INFN in trouble?

What sez you, Kevin?…

Microsoft Drops a Data Center Interconnect Bombshell

Just as the data center interconnect (DCI) system market was taking off, one of the biggest potential buyers of dedicated DCI systems boxes, Microsoft, just wiped out a whole chunk of the market courtesy of an alternative R&D collaboration with optical components vendor Inphi Corp.

Speaking here at the OSA Executive Forum Monday afternoon, Tom Issenhuth, Optical Network Architect, Azure Networking, at Microsoft Corp. (Nasdaq: MSFT), laid out the hyperscale data center operator’s DCI technology needs. All well and good – the audience of about 250 attendees were suitably rapt.

But then he dropped the bombshell: For Microsoft’s “sweet spot” of less than 80km inter-data center links, it has worked directly with Inphi to develop a new module that plugs directly into data center switches, obviating the need to deploy a dedicated DCI box.

The product developed by Inphi and as yet unidentified partners is a 100Gbit/s QSFP28 DWDM module that slots straight into data center switches from the likes of Arista Networks Inc. and Cisco Systems Inc. (Nasdaq: CSCO). Inphi and Microsoft are going to demonstrate the product and how it changes the DCI landscape during the OFC show here in Anaheim – more details are set to become available Tuesday.

And of course Microsoft might not be the only company that wants this product – it is not exclusive to Microsoft. Light Reading asked Issenhuth if Microsoft had held discussions with the likes of Google (Nasdaq: GOOG) and Facebook about the development: He said they chat about general market developments but that this particular move was one it was leading on its own. But “if it works for us, it should work for them too.”

And then, of course, there’s the impact on the systems vendors such as ADVA Optical Networking (Frankfurt: ADV) , Ciena Corp. (NYSE: CIEN), Coriant , Fujitsu Network Communications Inc., Infinera Corp. (Nasdaq: INFN) and Juniper Networks Inc. (NYSE: JNPR) (courtesy of its acquisition of BTI Systems), that have been developing and building dedicated DCI boxes to sell to the likes of Microsoft as well as the telcos. (See Juniper Flies Into DCI With BTI Acquisition.)

“This has potentially huge implications for the purpose-built DCI box vendors,”

Hmmm! This may not be good news for Infinera



First it’s important to note that this is the area between datacenters that are less than 30 miles or so of reach.

This are of the market has lots of players in this space, Arista is one of them. I remember Stuart Elby had a comment about Arista’s DCI product, which serves the same application, when he was asked if they would now be in competition. He had something similar to say about the reach - but for longer than those 30 mile distances Cloud Xpress is still the clear winner.

I’m not concerned about this development for Infinera. I’d be more concerned about the other players in this space that already have problems at longer distances.



With INFN the stock being down close to 9% on the news, Monkey here is just seeking (one more time) to confirm that this seems to be the market’s misunderstanding of Infinera’s strengths and niche; and that the market is overreacting based on a misunderstanding of the core business which is probably triggering further selling.

How confident are you, Kevin, that your assessment is closer to reality than the market’s perception of it? Monkey doesn’t mean to put you on the spot (again), but if you’re right–which you usually seem to be because of your tremendous expertise, then this is another wonderful opportunity to grow a few more bananas on the cheap.

Just looking for a bit more confidence in your assessment is all.




I’m not concerned about this development for Infinera.

It does sound like it shrinks INFN’s potential DCI market some, though, right?

Not saying it shrinks the company’s value 10% like the market is saying (seems a bit overdone if you ask me, from an already low starting point), but it does sound like it takes away some % of potential sales.

1 Like

Very confident.

Here is the reference. It took a while to find:…

Comments on the Arista DCI line card announcement. Last week, Arista introduced a six-port 100G DWDM line card for its 7500E Spine switch targeted at the Data Center Interconnect market. The card has coherent DSP (digital signal processing) supplied by a third-party merchant vendor and claims transmission distances of up to 5,000 km. However, when we spoke with Elby of Infinera, he pointed out that Arista does not have a line system (i.e. amplifiers) and therefore, in practice, is limited to only very short distances (i.e. less than 20 km.) Further, the major Data Center builders are likely to deploy an integrated Switch + Optical solution only in very short reach on-campus building-to-building applications. Finally, Elby noted Infinera’s DCI business plan is much more focused on the more than 20 km market, and particularly 50-500 km, which is the large majority of the total opportunity.

