Comms chips

Power X Powers Down

Switch fabric vendor Power X Networks Ltd. ran out of money today. And with no potential buyers or investors in sight, the company has been put on life support.

Nearly all Power X's 122 staff were given their marching orders this morning, according to several sources. Only five or six engineers are still on the payroll, their job being to take care of the company's intellectual property in the hope that a buyer can still be found for it.

Power X's story doesn't appear to be one of a company starting out with overinflated, overoptimistic ideas -- far from it. Founded in 1996 by a group of executives from ICL, a subsidiary of Fujitsu Ltd. (KLS: FUJI.KL), Power X was one of the first vendors to develop off-the-shelf chips that support packet processing. Its product, which started shipping in volume last year, was an intelligent switch fabric (see Power X Intros Chip Suite).

Power X was also one of the founders of the Common Switch Interface (CSIX) consortium, now the Network Processing Forum (NPF).

Estimated figures from RHK Inc. put Power X's revenues at around $350,000 in the first half of last year, and it's likely that its revenues were greater in the second half of the year, after general availability of its product was announced (see RHK Reports Packet Silicon Revenues). But it takes more than that to reach profitability, so Power X was trying -- unsuccessfully, it seems -- to raise a third round of funding.

This goes to show that meeting milestones and shipping product are not necessarily enough to satisfy investors anymore. And that's a worrying thought for a lot of startups.

Power X's announcements this week -- of its next-generation switch fabric architecture and of support for Intel's new IXP network processors -- also didn't sound like the kind of thing to come out of a dying company (see Power X Shows Off Next-Gen Architecture). But reportedly, those press releases were finalized just days before the administrators were called in.

Backruptcy advisors BDO Stoy Hayward were called in on February 1 to try to sort things out (see Power X Restructuring). BDO didn't return our call, and one ex-employee says that the administrators kept everyone in the company in the dark. It was reported in the trade press that BDO had talked to potential buyers but that there were no really hot prospects.

Despite BDO's stated intention that it would try and keep the company together as a unit, it now looks too late for that. "I think the IP will find a home, but the company won't," says one U.K. analyst familiar with Power X's situation. However, he does wonder if there are buyers waiting in the wings, hoping to get the company at a knockdown price.

Power X was funded by Intel Capital.

— Pauline Rigby, Senior Editor, Light Reading
dietaryfiber 12/4/2012 | 10:51:29 PM
re: Power X Powers Down The problem with this and we shall see how the NP business plays out as well is that as stated Switch Fabric design has allowed for competition on architecture cost and features.

ASICs have traditionally cost an order of magnitude less than off the shelf switch fabrics. They also contain the elements that the architecture requires. Thus, the platform cost and feature set is tuned to the switch fabric design.

On the NP front, its not clear to me that any of the successful mainstream designs (i.e. Juniper and Cisco) use NPs and if they do in what volume. I think that the jury is still out here, especially as IP technology pushes deeper toward the network edge (Tandem Offices and ISP POPs are hardly the edge).

dietary fiber
gea 12/4/2012 | 10:51:36 PM
re: Power X Powers Down That Toonces thing was one of the most hilarious things I ever saw. In the skits the characters always thought it was so great and amazing that Toonces could drive, but he SUCKED and he'd always wreck the car (killing everybody inside). The Toonces puppet was also wacky...right before another car wreck its mouth would open up in shock. The skit never made a hell of a lot of sense, and there was no punchline per se, but it never got old. (Oh yeah, I think I remember Toonces running over some bouncing dog puppet in jealousy).
OneMoreByte 12/4/2012 | 10:51:38 PM
re: Power X Powers Down > The Semi's game is volume and they learn it
> well. It only take 1-2 engineers to develop a
> switch fabric. How much you can save for that?

Yeah, I couldn't understand why it took Power-X 4+ years and 30+ people to produce a 16xOC-48 fabric chipset. I know a company where they cranked out a similar chipset (for internal use) in just over a year with only four people. In general, no startup's product should take more than two years to bring to market.
lightbulb0 12/4/2012 | 10:51:42 PM
re: Power X Powers Down If you take a look of all those fabric company's solutions, you don't see much difference. What they bet on was selling them to some box startups when they try to beat the clock by cramming out junky solutions. If the startup get bought, it will be a real design-in of big companies. That is not happening, box startup are fading away because of no differentiation and weak market. It may reach the level of attention for a couple of years.

Speaking of physical layer and framer. Those had never been the core competency of the system companies and they are not interested in owning them. Those problems alone are enough to make a business case.

