990s and there has been no compelling event to force enterprises to make vast, capital intensive changes to their network using a supplier other then Cisco.

John Chambers handles his job as master of ceremonies exceptionally well during conference calls and while speaking on TV. He even managed to squeeze a reference to Web 2.0 into his CNBC interview, ensuring his dialogue delivers the buzzwords the throbbing masses demand.

What you won’t hear Chambers say is that Cisco is successful because they are 80% of the market and the market for networking equipment is growing at 15-20% a year. This incumbency position provides them with several advantages, including monopsony power over suppliers and the ability to turn commodity items into value added products.

It also gives them the mass to shoulder aside competitors and pull the rug out from under them by acquiring key suppliers. People harbor crazy ideas about oil companies buying and burying 200 MPG engine technology. Cisco does this routinely with silicon suppliers.

Cisco pulled the exact move in 1997, when they acquired Skystone, an Ottawa based chip maker that was the sole source of OC-48 POS/ATM framer silicon. When Cisco assimilated Skystone, several other tier 1 equipment vendors were left without framer silicon and effectively gave the GSR12k a 6-9 month lead in the exploding (at the time) Telco market. They followed it up in 1999 by acquiring StratumOne. These events also catalyzed a gold rush for SONET/SDH silicon suppliers, resulting in an oversupply of products that still exists today.

The game plan for a comm semi supplier looking for liquidity is pretty straightforward.

  1. Design compelling silicon that enables a valuable networking equipment feature (sorry, no getting round this one)
  2. Get product thoroughly designed into Cisco Competitor to the point where they are 9 months pregnant with your silicon.
  3. Get product designed into Cisco (optional)
  4. Get acquired by Cisco
  5. Write note of apology to Cisco Competitors #1, #2, #3 . . . .

Unfortunately for Netlogic (NETL), they completed step 3 before step 2, and derive 60-70% of their historical revenue from Cisco. Cisco competitors are wary of building systems around their devices. And Netlogic is stuck fighting Cisco purchasing for a few extra points of margin day in, day out.

Meanwhile, Greenfield played the game right. And Huawei/3Com got pwned.

Full Disclosure – I hold Cisco puts.


Comments are disallowed for this post.

  1. Sorry Andrew but this one you got wrong. Though I agree that was Cisco’s strategy in the mid-90s when they acquired every switching vendor under the sun (not to mention right under the uppity noses of their competitors) this is not the case today. Cisco’s acquisition of Greenfield signals the end of the generic network processor as the building block for next-gen switches.

    Greenfield, Fulcrum Micro and several other silicon switching start-ups are proving that the generic NPs cannot keep up with market demands; there was a reason why we paid a king’s ransom for good ASIC designers and, thankfully, they are back in vogue. As we say in Hollywood, custom silicon is the “new black” The fact that this acquisition sticks it to their competitors is just icing on the cake; pretty tasty icing at that. Just watch to see what happens to Fulcrum and a couple of other players I won’t mention (if only to protect the identity of a slow-to-react Cisco competitor). But it does bode well for these chip start-ups for Broadcom (and others) will be calling soon, very soon.

    Bill Baker

    Posted by Bill Baker | November 14, 2006, 3:00 PM
  2. Andrew…I just posted a comprehensive list of Cisco deals since 1993 at Enjoy!


    Posted by Bill Koss | November 14, 2006, 3:15 PM
  3. I thought about folding my thoughts on NPU’s into this story but decided it was a completely separate issue. I am convinced NPU’s will remain “exception processors” (my term) and standard products will rule.

    But if that was the case, they why would Cisco take a standard product maker in house? They would leave them outside, and encourage competition to ensure low cost high innovation suppliers. Where am I wrong exactly?

    Posted by Andrew Schmitt | November 14, 2006, 3:31 PM
  4. Given your analysis of why Cisco is dominant and will continue to be dominant, why would you be shorting them? I can see not taking a long position, especially give the little run up in the stock the last few months. However, shortening them seems contrary to what you are saying.

    Posted by Sak | November 14, 2006, 4:20 PM
  5. Risk management.

    Posted by Andrew Schmitt | November 14, 2006, 4:30 PM
  6. I think your initial premise is false: if Cisco picked up Greenfield for ~ $10M as rumored, then they weren’t important to anyone. Surely if Greenfield were strategic to Huawei-3com (or any other vendor), those vendors would have negotiated a secure supply agreement, and rights of first refusal.

