NVIDIA DGX Spark
GB10 Grace Blackwell
- VRAM
- 128 GB unified
- CPU
- 20-core Arm (10× Cortex-X925 + 10× Cortex-A725)
- Memory
- 128 GB LPDDR5x unified system memory
- Location
- United Kingdom
ML servers
Dedicated GPUs for the full ML lifecycle — from data pipelines and training to serving. OS options confirmed per configuration, with a managed stack.
Workloads
Train models from scratch on dedicated GPUs with consistent performance.
LoRA, QLoRA and full fine-tuning on your proprietary datasets.
Long-running RL and simulation jobs without preemption from other tenants.
PyTorch DDP, DeepSpeed and similar frameworks across GPUs and nodes.
ETL, Spark jobs and vector databases close to your compute.
FastAPI, Triton and vLLM for putting models into production.
Operating systems
Ubuntu, Proxmox or a custom OS image on V100S, A100 and L40S servers; DGX Spark ships with NVIDIA DGX OS. Full root or administrator access on every machine.
Scaling path
Pick the server that fits your current workload and order online.
Scale out with additional nodes as experiments turn into products.
Talk to an engineer about multi-GPU and custom builds for larger jobs.
GB10 Grace Blackwell
2× V100S 32 GB
1× A100 40 GB HBM2
1× L40S 48 GB GDDR6
Dedicated GPUs from €326/mo
Pick a server and order online, or talk to a GPU engineer about multi-GPU and custom builds.