Name: Four-Cabinet Integrated AI Computing Infrastructure System
Brand: OPTIWAVE

Four-Cabinet Integrated AI Computing Infrastructure System

Make AI infrastructure easier to deploy, easier to scale, and easier to run for the long term.
At OPTIWAVE, we don't build a single product; we build system-level AI infrastructure, integrating compute, high-speed optical interconnect, power, cooling, energy storage and monitoring into ready-to-deploy system units.
In many AI projects, the real challenge is not a shortage of compute but the complexity of system integration.Power, cooling, networking and operations are often handled separately, and the problems surface only during deployment and scale-up.
OPTIWAVE takes a different approach.
From day one we define clear system boundaries, a modular architecture and a predictable scaling model, so customers can grow from one rack to a Pod, and even to multiple sites, with the same logic.
Energy here is not a “backup battery”,
but a system-level energy buffer that improves power continuity, load stability and operational confidence in high-density AI scenarios.

Request a Test

Product Features Specifications Applicable Scenarios

Product Features

System Overview

The system consists of four core cabinets: the AI optical-cooling integrated cabinet, the Micro-MDC management and storage cabinet, the ORV3 high-speed optical interconnect cabinet, and the energy storage and power management cabinet.

Design Objectives

High power density of 80–150 kW per system, high-efficiency liquid cooling (target PUE ≤ 1.15), 400G/800G optical interconnect, industrial-grade energy storage and UPS for continuous operation, and plug-and-play rapid deployment.

The Four Cabinets
AI Optical-Cooling Integrated Cabinet	Micro-MDC Management and Storage Cabinet	High-Speed Optical Interconnect Cabinet	Energy Storage and Power Management Cabinet
Supports 4–8 GPU servers	Management servers, storage and BMC systems	Supports 400G/800G spine-leaf architecture	Industrial-grade UPS or PCS + battery system
Cold-plate liquid cooling + CDU secondary loop	Out-of-band management network	High-density fiber distribution system	Backed by mature domestic manufacturing capability
Power capacity: 30–80 kW (per cabinet)	Cloud-native operations capability	Latency target ≤ 2–5 µs	Supports 0 ms transfer (target < 4 ms)
Full-load GPU temperature target ≤ 70 °C			Provides 5–30 minutes of ride-through

System-Level Advantages

High power density: A single system supports 80–150 kW.

High-efficiency liquid cooling: significantly lowers PUE and supports sustained full-load GPU operation.

High-speed optical interconnect: Supports AI training and large-scale cluster expansion.

Energy storage for stability: Built on mature capabilities from the industrial energy storage sector.

Rapid deployment: factory pre-integrated; on site, only power, cooling and network connections are needed.

Specifications

Module	Parameter range
AI Rack	30–80 kW, GPU × 4–8
MDC Rack	5–12kW
Interconnect Rack	2–6 kW, 400G/800G
Energy Rack	20–100 kWh, UPS 5–30 min
Liquid cooling system	Inlet water 18–25 °C; ΔT 5–10 °C
Optical interconnect	BER ≤ 1e-12; latency 2–5 µs
System PUE	1.10–1.20 (target)