    The fact is that Greenfield’s VCs bailed on them, which either says something about the quality of their product in particular, or about the market for metro switching silicon in general.

    Posted by Schlettie | November 14, 2006, 4:37 PM
  7. Prior to NPUs the industry, including Cisco, developed its own ASICs. Today, NPUs do not provide clear differentiation since everyone is using the same fabric (note: I never bought the argument: ‘it’s the software stupid’). So, either Cisco can develop its own custom ASICs or acquire a standard product via a ‘custom’ acquisition; the end-result is the same — Cisco has it and their competitors don’t. This is very different than the mid-90s’ when Cisco did not only buy every switching vendor, they bought an entire industry (and market); Bill Koss is right about how they got to 80% market share. And they did it using “Cisco money”, which was better than real money; guaranteed returns of 300% plus during Cisco’s hyper-growth period.

    Today however things are different; Cisco money can have negative returns which is why they cannot (and will not) buy every com silicon start-up out there. You are confusing tactical moves (a couple of silicon acquisitions) and assuming they are a part of a grander strategy. Cisco’s grand strategy is software (SOA) centric for they now that ‘he who controls the network controls the universe’.

    Posted by Bill Baker | November 14, 2006, 4:44 PM
  8. Bill, a post like yours is not complete without a little music.

    Readers: Click Here for the full effect.

    Posted by Andrew Schmitt | November 14, 2006, 4:49 PM
  9. Not sure if i get the music reference, but whatever — seems like a weak comeback. That said, Bill Baker could definitely explain his rationale better, because i don’t follow it.

    My limited knowledge of Cisco is that they maintain a healthy ASIC capability internally but also leverage external chipset solutions (e.g., Broadcom’s StrataSwitch) with a custom veneer as needed. There is no all or nothing strategy from a component sourcing standpoint, but the preference is definitely for internally developed stuff.

    I understand Andrew’s opinion a little better and it rings true, but Schlettie has a good point that the low price doesn’t seem like it was that critical to Huawei/3Com. Seems like it would cost Huawei/3Com much more than $10m to develop similar silicon internally, plus the lost time to market.

    I think i would still classify this deal as a question mark.

    Posted by ChipGeek | November 14, 2006, 5:29 PM
  10. very interesting blog Andrew, all the more so because ur views are well reasoned but i disagree with far more of them than i would expect :-)

    first, i agree that in some situations, taking a player out the game that could help a big competitor is a significant motivation for cisco. the procket acquisition is another example.

    however, there are lots of other points where i think u underestimate what it takes to get successful products out there, especially in the service provider market (eg metro ethernet). if u say u can’t see why metro requires different asics from what powers an enterprise switch, then i don’t mean to be smarmy, but i’d suggest you get someone to talk u through what it takes to implement the specs. IEEE 802.1{ad, ag and soon ah} are all baseline requirements for starters, and they simply won’t run on silicon burned to enterprise standards like 802.1Q. non-programmable asics have a place in cost-sensitive markets, but they always bite both vendor and customer in the end, and certainly for markets with the sales/validation/longevity cycles of service provider kit. deciding what functions to burn into gates and what to leave programmable, and in what way, while keeping within severe power/gate/pin constraints, is a fine art, and if you see inside any successful networking vendor you will find truly exceptional engineers making those calls. the best asic architectures last for many years, despite bandwidth doubling every year. saying npus will be “exception processors” shows no appreciation of this. if we knew what the non-exception path was, that would be safe to burn in, it would be a different industry.

    a consequence of this is that great asic designers are still exceptionally valuable. ethernet is about price, and asic price is about volume. cisco can afford to invest in its ethernet asics in a way few can (broadcom and marvell maybe) because of its volume, and thereby can burn in features that the others don’t have but integrate with its overall value proposition, thereby maintaining its market share and gross margins. a great virtuous circle. but it needs the asic design capability to keep ahead of the rest.

    so while i agree that the desire to take out a player that gives the competition an edge is certainly part of the calculation, i think you underestimate how much of a positive motiviation comes from buying great silicon and the people who developed it. i agree with bill – custom silicon has been the new black for some time now ;-)

    Posted by quefrency | November 14, 2006, 5:30 PM
  11. Excellent comments everyone….

    I am pretty familiar with the Metro Ethernet requirements. I think that carrier feature requests will be triaged as equipment volume goes up, and Enterprise silicon will be enhanced to meet these requirements.

