400GbE: Wakeup Call for Hyperscale Operators

When it comes to the critical development efforts for 400GbE, the big-four hyperscale data center operators do not appear to be doing much more than going through their usual motions. Despite their undeniable role in the destruction of the optical ecosystem, they seem inclined to believe that the mere expectation of high revenues from this gear will miraculously result in the arrival of these highly complex devices relatively soon. Incredibly, despite the arrival of 100G electrical possibly being as many as five years away, Amazon apparently expects to see 4x100G by the end of 2018. There is just no obvious sense of urgency by these buyers as it looks like their short-sighted goal, after they became dominant purchasers, to drive down devices to razor-thin margins, will result in about 1.5 large component vendors in the US remaining in the space. The longer these users wait to make the transition in their priorities to figure out a way to provide the necessary capital to ensure the indispensable development efforts, including re-creating a viable ecosystem, the harder it will be for them to ensure they are getting supplied with the required capacity in the coming years. For examples, optical engineers do not have a clue as to how they will pull off 200GbE in the distant future.

Although we are aware that there have been at least a couple of optical componentry suppliers, which have received funding from Google and Microsoft, in general, it has been the exception rather than the rule. It is also hard to imagine that these operators will get bailed out by the system vendors, which are having their own problems with getting margins higher, by acquiring these firms lower down on the food chain. There is also the matter of Oclaro, which has taken on a lot of the system engineering role at the chip level, being eventually moved at least somewhat in a new direction by its acquirer, Lumentum.

Just looking at Finisar’s situation, the historically largest optical piece-part provider to the data communications space is instructive. Its revenues and profits have dwindled down to precariously low levels. The new management will be very selective in targeting only the most margin-rich opportunities in its traditional sector moving forward, as opposed to continuing to be an all-encompassing, vertically integrated company.

An entire cultural and organizational shift will be required by the hyperscale players. They will have to become more interactive with the vendor community, not just engage with them when there is a problem.

Secondly, they should abandon all of the extracurricular activities concerning networking in favor of focusing their resources on the practical matter of ensuring timely development of higher data rate gear. In addition, although they all have their share of talented engineers, they will need to bolster their in-house talent with individuals, who are more experienced with transceivers and lower-level devices, particularly chips.

For example, Google can fully stop pretending it is interested in providing fiber to the home as well as moving its attention away, at least for now, from turning telecom central offices into edge computing centers. Microsoft should abandon the COBO program, as it evidently has little to no viability at this time, and stop overly obsessing on insignificant matters, such as resisting the movement to multimode fiber/VCSELs, when it makes economic sense for shorter connection distances.

Regarding Facebook. it should cease disingenuously professing its wish to help out the ISPs with its TIP program, as well as only push co-packaged optics, if it is certain that the standard optical/electronic hybrid solutions cannot do the job to achieve the 400GbE speed. On Amazon, while the company’s low profile in the cloud equipment realm was previously refreshing, it has ostensibly been too far out of the loop in getting a realistic view on the pace of change (or perhaps more generously, it has to readjust its wishful thinking) – and Amazon’s exposure to the outside world should not just be confined to leaks of supposed entry as a direct competitor of equipment, if nothing else, to get pricing lower.

As always, fibeReality does not recommend any securities, and this writer does not invest in any companies being analyzed by us.

To follow us on our totally separate, quick update company blog, which is exclusively on fibeReality’s LinkedIn page, please click here.

[written by Mark Lutkowitz]