Switch Fabric is the exact mirroring of .COM in networking industry. Little technology and big hypes.
mrcasual 12/4/2012 | 10:51:43 PM
re: Power X Powers Down It is noticed that Broadcom was the only company which did not join the
switch fabric qcuisition party. The Semi's game is volume and they learn
it well. It only take 1-2 engineers to develop a switch fabric. How much
you can save for that?

Methinks you underestimate the size of the switch fabric problem by several orders of magnitude.

The issue with commercial switch fabrics has always been that most people use their fabric as a way to differentiate their product.

Not too long ago, everyone built all their own ASICs. Then they stopped building the physical interfaces and framers. AMCC, PMC-Sierra, Broadcom do that well enough. Now NPUs are starting to take over the "standard" packet processing functions.

Traffic management and switch fabrics will be the last to go to commercial silicon because the product requirements are too diverse across the spectrum of customers so the fabric vendors have a hard time pleasing enough people to make enough sales.

This may change eventually but the fabric business is not an easy one to win.
lightbulb0 12/4/2012 | 10:51:44 PM
re: Power X Powers Down Raid,

It is noticed that Broadcom was the only company which did not join the switch fabric qcuisition party. The Semi's game is volume and they learn it well. It only take 1-2 engineers to develop a switch fabric. How much you can save for that? Cost has never been a issue and it is not a technical problem which desparately seeks solutions either. The entry barrier is the low of the lowest. How much was Abrizio valued at the time, 400million? It is probably only worth 400k, i.e. two engineer's salary, I mean GOOD enginees.
siskodude 12/4/2012 | 10:51:45 PM
re: Power X Powers Down I will take the liberty of answering this question.

look here:


to think I was afraid I was exajerating when I said EVERYTHING is on the net...my fears are no more
Cyras Forever 12/4/2012 | 10:51:46 PM
re: Power X Powers Down Thanks for the explanation. Toonces was featured regularly on Saturday Night Live, yes?
toonces_has_the_car 12/4/2012 | 10:51:47 PM
re: Power X Powers Down I will take the liberty of answering this question.

look here:


Cyras Forever 12/4/2012 | 10:51:48 PM
re: Power X Powers Down gea - what is the signifiance of "Toonces has the Car"

Is it a children's book title?
gea 12/4/2012 | 10:51:49 PM
re: Power X Powers Down I think "Toonces has the car" deserves the "best handle" award from Lightreading!!!
toonces_has_the_car 12/4/2012 | 10:51:49 PM
re: Power X Powers Down Respectfully, you can't really compare Broadcom and PMC to Power-X. Both of these guys have broad product lines, and track records of delivering successful products. Power-X is (er...was) a one product company.

I agree with Lightbulb's assessment of the switch fabric market. The big equipment vendors mostly develop their own fabrics. I think this is less an issue of cost than of features. The fabric semi developers are trying to build a general-purpose fabric with broad market appeal. But each equipment vendor have their own set of specific requirements. So, if the feature set isn't very close to what the equipment vendor wants, they will look elsewhere instead of trying to put a square peg in a round hole.

Another consideration: A switch fabric is not easily replaced once designed in. Who wants to bet the farm on some small company that may get bought or fail?
raid 12/4/2012 | 10:51:51 PM
re: Power X Powers Down LightBulb
What about Broadcom? They seem to be doing pretty well.

The trick seems to offer a wide range of products
and make a some money in each segment.

To be profitable, the IP vendors will need to offer a wide range of products since the market is not that huge (atleast nothing like Intel's processor market). This inherently means that there is space for 2-3 strong players. The rest will probabaly not make it - hopefully, some will get bought.

Of late some system folks are learning some bitter lessons - buying IP from startup vendors running out of money. The whole project schedule and sometime the company's future can be compromised.

lightbulb0 12/4/2012 | 10:51:52 PM
re: Power X Powers Down It has been an inflated dream that the networking industry will follow the route of PC industry. All comm semi company developed or acquired some sort of swutch fabric technology when the economy was red hot. Unfortunately, for the high volume and low margin network segment, shared-memory switching dominates the backplane while in the low volume, high margin segment, i.e. metro/core routing market, each major network box company has their own switch solution. If they use the off-shelf parts, essentially there will be no money left for them to make. After all, what make it so difficult for box maker to develop their own ASICs? The semi comm does not even understand their needs. PMC-Sierra learned this in the hard way. Watch out, more are coming ...
Sign In