fibeRealty has speculated in the past about Infinera doing a spinoff of its DSP/PIC combination. However, given the current intensity in the consolidation of semiconductor vendors, and the supplier’s extremely low valuation, a chip supplier might consider buying the entire company, and then either manufacture-discontinue or sell off everything, but the ICE-related components. Unsolicited, the next CEO at Infinera, David Heard, volunteered the following during the last earnings call: “I think recently, we’ve seen the demand for optics, DSPs have a, let’s say, scarcity and a very, very high multiple in terms of the demand, as you’ve seen by the purchase of Acacia and the recent purchase from Marvell or intent to purchase Inphi. And so we think it’s a very, very good place to be.” He then also stated: “It’s good for our business model, good for the customers.” Heard also seemed to be quite happy during the discussion that seemed to go beyond getting his chance to finally run the show, and perhaps was indicative of a hope of finding a satisfying way out of competing in a very difficult optical systems market. Without the current merger craze, Infinera getting bought would definitely be considered a long shot right now in the same way as President Trump being able to remain in power.
A possible purchaser would be inclined to assume that Joe Biden will be the next president of the US, which would likely facilitate a quicker decision by China to approve the Cisco Systems-Acacia marriage, totally removing the latter as a merchant vendor. While Inphi has been expected to offer limited 800G capability for HPC applications to Google because of the MPI-mitigating patent owned by the former, we do not foresee the supplier providing full capability at that data rate until 2025. NEL will likely remain behind Inphi on 800-gig development as well.
Ciena’s Canadian executives will always stay box-centric and will have no desire to support a healthy “duopoly” situation projected with Infinera. A buyer of Infinera would have a complete monopoly at that 800G rate in supplying full optical system suppliers, presumably including Chinese firms (if restrictions are relaxed in a Biden administration). Once again, the international, antitrust approval process in this situation would not be anticipated to be held up, with the additional advantage it will not hit close to home in China, as with the Acacia announcement, regarding the impact on ZTE.
Certainly, we do not think an acquirement of Infinera would even be close to comparable to the pervasive negative reaction by the announcement of the Cisco pending acquisition, as it had immediate ramifications, such as for the webscalers, because of the effect on several providers of optical systems. Actually, while we see a gradually declining market in the hyperscale space for these transponder-based, optical transport systems, as these cloud operators migrate to pluggables, such as 400ZR, as another reason for avoiding the buyout of Infinera, the cloud operators in the meantime might consider purchasing the Infinera PICs themselves and working with white box suppliers, to design a less costly, and even more importantly, a lower power-consuming solution, particularly as 800ZR is hardly around the corner.
There is also the chance that Nokia Networks would reconsider its avoidance of 800G and be a customer as well. Nokia may be less hesitant because of the elimination of a direct competitor, and the ability to fully complement its wireless and router product lines. Although it would generate a higher margin through vertical integration, it may not be worth the investment in a DSP so late in the game.
Of course, it still begs the question on how many 800G-capable suppliers there needs to be on the line side to serve the world’s needs. There is the possibility of any vendor failing to get business in some situations, if they cannot at least offer the capability — not to mention, it being unnecessary to start out at the very high end. There is also the OLS and its ability to support third party and alien 800G channels when they have a programmable, flex grid filter architecture that will accept 800-gig operating specifications.
Naturally, there would be negative consequences to end users with an installed base of Infinera’s products, even though any customer that has stuck with the vendor understood that it was being done with some risk. These customers would obviously not accept such a new strategy, although the sense of abandonment might be ameliorated by vendors offering the same PIC/DSP capabilities, and bending over backwards to make the transition to their systems as easy as possible. Nevertheless, the problem of such buyers immediately looking to cap the growth of Infinera products, and then just moving on to say, Ciena, especially during the closing period, may not be avoided.
Still, the most compelling argument against an Infinera purchase is the big burden on the chip company to understand, support, consolidate and divest the myriad products just to essentially gain an 800G chipset. Infinera has legacy routers, 7100 systems, Groove devices, Sycamore switches, and Transmode gear, along with the XTM, DTN, and Cloud Xpress products.
While fibeReality has not been that bullish about Infinera’s XR concept, it still might intrigue a potential purchaser, which may be convinced that there is a longer-term advantage in owning Infinera’s PICs. Additionally, Infinera has nothing to lose by kind of just indirectly putting the idea out there during this aberrant time of almost desperation with chip suppliers concerned about their future competitive positions. The most unusual moves can be made, when the stakes are so high.