fibeReality, LLC

After his experience as a speaker at the OFC 2024 Rump Session (please see: https://lnkd.in/e-RP_cMf), the high-level executive from Lightmatter may decide to skip any future opportunities involving a technical audience (please see: https://lnkd.in/e7ZyJaf6). Tough questions were asked including on the viability of the applications targeted by the startup and whether Google Ventures (GV) ever asked the optical networking team at the hyperscaler for its views on the supplier. The answer provided to the second inquiry implied that it was irrelevant because a top technology exec from the hyperscaler had been hired by the supplier. However, that individual mentioned in the response does not appear to have a lot of intimacy with optics and seemed to have only peripheral participation in the design of TPUs at the cloud services provider.

Could Applied Optoelectronics be an acquisition candidate (please see: https://lnkd.in/eV45Wxne)? Initially, the answer may have been toward the negative side because it has no existing 800G business, yet. However, given that it has access to Microsoft’s intellectual property involving active optical cables at that data rate, it may have a real competitive advantage. If that is the case, then it is theoretically possible that even a market leader in the business, such as Coherent (please see: https://lnkd.in/eekhYmqC), could be interested in a buyout. While the Street is increasingly becoming attracted to the optical space in general, a way for an existing player to substantially boost its market cap would be to get directly into the 800G and higher datacom module business (please see: https://lnkd.in/ebZRuYAH). The technical team at AOI which appears to be quite competent could be given a fresh start at large customers that turned away from the company in the past. Also, AOI has a really attractive valuation (which only became more so within the last day or so) and would not have the challenges in buying out other established transceiver suppliers, especially by entities in the west, such as a Chinese public firm, like Eoptolink (please see: https://lnkd.in/emCf82Dm).

On Lumentum’s earnings call earlier this month (https://lnkd.in/eCHAbvqg), a couple of matters, one ignored, and one addressed, potentially reveal other implications for the supplier. Unlike at previous conferences, starting with the one related to the acquisition of Cloud Light, in which there was discussion by the supplier on optical switching, undoubtedly connected to Google (please see: https://lnkd.in/emqFuy3G), the topic was totally disregarded this time. A reasonable conclusion is that given the highly sensitive nature of the work that Lumentum is providing to the hyperscaler on its OCS, that the important customer may have been perturbed with the attention on this topic by the corporate leadership at the manufacturer – please see: https://lnkd.in/eNu3ASTk. (At least, fibeReality was able to discover that the pretense of a “secret” relationship (please see: https://lnkd.in/ehWnRdVr), did not prevent one investment analyst from Barclays, making a direct connection publicly between Cloud Light and Google, as pointed out at https://lnkd.in/eCjzY_qc on the 9th of this month.) Then, although 3D sensing for consumer applications was brought up by Lumentum in the usual, cursory way, it did go to the trouble of stressing guidance (albeit unimpressive) for the next couple of quarters. According to fibeReality’s latest intelligence gathering, there could be a sign of at least a limited bright spot, for the two incumbent facial recognition chip vendors to Apple, including Coherent (please see: https://lnkd.in/eNFxi9cR), resulting from the price war. Pricing has dropped to the point that high yield is essential to ensuring profitability, which would only result from historic intimacy with the design/manufacturing of the solution. Moreover, we understand that just those vendors with that experience would be able to take advantage of lowered cost through less inspection for quality at the customer (please see: https://lnkd.in/eVYVMHxF). We have also learned that once there is a dramatic alteration in the module design, then that significant barrier to entry for a third player, apparently like Sony, would be eliminated (please see: https://lnkd.in/erSwYjTw). At the time of the publication of this post, fibeReality was unsure whether the expected changes in 2026 with Apple’s phones would fall into that category. 

