Multi-Platform GPU Fleet

A Multi-Platform GPU Fleet Built for Lifecycle Resilience

closed AI is designed as a heterogeneous GPU fleet, combining multiple NVIDIA platform generations into one infrastructure backbone. This allows the platform to serve different workload classes, customer profiles and lifecycle stages without becoming dependent on a single hardware generation.

Vera Rubin NVL144

Flagship layer for ultra-high-end AI training, premium inference, and next-generation model workloads.

GB300 NVL72

Premium workhorse layer for large-scale training, high-performance inference, and high-memory AI workloads.

GB200 NVL72

Scalable production layer for broad AI workloads, fine-tuning, inference, and enterprise capacity.

H200 SXM

Stable baseline layer for development, research, HPC and proven production workloads.

Core principles

Built Around Control and Responsibility

No Black Box Systems

Systems remain transparent, auditable, and understandable across every layer.

Scalability by Design

The platform is modular by design, enabling structured expansion through defined capacity stages.

No Trade-Offs

Built for performance and sustainability, with no compromise across system design or operation.

Digital Operational Modeling

Systems are modeled as digital twins, enabling automated operation and predictive maintenance.

Industrial-Grade Operation

High-performance computing is treated as an industrial system, planned, operated, and managed with precision.

Zero-trust Infrastructure

Infrastructure is designed for full control, operating independently without reliance on external platforms.

Integrated System

Energy, infrastructure, and operations are engineered as a unified system for stability and predictable performance.

Data Security

Data is handled within isolated systems, with controlled access and verifiable security at every level.

The first step Toward 25 MW

1 MW Go-Live PoC in Liechtenstein

PoC Liechtenstein ClosedAI

closed ai in action

With a 1 MW AI infrastructure, meum tec is the first company in Liechtenstein to implement such computing capacity. The proof of concept is the first operational stage of a 25 MW build. It establishes the model — sovereign, controlled AI infrastructure — at a working scale, proving the approach before it is expanded.

Rather than a standalone pilot, it is the entry point of a phased path toward full capacity, and the foundation for the development stages that follow.

Designed as an integrated system, the infrastructure combines compute, operations, and long-term scalability within a structured environment. The deployment serves as the foundation for future expansion phases and the continued development of sovereign AI systems.

1 MW

AI Infrastructure Capacity

Dell

H200 SXM, GB200 NVL72, GB300 NVL72

4-8

Number of Racks

46-110

Number of Nodes

CLOSEDAI SOFTWARE · DIGITAL & AI SYSTEMS

Built to Run Our Own Software

ClosedAI infrastructure is the foundation for the software built on top of it. The applications developed under ClosedAI run on meum tec’s own sovereign, controlled infrastructure, so the full stack—from compute to application—stays transparent and under our control.

Alongside the infrastructure platform, our dedicated software group develops proprietary products, explores high-value AI use cases, and turns promising ideas into deployable systems. We work with enterprises, research institutions, public organizations, and technology partners—from feasibility studies and concept development through prototyping, implementation, integration, and operation.

Explore ClosedAI Software →

SOFTWARE PRODUCTS