After the initial news report that Cisco Systems was working on an agreement to acquire Luxtera, most people in the industry, who had a least an adequate technological knowledge of Silicon Photonics (SiPh), and were familiar with the system house’s history with the solution, could not makes sense of such a deal. In fibeReality’s opinion, it was about a combination of motivations. First, Cisco wants to make its hyperscale data center customers happy, particularly Microsoft, with adequate financial backing of a CWDM4 player, which was probably making little to no money at best, and as the large optical component firms have been running from the space with the breakdown of the ecosystem. As a result, in paying such a high premium, Cisco’s move will stand out even more as a loss leader in the same way it does for Intel. Secondly, it is a defensive strategy for the former in that the vendor has obtained the most conspicuous supplier of SiPh, in order to deal with any questions that come up about the technology’s hyped use in both co-packaged optics and in support of 5G wireless. Nevertheless, a number of executives and academics directly working with SiPh, along with investors, reporters, and analysts in the data communications space, have or will point to this purchase as a seal of approval of these chips. However, there are four major reasons to dispute such an assertion:
1. SiPh is on its last legs;
2. one would find it difficult to find even a passing, public reference, by Cisco about SiPh for well over three years (not to mention the supplier’s lack of emphasis on optics in general) before the announcement of the buyout;
3. the press release being issued only a few days before the holidays could not have been a coincidence; and
4. Cisco had clearly made a concerted effort to move away from the SiPh technology, as its return on investment with its previous purchase of Lightwire, had been deemed inadequate.
Therefore, the “strategic fit is [not] sound;” SiPh is not “mainstream” as well as any “offers [to Luxtera] in the past,” likely were inadequate, as the vendor would have loved to exit much earlier in struggling to be profitable; Cisco’s action is hardly “a validation of SiP[h];” and Cisco taking “a break from its software-buying spree” (in the December 18, 2018 issue of The Deal) needs further examination – as with so much behavior by companies, it should not be taken at face value.
Getting back to the Lightwave article, it is unclear that “Mellanox shut down [its] SiP[h] activities,” and even if “Nokia [Networks is] working on silicon photonics in-house,” such R&D is very unlikely to become commercialized. Most critically, a causal link has to be established between “Cisco…struggling to gain traction in the hyperscale data center market against Arista Networks,” and obtaining Luxtera, for the transaction to become logical.
Yet, optical networking remains only a small percentage of Cisco’s overall business. The long-term impact of taking a loss should be as negligible in the long term to the supplier, as it is to Intel.
In 2015, we wrote that “at the OFC Rump Session, somebody sarcastically congratulated [Luxtera] for achieving volumes that equaled the volume of VCSELs produced in a relatively short period of time.” Now, there can be derisive compliments about Cisco spending close to a billion dollars on SiPh, a much greater number than the US government has thrown away on these components. Moreover, while in effect, Luxtera “substantially increase[d] its manufacturing output at a quicker pace” in just exceeding two million photonics transceivers to date, even after shifting to a hybrid approach, and being around since 2001, one naturally wonders about inadequate yields adversely affecting costs.
Lastly, with the new year, we are getting that much closer to the 50th anniversary of the publication of the influential article in The Bell System Technical Journal in September, 1969, by Stewart E. Miller, “Integrated Optics: An Introduction” (not too long after the creation of the laser). While there has been plenty of successful photonic integration efforts since that time, the “economy…ultimately result[ing]” from SiPh has predominantly remained a huge challenge. How much longer will the cheerleaders hang in there, when the fundamental physics problem, especially with loss, will never go away?
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]
SHARE THIS