Eliminate Patching. Optimize Your Fabric.

Infinity Shuffle OXC — Passive optical cross-connect using precision flex-circuit technology.

Replace manual fiber patching in spine-leaf and dragonfly architectures with pre-routed, high-density fiber shuffles that integrate seamlessly with ADTEK GPX patch panels. Reduce CAPEX by up to 40% while simplifying deployment and maintenance.

Passive Flex-Circuit

1200+ Fibers

0.05dB IL

Integrates with GPX ODF

Shuffle + Patch Panel, Pre-Integrated

Unlike other solutions that require adapter cassettes from third parties, the Infinity Shuffle OXC integrates natively with ADTEK’s GPX patch panel family (GPX51, GPX58, GPX59, GPX61, GPX62, GPX70). Choose MPO, MTP, MMC, or SN-MT interfaces—all from one vendor, one support team, one supply chain. Parallel architecture for front-access ODFs; inline architecture for rear-access ToR panels. No compatibility guesswork, no multi-vendor finger-pointing.

40% Infrastructure Savings

Theoretical modeling shows significant CAPEX reduction by enabling 400G/800G consolidation: fewer switches (8 vs 24 in spine-leaf), fewer transceivers (320 vs 1,280), lower power consumption (POEX), and reduced maintenance complexity. The shuffle enables higher-speed, higher-port-count architectures that maximize hardware utilization. Example: A 2,000-server deployment saves ~$342,400 in direct CAPEX by supporting 400G spine-leaf with 3:1 consolidation versus traditional 100G architectures.

Factory-Routed, Field-Reliable

Automated substrate cutting (±0.5mm), biomimetic fiber routing (±0.1mm), and precision glue spraying (±0.5mm) ensure consistent quality at scale. Each shuffle is tested to ≤0.05dB insertion loss (excluding connector loss)—critical for maintaining signal integrity across complex dragonfly all-to-all meshes. No field splicing, no routing errors, no patch cord spaghetti. Just pre-validated fiber paths from spine to leaf.

Inline or Parallel, You Choose

Inline architecture (recommended for ToR panels): No port loss, rear-fed cables, cassette removal for troubleshooting. Parallel architecture (required for ODF front-access): 50% port efficiency loss, but all patching from front—ideal for traditional data center ODF layouts. Match the architecture to your cable management topology, not the other way around. Both options available with MPO, MTP, MMC, or SN-MT connectors.

Technical Specifications

Standards: Telcordia GR-63, GR-1435 (MPO), IEC 61300, IEC 61753-1, IEC 61754-7 / TIA-604-5

Connector Options (Head & Tail):

MPO/MTP (8, 12, 16, 24 fiber)

MMC (Multi-fiber Mini Connector)

SN-MT (VSFF)

Bare fiber (for fusion splicing)

Optical Performance (IEC 61753-1):

Type

Insertion Loss (Typical)

Insertion Loss (97% Random Mated)

Premium (UPC/APC)

≤0.12 dB

≤0.25 dB

Premium Low Loss

≤0.15 dB

≤0.18 dB

Return Loss (APC)

≥65 dB

Return Loss (UPC)

≥60 dB

Patch Panel Integration:

Cassette Type

Configuration

Density

Architecture

GPX51 / HD51

MPO/MTP, 6In/6Out

6 cassettes/RU

Inline

GPX58 / HD58

MPO/MTP, 6In/6Out

4 cassettes/RU

Inline l

GPX59 / HD59

MPO/MTP, 12In/12Out

3 cassettes/RU

Parallel

GPX61 / HD61

MMC/SN-MT, 9×4In/9x4Out

3 cassettes/RU

Parallel

GPX62 / HD62

MMC/SN-MT, 9×4In/9x4Out

8 cassettes/RU

Inline

GPX70 / IHD70

MPO/MTP, 6In/6Out

18 cassettes/RU

Inline

Flex Circuit Specifications:

Parameter

Specification

Substrate Material

Polyimide Film

Conformal Coating

Green Glue

Max Dimensions

1000mm × 800mm

Max Fiber Count (Single Layer)

1200+ @ 1000mm × 800mm

Fiber Types Supported

Single-mode (G657.A1/A2), Multi-mode (OM3/OM4), Special fiber

Fiber Mapping

Fully customized

Insertion Loss (Shuffle Only)

≤0.05 dB (excluding connector loss)

Operating Temperature

-40°C to +85°C

Deployment Configurations

Density for GPU Scale-Out

Training large language models requires thousands of GPUs interconnected via high-bandwidth fiber. Traditional single-fiber cables congest cable trays and create airflow blockages. MCF FIFO delivers 4x the fiber count in the same physical pathway—run 400G/800G links between GPU racks without adding cable management chaos. The factory-aligned FIFO breaks out to standard LC connectors at each end, integrating seamlessly with existing InfiniBand or Ethernet switches while you prepare for next-gen MCF-native hardware.

Maximize Conduit Fill

Campus and metro DCI links often face physical constraints—underground ducts, aerial pathways, and leased fiber with limited capacity. MCF quadruples the capacity per fiber strand without digging new trenches or leasing additional dark fiber. The compact 55mm FIFO device fits standard splice trays and ODFs, making it compatible with existing termination infrastructure. Deploy MCF between buildings now; upgrade to native MCF transceivers later without touching the cable plant.

Bridge Today, Native Tomorrow

MCF represents the physical layer for future 800G and 1.6T parallel optics—but the cables can be deployed today. Our FIFO solution lets you install MCF backbones during current build-outs, use them with existing LC/SC-based 400G hardware, and simply remove the FIFO devices when MCF-native transceivers become available. No stranded assets, no rip-and-replace. The 75% space savings start immediately; the protocol upgrades come when you’re ready.

Solve the Fiber Density Bottleneck

Ready to evaluate Multi-Core Fiber for your AI training or DCI infrastructure? ADTEK is accepting partners for the MCF FIFO pilot program. Receive engineering samples, assembly guidelines, and application support for integrating 4-core or 7-core MCF into your existing LC/SC infrastructure.