100G Ethernet

Infinera Ditches 40G, Talks 100G

Infinera Corp. (Nasdaq: INFN) has at last answered questions about its 100-Gbit/s plans, but its ditching of in-house 40 Gbit/s is sure to cause some controversy.

A week ago, Infinera was a company that had proven its worth at 10 Gbit/s, but hadn't articulated a convincing second act. Its 40-Gbit/s technology, in the form of a 10-lane, 400-Gbit/s photonic integrated circuit (PIC), lagged the market. And the company was reportedly going to be late to market with the Coherent technology that's taken center stage for 100 Gbit/s. (See Infinera Puts 100G Coherent on Pause.)

Infinera got a clean slate Thursday, saying it would skip ahead to coherent 100-Gbit/s technology, to be available in 2012 in the form of a five-lane, 500-Gbit/s PIC. Infinera won't be first to market, but could arrive in time for much of the early wave of demand. (See Infinera Sets 100G Agenda.)

So, why give up on 40 Gbit/s?

It's tempting to consider this as a defensive move. After all, Infinera still hadn't produced a 40-Gbit/s product "roughly four years after initial 40-Gbit/s systems started shipping," analyst George Notter of Jefferies & Company Inc. wrote in a research note on Friday.

Competitors, meanwhile, are already piling onto the 40-Gbit/s coherent bandwagon. CoreOptics, now being acquired by Cisco Systems Inc. (Nasdaq: CSCO), is offering transponder modules that are being used by Cisco, Nokia Networks , and, by all accounts, Fujitsu Network Communications Inc. (See Cisco Renews Optical Focus With CoreOptics and CoreOptics Does Coherent 40G.) Moreover, some analysts were under the impression that Infinera was going to have 100 Gbit/s out in late 2012 anyway -- which would mean the "new" timeframe has been the plan all along. (CEO Tom Fallon says Infinera never declared an official 100-Gbit/s timeframe.)

In that case, the only real news would be the cancellation of a late 40-Gbit/s product.

But Notter points out that Infinera's plan is the right one, regardless of motivation. It "gives the organization a better shot at getting to market on time and ultimately having larger 100 Gbps share vis-à-vis competitors," he wrote.

Timing is everything
Of course, Infinera says there's nothing defensive about its choice.

Executives claim the 40-Gbit/s PICs and modules were nearly ready, but Infinera has concluded that carriers are dead-set on 100 Gbit/s. Were Infinera to announce both PICs, carriers would ignore the 40-Gbit/s one and wait for 100 Gbit/s, Fallon tells Light Reading.

"They have made the mental leap," says the CEO. "Universally, we got the feedback that we made the right choice not coming at 40 Gbit/s in 2011 if we can come out with 100 Gbit/s in 2012."

Infinera says the work already put into 40-Gbit/s PICs can be applied to the 100-Gbit/s case; the transmitter side, in particular, is nearly identical.

The 100-Gbit/s plan came into formation when a digital signal processor (DSP) -- a key element of a coherent receiver -- "came in sooner than we had predicted," Dave Welch, Infinera's chief strategy officer, tells Light Reading.

That DSP was crafted by a team in Ottawa, the same engineers who'd brought coherent technology to Nortel Networks Ltd.

Infinera was originally going to develop coherent technology in-house, using an analog design in place of the DSP. But Welch says this group approached Infinera about the idea of joining forces, leading to the opening of Infinera's new design center last June. (See Infinera Does R&D in Ottawa.)

Because the 40-Gbit/s PIC is canceled, Infinera's 40-Gbit/s offering -- for the carriers that really need it -- will truly be a stopgap: a two-port card built from off-the-shelf technology. No PICs.

One door shuts, another one opens
So, the 100-Gbit/s question has an answer, at least. But that just means analysts can move on to the packet-optical question instead.

Consider this: DWDM boxes don't command the margins that, say, routers do. And Infinera, which owns its own chip fab, is doing DWDM the expensive way.

"I've always questioned whether they would get a return on that, doing a pure-play DWDM," analyst Michael Genovese of Soleil Securities Group Inc. tells Light Reading. (Notter likewise wonders whether Infinera can produce the kinds of earnings that boost stocks.)

Moreover, more carriers seem to be getting interested in a converged packet and optical core. (See Packet Optical Transport Goes for the Long Haul and Vendors Target the Packet-Optical Core.)

"They need an Ethernet switching story, and they need a packet-optical story. They need an OTN switch. That's what carriers want," Genovese says. "I've never had any clue from Infinera that they are either an optical OTN player or an Ethernet player."

— Craig Matsumoto, West Coast Editor, Light Reading

chechaco 12/5/2012 | 4:34:59 PM
re: Infinera Ditches 40G, Talks 100G

Proof was, is, and always will be in the pudding. Tme will tell.

Pete Baldwin 12/5/2012 | 4:34:59 PM
re: Infinera Ditches 40G, Talks 100G

So, I'm curious what you all think: Is this an offensive or defensive move by Infinera?

Pete Baldwin 12/5/2012 | 4:34:59 PM
re: Infinera Ditches 40G, Talks 100G

It was interesting to find out that Infinera wanted to try the analog approach to coherent.  That is, no DSP (and presumably no superfast ADC either? Didn't have time to get all the details).

Welch says the key was that it would have been lower power.  But that advantage goes away at advanced process nodes -- so the Ottawa team, working with line widths smaller than 90nm (40nm? I don't have my process nodes memorized like I used to) was able to keep the power down.  And of course, the analog design would have been difficult -- probably one of the reasons Infinera was reluctant to pin down a 100G arrival time.

paolo.franzoi 12/5/2012 | 4:34:58 PM
re: Infinera Ditches 40G, Talks 100G


I think it is neither.

It will be interesting to see if they hit the 2012 date.  Think of it this way, when do you think they thought they would have 40G available?  Do you think they missed their schedule and thus missed the market or that the market happened before they planned to be ready.

What they are saying is that there is not a 2nd gen opportunity on 40G that creates enough business to continue justifying the investment in 40G.  The next opportunity is now at 100G.



Stevery 12/5/2012 | 4:34:56 PM
re: Infinera Ditches 40G, Talks 100G

Hmmm.  So 2007 they announce an inverse mux:


2+ years later, they had just delivered 40G PICs from development:


And now they ditch it altogether.

Forget offense/defense, I'd like to know who they think is accountable over there for delivering technology on time and on budget.  Whoever assembled the 40G plan should not be let anywhere near 100G.  Send him back to 10G where he belongs.

Pete Baldwin 12/5/2012 | 4:34:55 PM
re: Infinera Ditches 40G, Talks 100G

> Hmmm.  So 2007 they announce an inverse mux ... 2+ years later, they had just delivered 40G PICs from development ...

Ah, you read my mind - I was just writing something up on this. Check the blog later.

boozon 12/5/2012 | 4:34:48 PM
re: Infinera Ditches 40G, Talks 100G

That DSP was crafted by a team in Ottawa, the same engineers who'd brought coherent technology to Nortel Networks Ltd.


...the prelude to a patent war between Ciena/Nortel and Infinera?

Sign In