The Hidden Cost of the Wrong Optic: TCO Beyond the Price Tag [4/6]
The purchase price of a 100G ZR module is higher than the 10G SFP+ DWDM transceiver it replaces. That comparison is true and tells you very little. TCO in access networks is determined by five cost dimensions — only one of which is hardware price. This paper builds the complete model.
Why Unit Price Is the Wrong Starting Point
Access procurement teams instinctively treat the price delta between a 100G ZR coherent module and the 10G SFP+ it replaces as the cost of the upgrade. But this is not the complete story. Unit price captures one dimension of a five-dimension cost problem — and it is the dimension where coherent performs least favorably. The lowest-cost network is not the one with the cheapest components, but the one with the fewest layers to procure, manage, and eventually replace.
This paper models three migration paths across five cost dimensions over five years, in normalized form so the methodology is transparent. The scenarios where 100G ZR is not the right answer are stated explicitly — because an analysis that ignores boundary conditions is advocacy, not analysis.
|
3.6x More efficient per Gbit vs. 10G SFP+ tunable (power per bit delivered) |
25% 5-yr power cost vs. 10G baseline per 100G capacity unit |
$0.13 US commercial electricity per kWh (April 2026) up 27% since 2019 |
Power Is a Line Item — Measured Per Bit Delivered, Not Per Port
A tunable SFP+ 10G DWDM transceiver draws typically approximately 2W (Lumentum datasheet TRS70xxFxCPA000-x, 2020). A 100G ZR QSFP28 coherent module draws less than 6W. On a per-port absolute basis, coherent draws more. That comparison is architecturally meaningless. The relevant metric is power per Gbit delivered. Ten 10G SFP+ ports are required to deliver 100G of aggregate capacity: 10 × 2W = 20W. One 100G ZR port delivers the same 100G in a single QSFP28 slot at under 6W — a 3×+ improvement in power efficiency per bit, at identical port density, with nine freed QSFP28 slots for additional services.
Two factors compound the advantage. Every watt consumed in an access enclosure must be cooled — a PUE of 1.5× is appropriate for air-cooled street cabinets and aggregation enclosures. Multi-port 10G deployments frequently hit the thermal ceiling of these cabinets before the capacity ceiling; a single coherent port does not. Additionally, US commercial electricity averaged $0.13/kWh in April 2026 (PowerOutage.us, April 2026), up 27% since 2019 (EESI/EIA, 2026). Power cost is a rising line item, not a stable planning assumption.
Figure 1: 5-year power cost per 100G capacity unit — 10 × 10G SFP+ tunable (legacy) vs. 1 × 100G ZR QSFP28 (<6W). PUE 1.5× cooling included. $0.13/kWh US commercial rate (April 2026). Sources: Lumentum TRS70xxFxCPA000-x datasheet (2020); PowerOutage.us (April 2026). Arycs Technologies analysis..
| “Every watt we can save is good.”
— Mattias Fridström, VP & Chief Evangelist, Arelion (Cisco case study, 2025) |
The Forklift Decision — Strategic When Deliberate, Expensive When Undermodelled
Forklift upgrades are not always mistakes. Operators sometimes choose them deliberately — for vendor consolidation, lifecycle alignment across a node cluster, or an operational reset that clears years of accumulated technical debt. When the rationale is strategic and the cost model is complete, a forklift can be the right answer. The problem arises when the cost model is incomplete. In multi-vendor brownfield access networks, operators consistently report that EMS and NMS re-architecture — bringing a new equipment generation under management — runs 30–50% of hardware cost. That figure reflects real integration complexity in environments where access nodes from multiple vendors share a common NMS, often customized over years of incremental configuration. It does not appear on a hardware price list.
A common and expensive pattern: operators overbuild 10G parallel capacity because each link appears incremental, yet the cumulative result locks in the highest long-term cost structure — full 10G management overhead, full power draw, zero capacity headroom improvement. The forklift then becomes reactive rather than strategic, triggered by a crisis rather than planned. On the coherent side, operators should also account honestly for what changes when pluggable coherent arrives at the access edge. Tunable coherent modules require wavelength provisioning on the DWDM grid — a workflow that 10G access teams often do not have. Troubleshooting OSNR, chromatic dispersion margin, and FEC pre-FEC BER is a different skill set from diagnosing a direct detect link. Phase 1 pilots should budget explicitly for this learning curve alongside the technical validation.
