Investment | Price benchmarks | 05
Rack row
A row is the first system that behaves like a real cluster: scheduler, parallel file system, per-job telemetry and a growth plan. We design it as such, with headroom to extend onto the next row.
Typical configuration
Departmental cluster / AI pod with parallel storage
Investment range
€1.2–8M
What this tier involves
- Typical role
- Production HPC & AI clusters
- Storage
- Lustre, Spectrum Scale or BeeGFS, with tiered NVMe
- Scheduling
- Slurm / PBS with fair-share and accounting
- Fee
- Basic €5–25k | Platinum €90–400k | Supreme €35–90k/year
Indicative figures, August 2026, excluding VAT. High volatility on GPUs and memory — firm quotes are valid for 30–60 days. The final price is set at scoping.
What matters at this scale
A row is designed as a complete cluster: separate roles per rack, liquid distribution sized for maintenance and usage policies defined from the start.
- Pod architecture
- Compute, parallel storage and management on racks with separate roles, with space and cabling planned as a whole
- Liquid distribution
- CDUs are sized with maintenance margin, so interventions do not stop the row
- Parallel storage
- The NVMe tier is sized on per-client bandwidth; protection is set per data class
- Usage policies
- Fair-share, quotas and accounting, configured from commissioning
- Growth
- Floor space, power, cooling and fabric ports for the next row are reserved at design stage