    The uncertainty that exists today will not last. Once it is gone the programmable solution will not be as compelling. The number of possible ‘exceptions’ will be reduced.

    You all clearly know my position on NPUs for Metro Ethernet. But the post was more about Greenfield getting bought and Huawei/3Com getting hit from the blind side (Stop before 50 sec mark to avoid the gore). And I can see the point that disabling Huawei/3COM could have just been a positive outcome, and not the motivating factor.

    What is this stuff about Greenfield going for $10M? If that is true, then everything we are saying here is pure horseshit as that is a fire sale valuation.

    Posted by Andrew Schmitt | November 14, 2006, 5:52 PM
  12. Rationale: Standard/custom chip discussion notwithstanding my point was that the acquisition in question did NOT signal a “scorched earth strategy” by Cisco (as it relates to com silicon, which was Andrew’s main point). Andrew also backed up his argument with a similar strategy implemented by Cisco in the mid 90s (per Bill Koss reference) and I simply took an exception to his supporting argument; the two are not related. Hope this clarifies my view.

    Posted by Bill Baker | November 14, 2006, 5:54 PM
  13. Andrew-

    I think you have misread the implications of Cisco’s acquisition of Greenfield. It was a fire sale. Greenfield did not “play its cards right” and Huawei-3COM did not “get hit from the blindside.” Greenfields VCs refused to invest more capital in the struggling business and the company could not secure a strategic investment from their customers either (including Huawei-3COM). The only one that seems blindsided by this was Netlogic, which for some inexplicable reason chose to tie the commercialization of their Netlite processor to an undercapitalized start-up.

    The EZchip representatives at the Gilder conference were fully aware of Greenfield’s impending demise and indicated that Cisco was likely to absorb whatever IP was there under the headline of an acquisition. Greenfield’s ASIC approach did not afford OEMs the programmability they currently need which made securing design wins impossible. The “acquisition/fire sale” of Greenfield higlights the ascendancy of fully-programmable NPU it does not challenge it.

    Posted by Michael | November 15, 2006, 11:19 AM
  14. I suggest everyone to listen to Netlogic earning conference
    call where Greenfield problems were discussed by the Netlogic
    CEO. I am in complete agreement with Michael. Greenfield
    sale was a fire sell and was known to people for a while.

    Posted by Peter | November 15, 2006, 12:51 PM
  15. Okay, consensus seems to be that Greenfield was a fire sale, but i am amazed that that therefore proves the contention that NPUs are on the rise. Where? Please don’t say EZ-Chip, which has a barely detectable revenue stream.

    Posted by ChipGeek | November 15, 2006, 3:04 PM
  16. Distinguish NPU as a technology from NPU as a product category for chip vendors. Cisco/Juniper/Alcatel all have internally designed NPUs in their various boxes.

    Posted by Schlettie | November 15, 2006, 3:43 PM
  17. Hey for $10M to slow down Huwaii/3Com positiong – nothing worng with that Huwaii does the same in China. Cisco is #1 and it makers perfect BUSINESS sense to do this.

    The larger problem is CISCO manipulates the silicon supplier market with questionable business practised. They where one of the main catalysts for the market crash in high tech and the unwarranted buying frenzy and escalating purchasing prices for companies. Ask any major CEO of silicon vendors why they did certain things and the route goeas back to Cisco.

    Intentionally I think not but when your the big bully on the block its difficulty to recognise the damage you do – let alone be accountbale.

    Posted by BritishBullDog | November 17, 2006, 2:03 AM
  18. NPUs will win out because of programmability while still being able to act as a switch as chip processes continue to shrink and increase in performance. The first company to come out with a god-box like NPU/ASIC will do well. Network convergence is the name of the game from here on out.

    While some of Cisco’s practices seem questionable, they know they must head off the upstarts like Huawei at the pass. $10M is probably one day’s interest of Cisco’s coffers. Cisco today is the Microsoft of the 1990s.

    Does anyone remember a 10G Ethernet processor startup Cisco acquired a few years ago in the Dallas area ?

    Posted by NetworkObserver | November 28, 2006, 9:24 PM
  19. Navarro.

    Posted by Andrew Schmitt | November 28, 2006, 9:40 PM
  20. Trackbacks / Pingbacks

  21. Cisco’s / Greenfield & The Chess Game ;-) « Iain’s Chips & Tech | November 15, 2006, 2:23 PM