Lumentum is expected to implement further acute cost controls, although it is uncertain how exactly they will be carried out (please see: https://lnkd.in/ehWnRdVr). The current financial strategy of the corporation is to lower expenses on labor and travel, as well as employee support, as it focuses outlays on acquisitions and stock buybacks. There are already restrictions on Lumentum’s engineers attending OFC 2024 (please see: https://lnkd.in/eJSJZNSH), as approvals for some of them will allow for day trips to San Diego with no reimbursement on staying at a hotel overnight and/or attending the show virtually. (One would hope that CEO, Alan Lowe, will pay out of his pocket, if necessary, just out of principle – please see: https://lnkd.in/eDUiz5Kc). fibeReality also understands there has been a policy in place for a while at Lumentum to convert bonuses to restricted stock units, starting at the director level, with other workers having the choice to get cash (although it remains to be seen what will occur this month). Suffice it to say, that the firm does not appear to be positioned right now to fully take advantage of the optics boom in 800G transceivers occurring in the market (please see: https://lnkd.in/eG7YabZK). The special, “captive” relationships between the transceiver supplier and the customer, such as Google’s new relationship with Lumentum, frequently have not meant that the vendor will receive the largest share of the business at the buyer. These suppliers tended to be rewarded in other ways in making the user’s development needs a priority (please see: https://lnkd.in/e3xqpTWZ). With Google as the leader of the hyperscalers in technological advancements in the optical equipment space, there will inherently be an assumption of greater costs at times by Lumentum – or as Alan characterized it on the last quarterly earnings call, “that's on our dime” (https://lnkd.in/eCHAbvqg).

fibeReality has learned that in a presentation at Photonics West by Yole, it stated that although Broadcom's co-packaged offering remains available, the supplier apparently views its solution as an end-of-the-decade opportunity. We totally believe that note from the firm, just because it has a negative connotation for one of its potential customers. Despite the loads of money invested by Broadcom into CPO, fibeReality never thought there was an earnest commitment to the technology (please see: https://lnkd.in/eaJQJNQ3). In addition, we have remained steadfast in our negative outlook on the concept in general for a long time (please see: https://lnkd.in/eeXSDgZY). fibeReality is now unsure what the gentleman from Broadcom will totally address at the Optica Executive Forum at OFC 2024 (please see: https://lnkd.in/e2GCudV9), other than to apparently state that there will supposedly continue to be a waiting process. Yet, large customers, such as Google, will be relieved to know that even a greater amount of attention will be given to pluggables (please see: https://lnkd.in/eQ53p6zQ). Furthermore, we suspect that the recent closing of the acquisition of VMware, in placing everything further into the proper perspective, was a factor in Broadcom’s decision to effectively end the charade on CPO.

Last decade, it would have been hard to remember corporate executives at the highest level of Nokia (please see: https://lnkd.in/eFr3s29R) characterizing optical networking as “obviously very important,” as it did during the session with investment analysts as part of a Progress Update this past December (https://lnkd.in/ewwUvy8Y). According to the vendor on that conference call, a reassessment starting in 2021 led to taking a decision to “double down on R&D in this area.” Nokia has certainly remained one of the main DWDM players. The supplier has been able to retain very capable individuals from Alcatel-Lucent (including the ex-Lucent portion in Germany) and many of its talented optical engineers in Europe are expected to remain with the corporation indefinitely. They consistently design competitive products for the traditional operators. With its IP routing group considered one of finest in the world, both businesses will be able to continue to piggyback sales off one another. Of course, Nokia’s strength in telecom access networks (originating with Alcatel’s PON dominance), will likely be especially important, as the tremendous upsurge in AI traffic would be expected to particularly affect that specific market segment. Also, fibeReality likes the fact although for the longest time, there have been unreasonable expectations on the incumbent service providers shifting to a focus on optics at the IP layer (please see: https://lnkd.in/eaiuzEJe), Nokia is apparently unapologetic and unembarrassed about being out-front with its emphasis on an OTN switch/architecture (please see: https://lnkd.in/eCEFWN-J).

Our recent post on Jabil (please see: https://lnkd.in/eUR6qGP3) was a reminder that the terms were not publicly provided for Intel’s divestment of its SiPh transceiver line. Is it possible that Jabil received the assets for free? Maybe it was not a direct cash transaction. In 2015, we understood that when Avago (now Broadcom) sold off its module business to Foxconn, the latter bought components from the former at a premium price until the agreed upon value was fully paid off (please see our article at the time, “Avago Optical Assets Sold, Pointed Out by FoxConn PR -- https://lnkd.in/ewRPPySZ).

