Use Case · HPC

High-Performance Computing at Resita

Grid capacity, low OPEX, and EU jurisdiction for scientific simulation, financial modelling, and energy sector HPC workloads.

HPC Workload Profile

High-performance computing covers a broad range of compute-intensive applications: climate and weather simulation, pharmaceutical molecular dynamics, seismic processing for energy exploration, financial risk Monte Carlo, and engineering finite element analysis. These workloads share key infrastructure requirements:

  • Large, stable power blocks — HPC clusters run at near-100% utilisation; power intermittency causes job failure and data loss
  • High-bandwidth, low-latency networking — InfiniBand and RoCE fabrics require predictable inter-rack latency
  • Dense cooling — Modern HPC nodes (AMD EPYC, Intel Xeon, NVIDIA A/H series) exceed 5 kW per node in dense configurations
  • EU jurisdiction — Government, defence, and academic HPC increasingly requires EU data residency
Confirmed
650 MVA
Substation capacity
~€0.14/kWh
Industrial electricity
Romania
EU
Full GDPR &
data residency

OPEX Advantage for HPC Operators

HPC facilities operate at high utilisation factors (80–95%). Electricity is typically 60–70% of OPEX. At Romanian industrial electricity prices (~€0.14/kWh), a 50 MW HPC facility saves approximately €22–35M per year in electricity costs compared to equivalent facilities in Germany or the Netherlands.

Related: AI Training · PUE Guide · Request Briefing