California startup ODINN has unveiled an AI supercomputer in the form of a carry-on suitcase ahead of CES 2026. This supercomputer provides AI power without the extra cost and time required to build stationary data centers.
California startup ODINN has unveiled an AI supercomputer in the form of a carry-on suitcase ahead of CES 2026. This supercomputer provides AI power without the extra cost and time required to build stationary data centers.
OMNIA is a self-contained system that can be deployed in minutes. It is equipped with the same CPU, GPU, memory, and storage found in data centers, writes Interesting Engineering.
Thanks to its patented closed-loop liquid cooling system, OMNIA operates quietly even under extreme loads — whether it's working with petabytes of data or running the most advanced AI algorithms.
The biggest advantage of using such a system is that it can connect to standard power and network infrastructure, and can be placed in an office, hospital laboratory, or even a secure facility without the need to build a separate server room.
A data center the size of a carry-on would certainly have limitations in the amount of data it could handle. But ODINN anticipated this and created the Infinity Cube, an AI cluster that combines multiple OMNIA systems in a single glass enclosure.
Infinity Cube is a modular data center consisting of multiple OMNIA units mounted in a cubic structure. Each module has its own cooling and computing power, eliminating the need for external infrastructure such as cooling towers or raised floors. Infinity Cube gives a company complete sovereignty over its AI cluster without having to spend months building it.
Cloud-based AI data centers allow organizations to focus on developing their models, rather than managing the infrastructure they run on. Bringing infrastructure “on-premises” can increase operational challenges for organizations. ODINN addresses these issues with its software layer — NeuroEdge
The software integrates well with NVIDIA’s AI software ecosystem and other frameworks, and coordinates task scheduling and deployment to get the most out of the AI cluster. Institutions can focus on using AI, not on tuning systems for optimal performance.
