This is Cignal AI’s second ECOC 2017 brief, focused on news out of China.
Topics include:
- China Inventory and 2018 Demand
- Ciena Wavelogic AI Licensing Update
- China ROADM Deployment Outlook
- Huawei 100G+ Coherent Technology Roadmap
Our first brief, issued on October 3rd, focused on coherent 100G+ technology updates.
China Inventory and 2018 Demand
Cignal AI had extensive discussions up, down and across the industry at ECOC 2017. We attempted to stitch together a coherent picture for 2018 demand, but predictions, opinions, and data varied widely. Facts are in short supply.
That said, there were a few points on inventory conditions in China for which we determined consensus:
- CFP/CFP2 Client optical inventory is still abundant (modules and components), with up to 6 months of inventory remaining.
- Huawei is migrating to QSFP28 client-side optics. Existing CFP2 inventory may be sufficient to bridge this transition.
- Inventory is depleted for the CFP-DCO that Huawei builds internally. Companies are shipping components into this design opportunity now.
The big question is what demand in 2018 will be. The short answer is we still don’t know, and we’d demand facts and rigor from anyone who claims to have the answer. One must differentiate between what’s true and what sounds like it should be.
The intense focus on the timing and size of equipment purchasing tenders from Chinese carriers is an example of something that sounds like it should be true. This includes the hunt for the mythical Phase 13 tenders. What’s overlooked is that the tenders issued by Chinese carriers do not always represent what is eventually purchased. The only thing the tenders guarantee is market share among the awarded suppliers. They do not accurately predict volume.
In the absence of more information, our explanation is biased towards a simple and straightforward theory. For over 17 years, I’ve spoken with Chinese vendors about their outlooks, and never have I found them to be without a plan – they always have a roadmap for the year ahead and excel at meeting it. This year is the first time I have found Chinese vendors to be without a plan; a sure sign of uncertainty that could mask negative expectations. In short, we don’t think that the uncertainty in China is about how good things will be in 2018.
Ciena Wavelogic AI Modules
Ciena’s component partners see some minor traction for the 400G Wavelogic AI based 300-pin MSA. It was the attention-grabbing announcement at OFC 2017, though the company was quick to temper revenue expectations in 2018. Ciena now has moved Wavelogic AI-based equipment to production and is making progress with its component partners. It’s also hiring people from outside the company focused on supporting the component licensing business.
At this point, I expect that you are asking yourself why we have an update on the Wavelogic module in a China-focused brief. Well, at ECOC we learned that Ciena provided working alpha samples to its component partners and some of this hardware has made its way to China. Cignal AI believes both ZTE and Fibrehome now have this hardware under evaluation in their labs. These companies are primarily interested in the Ciena solution for 400G operation in shorter reach DCI applications. We do not believe that in 2018 Ciena is in a position to displace the higher volume incumbent DCO based pluggable solutions, at least not with its current DSP.
Acacia should be the next 400G DSP supplier after Ciena to reach the market. Ciena’s partners expect to deliver production solutions in 1Q’18. Ciena, therefore, has approximately a 6 to 9-month lead on Acacia for bringing 400G first to the market, though the Acacia solution is expected to be technically more advanced and likely more adaptable to pluggable solutions.
Ciena has not provided any guidance on its 400G pricing, but the larger OEMs that Cignal AI has spoken to indicate that in 2018, the MSA module isn’t attractive in volume for more than $8k for a 400G MSA module. What share of this revenue would flow through from the component partner to Ciena has yet to be determined.
The alpha-samples that Ciena just provided to its partners are slated to be replaced with MSAs designed by the partners themselves by the end of 2017. In line with the unconventional nature of this arrangement, Ciena is outlining the specifications of the module, right down to the components used inside. Cignal AI believes that each module from the three component partners will initially be both functionally and physically identical.
So far, Ciena has executed on the plan it described at OFC and has captured the attention of China OEMs. But high volume applications with the MSA module approach will likely remain out of its grasp unless Acacia’s solution experiences a delayed arrival.
China ROADM Update
Cignal AI had an opportunity to discuss the outlook for ROADM deployment in China during 2018 with Huawei, which concurred with our opinions outlined in our China ROADM Deployment Outlook issued on July 20th, 2017.
Huawei remains confident that it will deliver the MxN WSS based ROADM platform in 2018, with volume production beginning in 2019. Huawei still expects limited colorless-directionless (CD) ROADM deployment in regional and metro-core networks during 2018 in the interim. The company believes demand for contentionless (CDC) ROADM nodes using a multi-cast switch (MCS) is very limited and expects the MxN WSS based ROADM to rapidly obsolete MCS based ROADM design once it is available.
Western equipment vendors have been silent on MxN based ROADM technology. We believe they will take the same approach and follow Huawei if the MxN WSS is reliable and cost-effective. Ultimately, the decision to choose CD vs. CDC ROADM architectures will go away, as CDC could reach parity with CD, enabled by the crucial MxN WSS component.
Huawei 100G+ Strategy
Cignal AI’s 2016 ECOC report outlined Huawei’s plans to transition to pluggable CFP-DCO technology and its interest in moving to CFP2-DCO in 2018. The company planned to evaluate the feasibility of a CFP2-DCO transition in 2017.
It appears that this transition will take longer than anticipated. Huawei mentioned that having a single form factor that can address both short and long reaches (multi-haul) is vital. Until CFP2-DCO solutions can replace CFP-DCO across all applications, Huawei plans to keep using the larger CFP format.
One reason for Huawei’s multi-haul requirement is the advent of ROADM networks in China. Optical links to date, once provisioned, were rarely re-configured. Introducing ROADMs into the network changes this – when they are reconfigured the optical link must be capable of accommodating different distances.
Internally constructed CFP-DCOs are close to becoming the majority of Huawei 100G shipments. The company is pleased with the technology roadmap and performance of the CFP-DCO solution. Our belief is that Huawei is planning to move to CFP-DCO designs that can support up to 200G, and delay CFP2-DCO introduction until it can reach performance parity with the CFP-DCO form factor.
Of course, Acacia has an excellent 200G CFP2-DCO module that would fit the bill. But Huawei, at least historically, prefers to source devices internally and is likely working with its ASIC partner Inphi and external DSP supplier NTT Electronics to field a comparable solution.