With the opportunity provided at the Optica Executive Forum on interconnects (https://lnkd.in/eWUcS9DS) at OFC 2024 (please see: https://lnkd.in/e8GW65v8), Nubis Communications, should strongly consider pivoting on its marketing strategy to focus directly on its intellectual property assets related to vertical line cards (please see: https://lnkd.in/eFRXuwvE). While at ECOC 2022, a technical presentation sufficed (please see: https://lnkd.in/eiWetzCX) as a trial balloon when still in stealth mode, to try to stimulate discussion in the market, the startup needs to become more assertive now in taking advantage of the increasing amount of industry attention in the direction of 200G SerDes/100T switches (please see: https://lnkd.in/e5jttmfR), to promote the notion that pluggables will be further enabled (please see: https://lnkd.in/eQ53p6zQ). Otherwise, there could be a missed chance in potentially appearing to be just another me-too silicon photonics player (please see: https://lnkd.in/eHnHbAFv and https://lnkd.in/ehbdeQMY). In the same way that Peter Winzer, Nubis’ CTO and founder, has used his influence in advocating for new technological directions in the past, he needs to take the lead role now at the supplier in general in articulating the compelling case for VLCs (please see: https://lnkd.in/er9jEdZQ), based on his personal idea encompassing the unique placement of the ASIC. His direct involvement would be more apt to facilitate the eventual public support of large players, including prominent users in the industry, while hopefully resulting in gaining additional attention from investors, despite their natural tendency to focus on the bleeding edge.

In returning to the OFC 2024 workshop on generative AI and infrastructure operations (please see: https://lnkd.in/eaiuzEJe), regarding the big traditional operators, “automatically generat[ing]…RFP…documents by identifying the network's needs and matching them with available technologies” would certainly be feasible. It would be easy enough to automate the process of customarily rejecting a vendor totally for responding “No” to any question concerning whether it possessed certain feature/functionality. (The correct answer is always that the supplier will have that capability by a particular date.) Another easy enough task for such operations-centric service providers (please see: https://lnkd.in/e8GW65v8) is to substantially favor incumbent equipment vendors by automatically starting them with higher scores. Naturally, the human element cannot be eliminated in this process as there are always specific customer attitudes and preferences (often embodied by the executive in charge of procurement), and the inclination would likely be leave them out of the official, corporate AI setup anyway. fibeReality recalls a situation a long time ago at a regional holding company in which the original Nortel (please see: https://lnkd.in/enwCQ4wv) was invited to only bring a couple of people to a meeting related to the selection of gear. The supplier apparently felt that it could not possibly give a presentation without having every aspect of its device represented, from R&D to after-sales support. After showing up in person with individuals from Nortel in the double-digit range, the conference was immediately canceled. 

In addition to our views on the Optica Executive Forum (please see: https://lnkd.in/eDUiz5Kc) as well as on the show floor programs (please see: https://lnkd.in/eUEtSuf9) at OFC 2024, we have also been looking at the Rump Session (https://lnkd.in/eKfHJXVg). The answer to the big question – how much optics does AI need? – is a historically, unmatched, huge amount at the high end (please see: https://lnkd.in/e_bQbnPE). At the same time, the opening query hardly invites the free-for-all debate associated with the event. At its best, the topic should provoke a discussion of people (companies) and money in conflict – and even occasionally, involve a thorough examination of sacred cows in the industry. We believe that the level of intensity generated last year was that it focused solely on various optical materials in competition with one another (please see: https://lnkd.in/eR_Q7s7s). It may have been better in the upcoming event to target from the start, a single subcategory listed, and then incorporate it as part of a provocative introduction to the meeting. For example, let’s have it fully out on what we believe is any wrong-headed move to abandon CMOS, while supposedly shifting to an optical computing model (please see: https://lnkd.in/eFRXuwvE).