Specifications

Product

Coherent 1.6T OSFP optical transceiver

Form Factor

OSFP (1.6T class)

Aggregate Rate

1.6 Tb/s

Per-Lane Rate

200 Gb/s PAM4

Architecture Options

Fully retimed, transmit-retimed (LRO), linear pluggable (LPO)

LPO Power

Approximately 10–12 W

Bit Error Ratio

~8 × 10⁻¹⁰ demonstrated in the LPO configuration

Demonstration

Live 200G/lane link at OFC 2026

Optical Technology

EML-based and silicon-photonics portfolios

Internal Lasers

Coherent-designed CW lasers, EMLs and photodetectors

Target Platforms

Next-generation AI switch and NIC cages (1.6T / 800G class)

Latency Benefit

LPO removes the DSP retiming stage from the optical path

Thermal

Heat-sink options per cage type (IHS / RHS)

Strategic Context

March 2026 NVIDIA partnership with Coherent and Lumentum on AI datacenter optics

Ecosystem

Multiple 1.6T architectures so a single vendor covers retimed and linear designs

Compliance

OSFP MSA, CMIS

Application

AI data center interconnect, scale-out fabric

Availability

Sampling / early production — quote for allocation

Overview

Coherent's 1.6T OSFP module is the vertically integrated option in the 1.6T optical market: the company builds its own CW lasers, EMLs and photodetectors, so the module is not dependent on a third-party optical supply chain. The portfolio deliberately spans three architectures — fully retimed, transmit-retimed (LRO) and linear pluggable (LPO) — because AI fabric operators are splitting on how much DSP they want in the module.

The LPO configuration is the interesting one for power-constrained racks. By removing the module-side DSP and letting the host switch ASIC drive the optics directly, Coherent demonstrated a live 200G-per-lane link at roughly 10–12 W with a BER around 8 × 10⁻¹⁰. That is a materially lower power and latency point than a retimed module, at the cost of tighter host-side equalisation requirements and a more constrained set of supported switch platforms.

Coherent's position in AI datacenter optics was reinforced in March 2026 when NVIDIA announced strategic partnerships with Coherent and Lumentum to accelerate advanced optics development. QS Compute supplies Coherent 1.6T modules alongside the compatible third-party OSFP224 portfolio, so mixed retimed/LPO builds can be sourced on one order.

Key Benefits

Vertical integration — in-house CW lasers, EMLs and photodetectors reduce supply-chain exposure in an allocation-constrained market. Three architectures from one vendor lets you standardise the optical vendor while still choosing retimed or LPO per rack. 10–12 W LPO cuts module power at the rack level, and no module DSP removes a latency stage from the fabric path.

Applications

1.6T scale-out fabric between AI switch tiers; switch-to-NIC links in liquid-cooled and air-cooled racks where per-module power budgets are binding; and multi-vendor builds where a single optical supplier must cover both retimed and linear designs.

Request a Quote — COHERENT 1.6T OSFP OPTICAL TRANSCEIVER — 200G PER LANE AI DATACENTER MODULE

QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.

Get Your Quote →

Related Products

NVIDIA Quantum-X800 Q3400-RA InfiniBand Switch NVIDIA ConnectX-8 C8220 SuperNIC NVIDIA BlueField-3 B3240 DPU NVIDIA Spectrum-6 SN6800-LD Ethernet Switch Molex 224G OSFP1600 Interconnect System Credo ZeroFlap AEC