Notwithstanding the title of this article, a good number of optical suppliers, which want to do business with Microsoft, are likely to be too wary about totally ceasing resource allocation to both the Co-packaged Optics (CPO) and the Consortium for On-Board Optics (COBO) concepts. fibeReality understands that the head of COBO, Brad Booth, no longer reports to the top cloud infrastructure executive, Jeff Cox, who is obviously pushing the CPO effort. Unsurprisingly then, we have also just learned conclusively that COBO will not be deployed at Microsoft. Concerning CPO, which of course, also has the support of Facebook, it is the holy grail, and we started addressing the major technological challenges over a year ago. However, even though at ECOC 2019, one or more speakers entertained at least the partial employment of CPO at 51.2 terabits per second, our confidence level in the use of pluggables has risen greatly, even for the highest capacity switch under active discussion in the marketplace.
In fact, the top engineers in the industry are certain that they will be able to avoid the need for CPO up to 51.2T. While 102.4T is not currently ready for a pluggable solution, by the time of the arrival of that speed, it would be hard to believe that a single innovation required in the interconnect for the printed circuit board assembly will not have been accomplished. Naturally, just questions about the return on investment make CPO unattractive to vendors anyway.
Cisco Systems’ Bill Gartner, SVP/GM Optical Systems & Optics Business Unit, has been accurately talking about at least two generations for a CPO need. It gives even more weight to our ultimate conclusion that the Luxtera acquisition was principally defensive in nature in keeping it out of the hands of Broadcom, and to neutralize the perception that silicon photonics would become a significant factor in the business.
Thus, contrary to suppositions in the space, we suspect there is an absence of truly serious CPO R&D efforts at Cisco, although we doubt it will ever get to the extreme of the Chairman of Arista Networks, Andy Bechtolsheim, in totally reversing his perspective. Also, Cisco has other important priorities, including in going head to head with Broadcom on switch chips.
For all intents and purposes, especially given the expected schedule for the high-volume entrance of 400GbE, which will have the IEEE standardized PMD sublayers, and be utilized in up to 51.2 switches, even accurately projecting the use of pluggables in a 102.4T switch is impossible, let alone the next-gen device, which could hypothetically be the first definite candidate for CPO. Therefore, any development considerations concerning such a switch are so far out as to be categorically deemed inconsequential.
Returning to COBO, our intelligence points to Microsoft having solved the problem that the scheme was to address internally by just outsourcing the datacenter rack assembly. We have also heard that there has been a frantic effort at the consortium to revise the historic record of the MSA to make it seem that it was a stepping-stone to CPO right from the start, in order to remain plausible within the hypersclaler. Such an alteration would be a neat trick to pull off given that at best, a redefinition late in the game (which would have apparently applied only to that point in time and moving forward) appeared to be the only previous option.
Incredibly, as if it had not been a sham for quite a while, a plan has been devised to transform this walking zombie of an optics model into a CPO rival to stay pertinent in general. How would that be in Microsoft’s interest and how much longer will the company allow this debacle to continue before it has a potentially deleterious effect on its brand?
COBO still has several hundred thousand dollars in the bank to keep itself going for quite a while (partially for the very nice social events with an excess amount of food), including allowing it to organize an event at OFC 2020. Is there at least any concern at Microsoft about the possibility for conflicting messages being delivered at the show, and will there be a repeat of what happened at ECOC 2019, in which there was the appearance of indecision on the exact focus of the presentation?
Additionally, we had found out that there was the prospect at the COBO meeting in concert with the last ECOC that in light of Microsoft’s clear direction toward CPO that the consortium could have decided to disband and return the remaining money to the membership. Apparently, a few individuals active with the organization and representative of the key technologies were privately consulted in advance. It would have been the right decision because this is another case in which there is absolutely no longer even a legitimate pretense for saving face.
For our fast-growing, totally separate, quick update blog, which is exclusively on fibeReality’s LinkedIn page, please follow us here.
As always, fibeReality does not recommend any securities, and this writer does not invest in any companies being analyzed by us.
[written by Mark Lutkowitz]