Inside the AI Powerhouse: Aivres Brings NVIDIA’s Vera Rubin to Life

Aivres exhibited some serious computing capability at the 2026 OCP Global Summit in San Jose. The highlight of the stand was the KRS8000V4 which is a liquid-cooled rack based on the NVIDIA’s Vera Rubin platform. The system features 72 Rubin GPUs and 36 Vera CPUs, all linked via sixth generation NVLink.

The system includes 18 compute trays and nine NVLink switch trays. Together, they create a massive computing environment with more than 20TB of HBM memory. Aivres utilizes complete liquid cooling technology, where the rack can work with the coolant temperature as high as 45°C. The use of such a solution is used to handle heat generated from dense AI hardware.

Eight GPUs, Two CPUs, and a Fanless Design

Aivres has also introduced KL2288, a small 2U compute tray specifically designed for NVIDIA HGX Rubin NVL8. The tray comes with eight Rubin GPUs and two AMD EPYC 9006 “Venice” CPUs.

As claimed by Aivres, it can perform at 400 PFLOPS speed for NVFP4 inference and 280 PFLOPS speed for NVFP4 training. Furthermore, it also supports 2.3TB of HBM4 memory and 176 TB/s of memory bandwidth.

In terms of cooling, it makes use of complete liquid cooling technology for its processors and DDR5 memory. It has a modular design that is fanless and also uses quick disconnects and power busbar for easy rack integration.

This Hardware & AI Future

Aivres brings about a significant trend in AI infrastructure. With models getting larger and increasing workloads, performance of the GPUs is just one of the factors. Factors like memory bandwidth, power delivery, cooling, and rack integration will play a key role in determining the efficiency of AI operations in the coming years.

These developments could enable better AI services and bigger workloads for the organizations. For you, this would eventually translate into better capabilities of AI solutions and fast response time.

Find this helpful? Spread this article across your network and check out other articles on AI hardware and data centers.