So, while there is some potential to take 50-80km out of Infinera’s DCI bread and butter, it should be minimal.

Also, you have to look at the entire space of the market. DCI is one facet, and it is growing, but it is considerably smaller than the metro space. I would say the market size for metro is about the same as it is for long haul. If you put ratios on them it would go 5:5:1.5, where 5 and 5 are for longhaul and metro, and 1.5 is for DCI. Then, when you look at that part of the pie, you’re still talking about a small portion that has the potential to go to this Inphi solution.

I honestly don’t think this new product will have a “9% haircut” impact Infinera’s core position in this space - and it doesn’t even go near the other spaces Infinera is either 1) already dominating (long haul) or 2) about to get into (metro).

The market is also not pricing in the potential of what Infinera just introduced. The flexibility introduced with their new Infinite Capacity Engine offering may change a few minds on where customers want to deploy Infinera equipment.

I would view this as another opportunity to buy more bananas - but - I would love to have others chime in here as well.



This MSFT announcement is not a groundbreaking development IMO - Infinera has openly discussed this type of integration as a possibility for shorter distances. It isn’t good news, but does not impact longer distances or other Infinera markets such as metro and long haul.

I would view this as another opportunity to buy more bananas - but - I would love to have others chime in here as well.

I agree with Kevin, this is a buying opportunity. If my position in INFN were not already rather large I’d have bought when I saw the price drop to 15.25!

If you doubt that these are companies focused on different markets, just consider the recent press release for each:

MSFT: 100 GBit/second, less than 80km
INFN: 2400 GBit/second, up to 12,000km

I’m not worried.

Kevin, thank you yet again for your insights into Infinera!


Did a little more digging on the Inphi solution.

What they are building is not at all ideal. It seems like a stopgap solution to me, and it is far from flexible. The solution also requires equipment along the way to keep the signals in tact:

“with the only equipment between the switches being the necessary amplifiers and multiplexers/demultiplexers.”

So, they have to break the line along the way and install equipment in another powered housing unit/building (or two, or three,) to satisfy the requirements. Sounds like patchwork.

Here is another person’s take (who is more familiar with the equipment and needs) on the Silicon Investor boards:

“This seems like it might be a simple low cost way of connecting the data center, but it seems like it might be a very short sighted answer to the problem. INFN’s solution maximizes capacity on each fiber and gives you a way to manage it and add / change capacity, so it future proofs your investment. By not maximizing the capacity and flexibility of your fibers, when you upgrade you have to pull everything out and do it again, and to add capacity you have to do a physical upgrade that adds more and more fibers- It kind of gets away from the whole promise of SDN and takes us right back to the truck roll /digging physical upgrade and managing the network with hardware which is why the promise of SDN came about in the first place- The true cost of this solution needs to be examined- You might be saving on hardware costs, but management, installation, and upgrade costs need to be taken into account. With the data growth accelerating this seems to be going backward- When you hardwire your data centers together in this fashion it seems like you are pushing your costs back in the short term, but your setup will obsolete sooner and end up costing you more than if you had gone with cutting edge equipment- The idea that this announcement is a big surprise seems suspect to me-”

I don’t think there is much to worry about here.



Maybe Infinera will slide in some verbeage at the media event later today to emphasize that their NEW technology (infinite capacity engine), not Inphi’s is the big deal stuff that will change the networking world and the stuff that will produce the real results (earnings).

As excited as I am about all the stuff I hear about Infinera I do develop some fatigue over the constant fact that something always seems to come up to obscure their growth, new products, etc… I know the answer likely comes down to patience so I will sit and wait for the realization of things to run their course.