The TCO model below applies specifically to the access edge tier where 100G ZR is architecturally appropriate: mobile backhaul rings in the 80–200km range feeding aggregation hubs, rural aggregation spans connecting remote nodes to regional PoPs, and metro edge consolidation links where operators are replacing multiple 10G DWDM wavelengths with a single 100G coherent channel on an existing open line system. These are the topologies where chromatic dispersion has already foreclosed PAM4 direct detect, the installed QSFP28 base is dense, and the incremental migration approach delivers the most pronounced TCO advantage.
The Honest 5-Year TCO: Three Paths, Five Dimensions
Three paths modelled over five years. Normalized to 100% = one 10G SFP+ module at Year 0 list price. Five cost dimensions each.
Path A — Status quo with parallel capacity workaround. Continued 10G operation plus augment links. Capex 150%, power and management at full 10G baseline. Capacity problem deferred, not solved. Total: 450%.
Path B — Forklift upgrade. Full chassis replacement. Hardware capex 180% once installed-base write-offs are included. Power drops to 28%. Management adds 60% (NMS re-arch, retraining, cutover risk, parallel running). New sparing at 80%. Depreciation risk 30%. Total: 378%.
Path C — Drop-in 100G ZR coherent migration. Module capex 100%, no installed-base write-offs. Power 28%. Management 20% (SFF-8636 compatibility preserves NMS continuity in Phase 1; CMIS capability added on host’s own upgrade cycle). Single-SKU sparing 30%. Zero depreciation risk. Total: 178%.
Figure 2: 5-year TCO across three migration paths — five cost dimensions each, normalised to 100% = 10G SFP+ module capex at Year 0. Sources: Lumentum; Adtran/Coherent; Arycs Technologies; PowerOutage.us (April 2026). Arycs Technologies analysis.
Path C is economically compelling across most brownfield access scenarios in the 80–500km tier. Two honest caveats: for very short links where PAM4 direct detect is adequate (below ~40km on O-band infrastructure), the 100G ZR price premium is not recovered by power savings. For greenfield builds with no installed base, forklift economics reset. Neither exception changes the conclusion for the bulk of the 10G SFF DWDM installed base where chromatic dispersion has already foreclosed the direct detect option.
| “10G DWDM is widely deployed at the edge of the network, but carriers are now searching for higher bandwidth 100G solutions. Low-power QSFP28 is required at the edge, and the 100ZR solution is well-positioned to be the format of choice for such upgrades.”
— Scott Wilkinson, Lead Analyst, Cignal AI (December 2024) |
The Business Case Is Clear — Within Its Scope
In most brownfield access scenarios at 80–500km, the financial case for the drop-in 100G ZR coherent migration path is consistently favorable: roughly half the 5-year TCO of the forklift alternative, less than 40% of the status quo trajectory, with power efficiency, management simplicity, sparing consolidation, and zero depreciation risk combining to produce a structurally durable result. The caveat stated earlier is real — greenfield builds and short-reach links change the comparison — but they do not represent the bulk of the installed base that operators are actually trying to upgrade.
What determines adoption speed from here is less the economics and more the operational readiness of management infrastructure. The CMIS telemetry, adaptive FEC, and zero-touch provisioning capabilities embedded in these modules are only fully accessible once the surrounding EMS and NMS have been modernized to CMIS-capable platforms. Paper 5 examines how those intelligent pluggable capabilities feed into the AI-managed network architectures that operators are building toward.
#TCO #PowerEfficiency #BandwidthPerWatt #DataCenterInfrastructure #PerformanceWithoutCompromise
About Arycs Technologies
Arycs delivers power-efficient, coherent-class optical connectivity based on silicon photonics, coherent DSP, and advanced optical architectures. Our solutions provide industry-leading bandwidth per watt, deterministic performance, and flexible network evolution for AI, cloud, telecom, and edge infrastructure. Designed for real-world deployment, Arycs Technologies enables networks to scale with growing AI demand without disruptive redesign or hardware replacement.
arycs-tech.com | LinkedIn: Arycs Technologies
The Hidden Cost of the Wrong Optic: TCO Beyond the Price Tag [4/6]
