Broadcom has appeared solidly behind its public engagement with co-packaged optics, developing switches with optical I/O. In fact, the supplier seems to be pulling out all of the stops in at least leaving the impression that it is organizing the entire firm around CPO. There are also potential blueprints to compel customers to purchase the full CPO solution in order to obtain the switch ASICs. fibeReality is unaware of another sizable corporation that has ostensibly made such an immense commitment to this networking concept as its core strategy. One explanation that we have heard is that SVP and GM, Alexis Black, has been able to persuade CEO, Hock Tan, to adopt this new, all-encompassing vision, despite the latter always being adamant about high profitability. We cannot believe that Tan, who has a meticulous understanding of the marketplace would buy into the CPO hype, which has close to a zero chance of adding meaningful revenues to his balance sheet in any kind of foreseeable future. Just the complete alterations/challenges in operations, pricing, servicing, etc., alone help to exclude the possibility — never mind the lack of a clearly defined purpose for CPO. Then there is the very long-lasting promise of pluggables with the relatively high confidence level in getting beyond any future obstacles (not to mention one should always avoid betting against copper). Therefore, there must be another explanation for this perceived massive effort by Broadcom. At a minimum, there must be a backup plan, which likely included Tan telling Black that she has been given a fixed amount of time to make this scheme work. In the meantime, there will continue to be demonstrations, such as with the half-CPO switch at ECOC.
Abstractly, such an effort towards CPO looks fantastic. In addition to not only continuing to lock in the hyperscalers to a greater extent, Broadcom could now take hold of optical expenditures as well. In other words, the optics integrated would give it a much larger revenue stream for the same number of switches shipped.
The reality is that just the impression of Broadcom positioning itself as by far the dominant force in CPO, is the ultimate defensive move, along with the justification of the ludicrous amount of spending on packaging, systems, etc. by the company on this idea. The stakes in the supplier maintaining its position in high-end switching are incredibly high for two major reasons: 1) it is a major cash cow and 2) the immense amount of psychological warfare in the overall data communications networking space.
Regarding both points, Broadcom wants to preempt a comparable heavy investment by any other established supplier in CPO by being viewed as the clear, established leader in R&D activities. No large buyer will be able to accuse the supplier for doing the bare minimum before Broadcom declares it an unreasonable expectation.
At the same time, no other big vendor will have the necessary credibility to walk away from these highly expensive activities when it is most optimal — whether it be the arrival of the 51.2T switches (it is 100 percent a done deal for pluggable optics because of the market window and there is no question in our mind that 102.4T will fail to be a maybe for CPO) or if there happens to be a few consecutive quarters of rather disappointing financial performance. In the interim, Broadcom can control the CPO narrative in ways that are most beneficial to itself.
In our opinion, there can be no other legitimate explanation for Tan, who has always played hardball when it comes to margins being all-important. It would simply be irrational for him to play the role of a VC-funded startup, in purposely going after a science-fair type of project that would play out over a tremendously long period of time. Conversely, it should be considered a vital downpayment on helping to ensure a successful future.
It would additionally be quite baffling for Tan to be serious about CPO when the technical and operations leaders of the big four hyperscalers, Google and Amazon respectively, have made it clear that they lack a need for CPO. Moreover, with it being so early in the product development cycle, it would be astounding to believe that he has received an endorsement from a webscaler, along with the promise of a reasonable return on investment.
For one thing, we think Microsoft’s optical networking team is in a state of flux right now anyway. That leaves Facebook, which has the least need of the big four operators for CPO from a sheer capacity standpoint based on its business profile.
At the end of the day, the hyperscale data centers do not want CPO on the switches because that is one fewer product they can negotiate with and leverage when integrated with the switch silicon module. Recently, we have become more optimistic that there is over a 50-50 chance even at 204.8T that pluggables will remain viable.