Bandwidth is per-lane speed times lane count. Push speed and you hit the SerDes energy floor; push lane count and you hit fiber and connector cost. The new fact this year: the second wall arrived first.

At ECOC 2025, LightCounting, NVIDIA, Coherent, OpenLight, Mitsubishi Electric and leading labs debated plasmonic, SOH, TFLT, TFLN, InP and EML modulators. The verdict: no single winner — the best balance of system, packaging and supply chain will prevail.

With 512K-GPU clusters becoming the norm for AI factories, copper is hitting its limits. An OFC 2026 panel showed how CPO delivers a 3.5X efficiency gain, and how the 12.8T XPO rewrites the switch-density game.

As copper interconnects hit their physical limits, NPO is emerging as the practical lifeline for AI compute clusters. Tencent and Alibaba plan to start 3.2T NPO deployment by late 2026, while Huawei proposes a 7.2T spec aligned with GPU roadmaps.

When does copper run out? At OFC 2026, Meta answered with 90 million device-hours of field data: CPO cuts link power by 65% and, excluding one fixable component issue, is more than 10x more reliable than pluggable modules.

A structured recap of ECOC 2025: how AI is reshaping optical communications, from 400G/lane and 448G interfaces to pluggable vs. LPO vs. CPO, coherent ZR, OCS, and the new energy-efficiency KPIs.

At ECOC 2025, Cignal AI, Coherent, Terahop, Broadcom, TFC Communication and China Telecom debated how fast 1.6T is ramping, why power and dispersion limit 400G/lane, and what it will take to reach 3.2T.

AI compute is growing 2.5X a year and electrical switching is hitting its limits. An OFC 2026 workshop showed how optical circuit switching (OCS), with its low latency and low power, is becoming a core technology for NVIDIA and the hyperscalers building next-generation AI supercomputers.

At an OFC 2025 panel, leaders from Arista, Microsoft, NVIDIA, Celestial AI and Accelink debated silicon photonics, optical interconnect and advanced packaging — and why the network has become the core bottleneck of exascale AI.

AuthenX chairman 伍茂仁 on how silicon photonics is moving into AI GPU and switch systems, the alignment and fiber-coupling challenges it faces, and why Taiwan is positioned to play a key role.

Conventional EDA is hitting a bottleneck when simulating silicon photonics. At OFC 2026, GF and Cadence presented photonic-native modeling built on waveguide scattering physics, delivering 2x faster simulation and paving the way for large-scale electro-photonic integration in the 1.6T era.

At ECOC 2025, Broadcom and Corning laid out what it takes to build a 102.4 Tb/s CPO Ethernet switch: density, performance, reliability and ease of assembly — and why the real challenge is system design, not the module.

At OCP Global Summit 2025, Google updated the OCP Optical Circuit Switching (OCS) subproject: open hardware and software specs, a SONiC-based management stack with gNMI/gNOI/gNSI, and OCS as a structural foundation for AI data centers.

At OFC 2025, Intel unveiled its second-generation 300mm heterogeneous silicon photonics platform, with 50% more laser power, 200G/lambda modulators and a full passive component library for pluggables, CPO and optical compute interconnect.

Larry Dennison of NVIDIA at OFC 2025 on why latency, error rates, packaging and power are pushing large AI systems beyond "copper where you can, optical where you must" — from 1.6T optical engines to CPO and the full menu of photonic options.

At ECOC 2025, Furukawa showed a 50 Gbaud × 16-channel CPO transceiver built on 1060 nm single-mode VCSELs and 19-core fiber: a 7.7 × 15.9 mm footprint, 2 km reach and 3.95 fJ/bit.

Once the laser is moved off the chip, someone has to catch that +20 dBm high-power light. AIST shows that polymer waveguides on low-cost glass-epoxy substrates can do it: zero degradation over six hours and only a 4.4°C temperature rise — a key feasibility endorsement for the packaging-centric CPO route.

While optics dominates the AI infrastructure conversation, Lotes used OCP 2025 to make the case for copper: CPC, NPC and active copper cables can still cover in-rack and adjacent-rack 200G scale-up links reliably and cost-effectively.

As AI training outgrows a single site, Scale Across is the new normal. How Meta defines the next generation of interconnect, why Google says "fiber is the new wavelength," and the strategy behind Lumen's push to add 41 million fiber miles.

At OFC 2026, Lumentum announced a $2 billion strategic investment from NVIDIA and a multi-billion-dollar OCS deal. We break down how its 400G-per-lane EML tackles the copper bottleneck and how the new Greensboro fab supports its path to $2 billion in quarterly revenue.