- Home
- Dedicated clusters
Bare metal. Single tenant.
The stack is yours.
From 32 nodes and 256 GPUs to 256 nodes and 2,048 GPUs. No hypervisor. You get the nodes directly and choose your own scheduler and training stack; we are responsible for the power, the facility, the hardware, the fabric and the operations.
The first cluster: 32 nodes, 256 GPUs.
Nachster AI B300 — standard node and first cluster
| Platform | HGX B300 NVL16 · 8-GPU node · 10U air-cooled |
|---|---|
| GPU | 8 × NVIDIA B300 · 288 GB HBM3e each · ~2.3 TB / node |
| CPU · memory | 2 × Intel Xeon 6 · 3 TB DDR5-6400 |
| Local storage | 2 × 960 GB M.2 (OS) · 8 × 3.84 TB NVMe |
| Networking | 8 × ConnectX-8 800G (integrated) · BlueField-3 DPU |
| Cluster · fabric | 32 nodes · 256 GPUs · Quantum-X800 InfiniBand · rail-optimized · 1:1 |
| Shared storage | Dedicated high-performance storage |
| Power | ~15 kW per node (reference) |
| Location | Japan · dedicated, single tenant |
| Operations · SLA | Managed 24×7 · ≥99.5% per allocation unit / month |
| Contract | Multi-year committed capacity · 5-year standard |
Standard node configuration; the final BOM is attached to the allocation offer. Later generations — GB300 · Vera Rubin — are allocated on subsequent windows.
What is yours, what is ours.
| Layer | Owner | Detail |
|---|---|---|
| Application & models | You | Training code, weights, data, serving |
| Scheduler & OS | You | Slurm / Kubernetes, containers, driver versioning |
| Base image & acceptance | Joint | A validated image at handover, and the acceptance test |
| Hardware & fabric | Nachster AI | Nodes, InfiniBand, storage, failure replacement |
| Facility, power & cooling | Nachster AI | Substation, cooling, physical security |
| 24×7 operations | Nachster AI | Monitoring, first response, spares, escalation |
On isolation.
- No hypervisor — Bare metal. There is no virtualisation layer we operate between your workload and the hardware.
- No shared namespace — Storage and fabric are dedicated to your cluster. You do not share a namespace with another tenant.
- Not released during the term — Nodes under contract are not turned over to another customer during the term.
- Wiped on return — At the end of the term nodes are wiped to NIST SP 800-88, firmware is reset to baseline, and the node is re-validated against the acceptance standard before it is used again.
- Annual facility audit rights — You, or an auditor you appoint, have a contractual right to audit the facility annually.
The commercial frame.
5-year standard
Three years is available. Dedicated, reserved and non-cancellable. JPY, billed monthly in advance. Billing starts at acceptance, not at signature.
Locks allocation and date
Applied against service fees per the agreement. A higher reservation and a longer term improve the terms.
Set in the allocation offer
Determined by delivery window, contract term, cluster size and reservation level. Qualified customers receive an offer.
Tell us what you need. We reply within 48 hours.
Qualified enquiries receive proposed times for a technical meeting within 48 hours, and an allocation offer after that meeting. Information is handled under NDA.