RebelRack™

The Turnkey AI Inference Rack for Production Deployment

The Definitive Turnkey Infrastructure for
Generative AI and Agentic Workflows

RebelRack is a fully integrated AI inference infrastructure designed to simplify large-scale deployment. By combining RebelServer, high-speed networking, and the Rebellions Software Stack into a validated rack-scale solution, RebelRack enables organizations to move from infrastructure setup to production inference faster and more efficiently.

Fully integrated rack-scale AI infrastructure

Optimized for agentic AI and distributed inference

Air-cooled deployment for existing data centers

Production-proven configurations with open software ecosystem

Key Features

Turnkey Deployment

Pre-integrated and validated infrastructure designed to reduce deployment complexity and accelerate production readiness.

Distributed Inference Ready

High-speed networking and optimized multi-node architecture built for large-scale inference workloads and modern AI pipelines.

Air-Cooled Infrastructure

Deploy into existing air-cooled data centers without requiring liquid cooling retrofits or specialized infrastructure upgrades.

Open Software Stack

Integrated support for vLLM, PyTorch, and open-source AI frameworks with optimized distributed inference software.

86%
RebelRack™
vLLM
PyTorch
TensorFlow
Huggingface
RebelRack™

Key Benefits

Industry’s Most Profitable
Inference Platform

Engineered for energy-efficient inference and agentic AI workloads at scale, RebelRack maximizes throughput while reducing operational costs and infrastructure overhead.

Proven Reliability,
Immediate Deployment

Production-proven and pre-validated configurations help organizations reduce deployment complexity and accelerate time-to-production from months to days.

Sovereign AI with
Zero Infrastructure Friction

Deploy fully on-premises AI infrastructure while maintaining operational control, security, and data sovereignty within existing enterprise and government data center environments.

Integrated Rack Infrastructure

Fully Integrated Rack-Scale AI Infrastructure

RebelRack combines compute, networking, management, and software into a unified AI inference platform optimized for production deployment.

The infrastructure integrates dedicated NPU and management racks designed to simplify operational deployment while enabling scalable distributed inference environments.

RebelRack™ NPU Rack
RebelRack™ Management Rack

RebelRack™ NPU Rack

Server Model 4x RebelServer™ (5U)
Accelerator 32x RebelCard™
NPU Memory 4.61TB total, 153.6TB/s HBM3e bandwidth
CPU 8x AMD EPYC™ 9355 Processors (256C/512T)
System Memory 6TB DDR5
Network Backend Fabric : 400G QSFP112-DD
Frontend Network : 25G SFP28
Storage Network : 100G (Optional)
Performance FP8 : 65.5 PFLOPS
FP16 : 32.7 PFLOPS
Cooling Type Air-cooled
Power Consumption 28kW max
Rack Height: 42U, 78.38” (1991mm)
Width: 23.62” (600mm)
Depth: 47.24” (1200mm)
Weight ~773 lbs (~351 kg), unpackaged
Cooling 105,094 BTU/hour (max)

RebelRack™ Management Rack

Management Servers 3x AMD CPU Servers (optional)
CPU 3x AMD EPYC 9254/9255 (24 cores)
System Memory 384GB Total
(3x Management Servers, each with 128GB [4x 32GB DDR5])
Network Backend Fabric : 800G OSFP
Frontend Network : 25G SFP28 + 100G QSFP28
Storage Flexible support for validated partner storage solutions
Power Consumption 5.2kW max
Rack Height: 42U, 78.38” (1991mm)
Width: 23.62” (600mm)
Depth: 47.24” (1200mm)
Weight ~558 lbs (~253 kg), unpackaged
Cooling 19,517 BTU/hour (max)

* Multiple rack configuration options are available to enable easier integration into existing data center environments

Rebellions Software Stack

Open and Optimized AI Software Stack

RebelRack comes pre-integrated with the Rebellions Software Stack, enabling organizations to deploy and scale AI inference workloads using familiar open-source frameworks and existing operational workflows.

Built for production deployment, the software ecosystem reduces infrastructure complexity while simplifying distributed inference and model deployment at scale.

stack graphic

Familiar
Open Ecosystem

Integrated support for vLLM, PyTorch, and open-source AI frameworks without requiring new operational workflows or proprietary software environments.

Distributed
Inference Optimization

Specialized distributed inference software manages high-speed interconnects and multi-node communication to simplify large-scale AI deployment.

Production-Ready
Model Support

Pre-validated and optimized model support helps organizations accelerate deployment timelines and reduce operational overhead.

Cloud-Native
Infrastructure

Built for modern AI infrastructure environments with support for scalable orchestration and cloud-native operational tooling.