Introduction
The 1.6T Ethernet standard is entering the home stretch. The IEEE P802.3dj task force — launched in 2020, with a formal study group standing up in 2021 — has now reached the ballot stage, the last procedural hurdle before official ratification expected before the end of 2026. Yet the ink isn't dry yet, and early products are already in the channel. A multi-vendor interoperability showcase is running this month in Málaga. For infrastructure leaders, the time to factor this trajectory into network roadmaps is now.
From Lab to Deployment in Five Years
The story of this standard is a useful reminder of how fast the network innovation cycle has compressed. Less than five years after initial work began, 1.6T direct-detect modules are already shipping to first buyers. Designed for links reaching up to approximately two kilometres over single-mode fibre, this technology targets intra-data centre use cases: spine-to-spine switching, high-density compute servers connecting to storage arrays, and chassis interconnects within a campus.
P802.3dj is not a single-speed standard. It simultaneously covers 200G, 400G, and 800G variants, establishing a coherent framework for the full Ethernet speed spectrum that will roll out over the coming years. The chip-to-module interface defined as 1.6TAUI-8 plays a pivotal role in this multi-speed architecture, ensuring silicon and optical module compatibility across the stack.
Coherent Optics Takes Over at Long Range
Where direct-detect reaches its limits, coherent optics picks up. In 2026, the OIF (Optical Internetworking Forum) published the 1600ZR implementation agreement, enabling 1.6 Tb/s transport over a single wavelength at distances up to 120 kilometres — double what the previous 800ZR architecture supported. In practice, that means an existing DCI (Data Center Interconnect) infrastructure built on amplified DWDM links can double its capacity by swapping out modules, with no changes to fibre plant or amplifiers.
The 1600ZR specification is complemented by a leaner variant, 1600CL, targeting short-range WDM links of around ten kilometres, and a future long-haul version called 1600ZR+, designed for spans up to 1,000 kilometres. This three-tier architecture gives metro network operators and large multi-site enterprises a consistent, single-family solution set across deployment scenarios.
Málaga, September 2026: Interoperability Under Real Conditions
Standards announcements are one thing; industrial reality is another. That gap is exactly what the Ethernet Alliance interoperability demonstration at ECOC 2026 is designed to close, running from 21 to 23 September in Málaga. Around a dozen manufacturers — including Cisco, HPE, Huawei, Amphenol, Synopsys, TE Connectivity, and Keysight — are running switches, OSFP modules, and 1.6T cables from different vendors together, using 224G-class SerDes and link training mechanisms. The demonstration also covers lossless Ethernet features — Link Layer Retry and Credit-Based Flow Control — both essential for HPC and AI compute clusters where even brief network congestion produces measurable throughput degradation.
A Roadmap That Doesn't Stop Here
While 1.6T has yet to receive its official ratification stamp, IEEE is already scoping the move to 3.2T, with work tentatively scheduled to begin in early 2027. That cadence demands that infrastructure teams think in successive technology generations rather than one-off capital purchases. Any investment made today in 800G or 1.6T equipment should be evaluated on its upgrade path — module-level refresh rather than chassis replacement — as speeds continue to climb. The question is no longer whether terabit-class speeds will become standard across enterprise networks. It is when to build that transition into the budget cycle.