A trusted partner for national-scale systems.
Kronsavier engineers lead programmes at the very top of the spectrum — including DEUS 11, an approximately 9 PF supercomputer programme in Cluj-Napoca, built for high-performance computing and AI. The details remain under client confidentiality; the capability does not.
- Programme
- National-scale HPC & AI supercomputers
- Facility
- New-build & retrofit data-centre engineering
- Assurance
- Independent infrastructure & specification audits
Kronsavier is part of a wider ambition — helping Romania become an internationally recognised centre for high-performance computing. We build the systems that turn that vision into real capability.
A heterogeneous system, and a facility engineered around it.
A complete engineering dossier governs the programme — from the system architecture and the facility that serves it to procurement, governance and operations — traceable from requirement to acceptance.
System architecture
- Deliberately heterogeneous and multi-vendor — no dependence on a single ecosystem
- Dedicated partitions per workload class, each on the hardware that suits it
- One interconnect fabric, one shared storage tier with a long-term archive, one scheduler
- A fair-share model that keeps idle capacity working
- Expansion designed in from day one and absorbed without redesign
Facility engineering
- Direct liquid cooling, with year-round free cooling as the primary thermal path
- Heat recovered into the building's own heating; minimal water use, no industrial effluent
- A redundant, concurrently maintainable power chain, from grid intake to standby generation
- Fire detection and clean-agent suppression, physical security, full facility telemetry
- Sized once for the ultimate load and populated in phases
Delivery & assurance
- Traceability from codes and standards to every design decision
- The system–facility interface documented in a demarcation register
- Procurement packages derived from the design, ready for EU public tendering
- Risk analysed at design time, down to single points of failure and failure modes
- Operations readiness: staffing, procedures and a human-capital plan for the specialists who run it
A programme delivered as a dossier, not a single document.
From architecture to operations, each discipline has its document — and the documents govern one another, with declared precedence and interface registers.
System architecture
Defines the machine — the partitions, the interconnect fabric, the storage, the software stack — and the rules by which it grows.
Facility document
Power, cooling, fire, security, monitoring — every design decision recorded with its rationale.
Procurement documentation
Tender packages and employer's requirements, with award criteria verifiable by benchmark, for EU public procedures.
Governance framework
Who owns, who decides, who is answerable — aligned with the European HPC hosting model, with capacity allocated by independent review.
Operations manual
Procedures for the operations centre, monitoring, access and maintenance — no work on a critical system without an approved procedure and a way back.
Dependency matrix
Shows which values move together, so any proposed change can be traced through every system it touches.
Human-capital plan
Fully costed scenarios for educating the specialists who will run the system, at first-rank institutions.
Decision materials
An executive synthesis and an interactive facility dashboard for the beneficiary and funders.
Published under NDA — no configurations, no quantities, no vendors, no performance figures
A regional supercomputer, within IPCEI AI.
Kronsavier is running a project of its own within IPCEI AI — the Important Project of Common European Interest on artificial intelligence — as a Romanian SME, leading a Bucharest-anchored, cross-border cluster of academic and industrial partners with a European dissemination programme. The project has two components that sustain each other.
The supercomputer
A regional supercomputer, built in Romania and designed to become the country's first state-financed operational supercomputer. Once live, the system opens to the market and operates commercially as an HPC service — compute infrastructure as economic capability, not only scientific capability.
The foundation model
The machine's reason to exist: training a European, sector-specific AI foundation model for the AI-assisted inverse design of next-generation photonic and metamaterial devices.
Project in progress, under confidentiality rules — architecture, partners, timeline and figures remain unpublished