Specifications
Part Number
MMS1X00-N5400
Manufacturer
NVIDIA
Form Factor
QSFP112 (quad-channel small form-factor pluggable, 112G per lane)
Aggregate Bandwidth
400Gbps
Per-Lane Signalling
4 x 112G PAM4
Wavelength
1310 nm
Optical Reach
DR4 — 500 m over single-mode fibre
Optical Interface
MPO-12; 8 of the 12 fibres used (4 Tx / 4 Rx)
Electrical Interface
4 x 112G PAM4 (QSFP112 host cage)
Typical Deployment
NVIDIA GB300 AI Station to NVIDIA Spectrum-X Ethernet switch
Thermal Design
External heatsink fins on the module shell — cools outside the host cage
Cage Compatibility
QSFP112 / QSFP-DD class cages; not OSFP
Signal Type
Active optical module (not a DAC)
Use Case
400G leaf/spine links and GPU-to-switch fabric in AI clusters
Interoperability
NVIDIA Spectrum-X / Quantum fabrics and ConnectX-8 class NICs
Form Factor Comparison
More compact than OSFP, which dominates 800G / 1.6T links
Supply
B2B volume supply; quote per quantity
Lead Time
15-day sample lead time on stock configurations
Overview
The NVIDIA MMS1X00-N5400 is a 400Gbps QSFP112 optical transceiver tuned for AI cluster fabrics. The QSFP112 form factor keeps the familiar quad-channel pluggable package while moving to 112G PAM4 per lane, so the same 4-lane footprint delivers 400G without the larger OSFP shell used at 800G and 1.6T.
The module is a 1310 nm DR4 design: four lanes in each direction over an MPO-12 connector, using eight of the twelve fibres. NVIDIA builds these with visible external heatsink fins, so a meaningful part of the thermal path sits outside the host cage rather than depending entirely on chassis airflow — a design choice that matters in dense 1U switch and server front panels.
In practice these are deployed to connect NVIDIA GB300 AI Station systems into Spectrum-X switches, which is exactly the role they play in a rack-scale GPU fabric: a short-to-medium reach, single-mode 400G link between compute tray and leaf switch.
Key Benefits
400G on a compact QSFP112 shell — the density of a quad-lane pluggable with 112G PAM4 signalling. DR4 single-mode optics cover 500 m over MPO-12, enough for intra-row and cross-row fabric. External heatsinks reduce dependence on cage airflow in dense 1U platforms. NVIDIA-fabric validated for Spectrum-X and GB300-class deployments, and available in B2B volume with per-quantity pricing.
Applications
400G GPU-to-switch uplinks in AI training and inference clusters; NVIDIA GB300 / GB200 AI Station fabric links; Spectrum-X Ethernet leaf-spine connections; ConnectX-8 host connectivity; storage and front-end network attached to AI racks; 400G test and validation labs.
Request a Quote — NVIDIA MMS1X00-N5400 400G QSFP112 DR4 OPTICAL TRANSCEIVER — 1310 NM
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →