Ciena Abandons Access

Ciena released their earnings on Thursday. They had a blow-out quarter, much of it tied to the AI frenzy. In the earnings call, they announced that they are effectively exiting the access business. Surprised/not surprised. Here are some immediate thoughts:

• Business 101: A new entrant to an established market can be successful only if they are able to significantly differentiate from the incumbents. Ciena’s story was innovative, but apparently not enough to gain a foothold.
• The official line is that tapering down a non-performing business allows them to divert R&D expenses to more promising opportunities in the AI space. That seems reasonable, although there is plenty of free cash flow to expand R&D.
• Ciena is retaining the current micro-OLT and ONT technology and tooling for a new mission, a PON to rationalize management traffic in large data centers. Data Center Out-of-Band Management (DCOM) is a collaboration with Meta, which is one of their largest customers.
• They claim that they are going to continue to sell and support the existing PON access portfolio. It is doubtful that they will get any new accounts, and likely that it will fade away. If the reaction to the DZS bankruptcy is any indication, their customers are now scrambling to select another vendor.
• Ciena made some risky decisions that limited their total addressable market (TAM). For example:

o   Their micro-OLT does not support GPON, which eliminates customers looking to replace incumbent vendors or migrate from GPON to XGS-PON.

o    They do not have a chassis-based OLT, instead offering stackable, modular units connected together with a proprietary interconnect; that is probably a hard sell to incumbents with big, urban central offices.

o   The economics of pay-as-you-grow micro-OLT ports in generic switch hardware worked out for networks with centralized split architectures, but for distributed split and tap architectures, high density line cards are the more cost-effective solution. Per-port pricing for the micro-OLT is said to be relatively high.

o   Part of Ciena’s strategy (and motivation for acquiring Benu Networks) was to knock down the silos between access, transport and IP switching/routing in customers’ networks, merging them into common hardware managed by a single pane of glass with a single operations team. To large incumbent service providers, that typically means organizational change of the sort that sets off immune responses in mature organizations. Ciena claimed one Tier-1 customer (unnamed but said to be AT&T) as a second source, but it is possible that this could have been an impediment to selling into other Tier-1s and Tier-2s.

o   One of Ciena’s strengths is its relationships with decision-makers and influencers in many service provider customers’ optical transport and IP switching organizations. Presumably, part of the strategy was to leverage those relationships in order to get over the usual initial hurdles to market entry. It is likely that at some larger accounts, Ciena’s friendly decision-makers and influencers in the transport and IP were not aligned their peers in access.

• Ciena/Tibit embraced 25GS-PON as a lynchpin of their strategy, a controversial decision across the industry and a risk for Ciena. 25GS-PON is, in reality, entirely a Nokia initiative, pursued outside the regular ITU-T standardization process. Ciena was the only major OLT vendor, other than Nokia, that supported it. Too much of their strategy relied on broad adoption, which has not (at least yet) materialized. The company blamed this for their inability to achieve traction.

o   Ciena’s exit puts a dent into the viability of a 25GS-PON market. Nokia is the sole-source for 25GS-PON OLTs. Tier-1 and Tier-2 operators typically require a second source, and the Ciena/Tibit 25GS-PON Micro-OLT was the only (announced) candidate.

o   Broadcom’s newly announced Higher-speed PON System-on-Chip (SOC) supports 50G upstream and downstream, as well as 25G upstream… but NOT the 25G downstream (per their data sheet). Broadcom is the only resort for OLT vendors that do not choose to develop FPGAs or custom ASICs. Ciena/Tibit’s exit leaves no merchant 25GS-PON OLT SOC in the market.

• Two years ago, a number of smaller players based new product on the Tibit/Ciena Micro-OLT. This saved them the cost and risk of developing and interop testing their own PON interfaces. As far as I know, all of them have exited the market.
• The next generation micro-OLT, with 25GS-PON support, apparently slipped its schedule by a year. It’s not clear why. Were development resources re-deployed or distracted by field issues? Did the first sample of their ASIC come back from the foundry with serious problems that required a major re-spin? Were there insurmountable mechanical or thermal problems with packing XGS- and 25GS-PON optics into the same SFP+ module?  Tibit’s team had been through several OLT SOC developments at various companies and clearly knew what they were doing, but was that kind of combo module a bridge too far?
• CableLabs is driving a new specification for a coherent 100G-PON (C-PON), ahead of longer-timeframe standardization in ITU-T, in order to meet nearer-term needs of incumbent HFC operators.  Coherent modulation/demodulation is one of Ciena’s greatest strengths, and its Wavelogic platforms are a solid foundation for C-PON product development. Ciena has apparently been the de-facto technology lead for the technical specification (architecture and requirements were driven by CableLabs’ operator members). Those people have been laid off or redeployed. The C-PON specification is in a final review stage, so their absence will not be felt quite as severely as it would have been when key decisions were being made but still poses an unexpected challenge for the project. CableLabs can hire some of them as contractors to complete the specification for publication. Beyond that, it is not clear who has both the capability and interest for prototyping and productization.
• I once thought that there might be enough room in the PON equipment market for as many as 6 or 7 vendors globally. The market apparently can’t support that much dilution. That speaks to the high cost of R&D, interoperability and pre-sales, the limited number of opportunities and the “all or nothing” nature of most tenders.
• It now looks like the PON equipment space is down to Nokia, Huawei, ZTE and Calix, with Adtran struggling to remain relevant, and a few much smaller regional players with insignificant market share. With geopolitical restrictions matrixed against the state of FTTH build-out by country working in their favor, Nokia has emerged as the clear leader. Huawei and ZTE enjoy state backing that allows them to price aggressively and offer attractive vendor financing. Calix has a large share of Tier-3, municipal, utility, overbuilders and WISPs, particularly in the very active North American region, largely on the strength of their go-to-market strategy. All of them have strong portfolios backed by large R&D investments, adequate sales and support resources in the field, and powerful selling narratives. There might be room for further consolidation, but with Ciena and DZS exiting the business, it is becoming increasingly clear that there is no viable room for new market entrants.