NVIDIA AI Factory MGX Ecosystem (Jentech)

Powering the Future of AI Factories: Advanced Liquid Cooling for NVIDIA MGX and Vera Rubin Platforms

As artificial intelligence rapidly transitions into the era of AI factories, data centers are evolving from traditional compute environments into highly orchestrated, large-scale systems designed to generate intelligence at unprecedented speed and scale.

At the center of this transformation is the NVIDIA’s NVIDIA MGX™ architecture—an open, modular platform that enables the industry to design and deploy accelerated computing infrastructure with unmatched flexibility and scalability.

AI workloads today are no longer limited to training. They now span pretraining, post-training, test-time scaling, and agentic workflows, all of which dramatically increase compute density, thermal output, and infrastructure complexity.

In this new paradigm, thermal management is no longer a supporting function—it is a foundational enabler of AI scalability.

An NVIDIA MGX Ecosystem Partner: Enabling Scalable AI Infrastructure

As an NVIDIA MGX ecosystem partner, Jentech specializes in delivering advanced liquid cooling solutions and critical thermal components for next-generation AI systems.

Our portfolio includes:

●       High-performance GPU and CPU cold plates

●       NVLink and switch tray cooling solutions

●       Liquid distribution manifolds.

●       Rack-level thermal integration components

All solutions are designed in alignment with MGX principles of modularity, flexibility, and forward compatibility, enabling customers to rapidly scale AI infrastructure while minimizing development complexity and cost.

By leveraging standardized building blocks, we help system builders:

●       Accelerate time-to-market

●       Reduce total cost of ownership (TCO)

●       Enable seamless multi-generation upgrades

Designed for the Vera Rubin Era: A New Thermal Paradigm

With the introduction of the NVIDIA Vera Rubin platform, AI infrastructure enters a new phase—defined by agentic AI, extreme-scale inference, and rack-level system orchestration.

The third-generation NVIDIA MGX  rack architecture introduces:

●       100% liquid-cooled system design 

●       Support for 45°C warm water cooling 

●       Fanless, cable-free, and hose-free architectures 

●       Rack-level optimization across power, compute, and cooling

These innovations fundamentally redefine thermal requirements:

Traditional cold plate designs are no longer sufficient.Cooling systems must now evolve in lockstep with AI architecture.

Introducing NVIDIA Vera RubinVERA RUBIN NVL72: A Cooling Platform Built for AI Density Extremes

To meet these demands, Jentech developed a cooling platform for NVIDIA Vera RubinVERA RUBIN NVL72, a next-generation liquid cooling platform engineered specifically for MGX and Vera Rubin-based systems.

NVIDIA Vera RubinVERA RUBIN NVL72 is not an incremental upgrade—it is a ground-up redesign for AI workloads at scale.

Optimized Flow Distribution (Hotspot Targeting)

Jentech's cooling solution for NVIDIA Vera RubinVERA RUBIN NVL72 features advanced flow distribution tailored to GPU hotspot regions:

●       Non-uniform flow routing for targeted cooling

●       Localized velocity enhancement

●       Reduced thermal gradients across critical components

This ensures optimal cooling performance under dynamic, burst-heavy AI workloads, including large language models and agentic inference.

Dual-Layer Manifold Architecture

A dual-layer manifold design separates:

●       Flow distribution layer 

●       Cooling micro channel layer 

Key benefits include:

●       Improved flow uniformity

●       Reduced pressure drop

●       Enhanced scalability for multi-GPU configurations

This architecture aligns directly with MGX’s modular, rack-scale deployment strategy.

Micro channel Geometry Optimization

NVIDIA Vera RubinVERA RUBIN NVL72 leverages advanced micro channel engineering:

●          Optimized fin density and channel geometry

●          Increased heat transfer coefficient

●          Higher cooling efficiency at equivalent flow rates

The result is a thermal solution purpose-built for high-TDP GPUs in Blackwell and Vera Rubin generations.

Production-Grade Reliability: Laser Welding for Zero Leakage Risk

Reliability is mission-critical in AI factories.

NVIDIA Vera RubinVERA RUBIN NVL72 adopts a fully laser-welded construction:

●       No adhesives or O-ring-dependent primary seals

●       Fully sealed metallic flow structure

●       Designed for long-term operational stability

This delivers:

●       Near-zero leakage risk 

●       Extended lifecycle durability

●       Hyperscale-ready reliability

Aligned with MGX’s emphasis on resiliency and production-grade system design, NVIDIA Vera RubinVERA RUBIN NVL72 ensures dependable operation at scale.

From Component to Infrastructure: Delivering Real AI Factory Value

In the Vera Rubin era and beyond, cooling solutions must do more than dissipate heat—they must unlock system-level performance and scalability.

NVIDIA Vera RubinVERA RUBIN NVL72 enables:

1. Thermal Headroom Expansion

Supports higher GPU power density→ Maximizes tokens per rack

2. Energy Efficiency Gains

Reduces cooling overhead→ Improves PUE and energy utilization

3. Deployment Readiness

Fully aligned with MGX standards→ Accelerates large-scale deployment

4. Future-Proof Design

Supports multi-generation platform upgrades→ Protects infrastructure investments

Conclusion: Cooling as the Backbone of AI Factories

As AI competition shifts from models to infrastructure, the defining factor is no longer just compute capability—but the ability to deliver and sustain compute at scale.

By combining:

●       NVIDIA MGX modular architecture

●       Vera Rubin rack-scale system design

●       And NVIDIA Vera RubinVERA RUBIN NVL72 advanced liquid cooling technology

Jentech is enabling the next generation of: Scalable, efficient, and production-ready AI factories