Optical components

Intel's 10-Gig Shopping Spree

Intel Corp. (Nasdaq: INTC) today revealed its intention to purchase three optical components companies (see Intel Scoops Up Chips).

The big question is: Can Intel use these acquisitions to put itself on the optical components map? Taken individually, the acquisitions don't rate as anything special. But the real value may lie in their combination as Intel's platform for attacking the OC192 (10 Gbit/s) components market.

Intel intends to pool expertise from its optical acquisitions to create low-power, small-footprint transponders that can send 10-Gbit/s signals as far as 80 kilometers. "We're focusing on industry trends to reduce power, cost, and increase reach," says Tony Stelliga, Intel's general manager of strategic marketing and business planning. "Carriers want to put more transponders on a line card, so they can generate more revenue per foot of colocation space. They also want to go longer distances so they can reach more customers, but, at the moment, they need more power to do it."

The biggest of the three outfits, with approximately 80 employees, is Newark, Calif.-based LightLogic Inc. The startup already has customers and revenues for its flagship product, a small form-factor optical transponder, which was unveiled in February (see LightLogic Launches DragonFly). The firm's price tag: about $400 million, according to Intel. The deal is expected to close this quarter.

"LightLogic has made major strides in automating the process of attaching fibers to the module, and it's got the alignment to a very high precision," says Stelliga.

The other two acquisitions, now completed, are Cognet Microsystems Inc. in Los Angeles and nSerial Corp. in Santa Clara, Calif. Both specialize in CMOS (complimentary metal oxide semiconductor) chip technology, a low-power type of circuit.

A developer of high-speed electronics, such as laser drivers and pre- or post-amplifiers, Cognet claims it has products shipping to customers in pre-production quantities. Financial terms of the deal were not disclosed, but, according to a source in the industry, it could be in the region of $200 million.

nSerial is a smaller outfit. It's a startup specializing in serializer/deserializer (SerDes) devices, which bridge the gap between high-speed serial traffic coming off a fiber or backplane and the slower internal workings of routers and switches. It's in a hot market segment but appears to be behind the competition in its product development as it has yet to ship products or even put content on its Website (see Accelerant Boosts Backplanes). nSerial commanded a purchase price of approximately $66 million, says Intel.

To send signals farther, Intel will use FEC (forward error correction) coding techniques it gained from its earlier acquisition of Denmark's Giga A/S, in combination with LightLogic's high-precision alignment techniques for getting more light into the fiber. Cognet and nSerial will contribute low-power chips to the overall solution.

The end game? Ultimately, Intel has designs on a 10-Gbit/s transponder for sending signals 80 km, which consumes 8 watts, occupies 12 inches, and costs $10,000. An equivalent device made with today's technologies consumes 22W, needs 127 square inches, and costs about $80,000, the company claims.

-- Pauline Rigby, Senior Editor, Light Reading http://www.lightreading.com

sonet49er 12/4/2012 | 8:31:17 PM
re: Intel's 10-Gig Shopping Spree Look at
rather than just accepting Intel claims.
waveform 12/4/2012 | 8:31:13 PM
re: Intel's 10-Gig Shopping Spree However, wouldn't a good point be that the Lucent transponder is only good to less than 12km and Intel is stating the desire to push the distance out to 80km? O.K., so they blow their own horn a little, but that is to be expected? Just a thought...
metroshark 12/4/2012 | 8:31:09 PM
re: Intel's 10-Gig Shopping Spree First, the Lucent data sheet posted by the original poster is for an IR (intermediate reach, 1310nm) transponder. I beleive the long reach (1550nm) transponders from Lucent, which would be more comparable to the LightLogic product, have a reach that is significantly more than 12Km.

There are a number of vendors who have been shipping long reach OC-192 transponders for more than a year. These products are priced somewhere between $15K and $20K and the power consumption is typically around 8W-10W. As the volume on these components pick up, especially as 10GigE applications start ramping up, it is reasonable to expect prices to fall down to $8K-$10K range. If Intel introduces a long reach 10Gb/s transponder with 8W power consumption at $10K price some time next year, I don't think it is going to make a major impact on the market. Their value add may be integrating all the 10GigE physical layer components to the transponder.
sanjiv 12/4/2012 | 8:31:08 PM
re: Intel's 10-Gig Shopping Spree I was wondering how much does it cost to do electrical regeneration at 10Gb/s every few hunderd Km or so (i.e. what type of components are present at the regeenration points and total cost for 10Gb/s regeneration). Can anyone please help.

PBC 12/4/2012 | 8:30:53 PM
re: Intel's 10-Gig Shopping Spree Reality check please...

Transponder, as in, taking in a 1310 lambda and mapping it to an ITU-T compliant 1550 lambda for transport on a DWDM network?


gladysnight 12/4/2012 | 8:30:37 PM
re: Intel's 10-Gig Shopping Spree Reality check please...

Transponder, as in, taking in a 1310 lambda and mapping it to an ITU-T compliant 1550 lambda for transport on a DWDM network?

Yes, that seems to be what they're talking about.
But it still seems logical to take:
- the fibre attachment advances (goes to insertion loss?)
- the low power chipsets
- other advances (?)

and try to combine them for a "new" product that is competitive across a number of dimensions.

Whether or not this is actually feasible, or in any way profitable, remains to be seen.

But I don't think they were talking about developing a new transponder, just about using some of the intellectual property that is in the transponder and applying it to other uses.

That's how it seemed to me, anyway, I could just as easily be dead wrong. It's been known to happen. :-(
PBC 12/4/2012 | 8:30:27 PM
re: Intel's 10-Gig Shopping Spree You go Gladys!


Sign In