Bare Metal GPU
Dedicated hardware infrastructure for maximum performance and control. Deploy your AI workloads on exclusive NVIDIA B300 and GB300 NVL72 superchips with full hardware access.
NVIDIA HGX B300 NVL16
Sovereign high-performance NVIDIA HGX B300 GPU instances in Europe, available as bare metal or GPUaaS. Train, fine-tune and deploy frontier AI workloads with the performance of a hyperscaler and the control of dedicated infrastructure.
Up to 2x attention performance
The HGX B300 doubles attention performance for transformer and long-context workloads compared to the previous generation, accelerating modern reasoning and retrieval models.
2.1 TB total GPU memory
With 2.1 TB of high-speed GPU memory, the HGX B300 supports larger models, longer sequences and higher concurrency without running into memory bottlenecks.
1.6 TB/s networking bandwidth
High-bandwidth interconnects deliver 1.6 TB/s of networking bandwidth per node, enabling seamless scaling across multi-node clusters for distributed training and inference.
Specifications compare the NVIDIA HGX B300 NVL16 to the HGX B200.
NVIDIA GB300 NVL72
Experience groundbreaking performance with the NVIDIA GB300 Grace Blackwell Ultra Superchip, featuring 72 Blackwell Ultra GPUs and 36 Grace CPUs with up to 40 TB of fast memory and 130 TB/s NVLink bandwidth.
288 GB of HBM3e
Larger memory capacity allows for larger batch sizing and maximum throughput performance. NVIDIA Blackwell Ultra GPU's offer 1.5x larger HBM3e memory.
Fifth-Generation NVIDIA NVLink
Unlocking the full potential of accelerated computing requires seamless communication between every GPU.
NVIDIA ConnectX-8 SuperNIC
The NVIDIA ConnectX-8 SuperNIC's input/output (IO) module hosts two ConnectX-8 devices, providing 800 gigabits per second (Gb/s) of network connectivity for each GPU.
Performance is based on the GB300 NVL72 Superchip, featuring 72 NVIDIA Blackwell Ultra GPUs and 36 NVIDIA Grace CPUs with up to 40 TB fast memory and 130 TB/s NVLink bandwidth.
Bare Metal GPU Specifications
Dedicated infrastructure designed for maximum performance, security, and control. Deploy your AI workloads on exclusive hardware with full access to every resource.
Hardware Options
NVIDIA GB300 NVL72 (72 Blackwell Ultra GPUs) or HGX B300 (8 Blackwell Ultra GPUs)
Storage & Memory
Up to 40 TB Fast Memory (GB300) or 2.1 TB (HGX B300), NVMe storage with enterprise-grade redundancy
Performance
1,400 PFLOPS FP4 (GB300) or 144 PFLOPS | 108 PFLOPS FP4 (HGX B300), dedicated hardware for maximum throughput
Networking
800 Gb/s per GPU connectivity (GB300) or 1.6 TB/s aggregate (HGX B300), NVIDIA ConnectX-8 SuperNIC, Quantum-X800 InfiniBand support
Security & Compliance
Full hardware isolation, EU data sovereignty, GDPR compliance, SOC 2 Type II certified infrastructure
Availability & Support
99.9% uptime SLA, 24/7 expert support, rapid deployment, custom configuration options
Ready to Deploy on Dedicated Hardware?
Get exclusive access to NVIDIA's most powerful AI superchips with full hardware control and expert support.
Why Choose Bare Metal Hosting?
For organizations that demand maximum performance, security, and control, bare metal hosting provides the ultimate infrastructure solution for AI workloads.
Dedicated Hardware
Exclusive access to entire NVIDIA GB300 or GB200 NVL72 superchips with no resource sharing or virtualization overhead.
- Maximum performancePredictable latencyFull resource utilization
Peak Performance
Achieve the highest possible throughput for AI training and inference with direct hardware access and optimized configurations.
- Zero virtualization overheadDirect GPU memory accessOptimized for AI workloads
Enhanced Security
Complete hardware isolation provides the highest level of security for sensitive AI workloads and data processing.
- Hardware-level isolationNo multi-tenant risksCustom security controls
Full Control
Complete control over hardware configuration, operating systems, and software stack for maximum customization.
- Custom OS configurationsHardware-level tuningFlexible deployment options
Expert Support
24/7 expert support from our AI infrastructure specialists with deep knowledge of NVIDIA hardware and AI workloads.
- Dedicated support teamRapid issue resolutionPerformance optimization
EU Data Sovereignty
Deploy in our European data centers with full GDPR compliance and EU data sovereignty guarantees.
- EU-based infrastructureGDPR complianceData residency control
The Polarise stack
- Rise built on Drive
- Drive built on Core
- Core built on bare metal, for managed GPU rental
- Bare metal the servers we operate
- Colocation space, power & cooling for your hardware
- Datacenter infrastructure built & owned by us
- Real estate rented or bought by us
Our AI Factories
Explore where Polarise operates AI-ready infrastructure — Oslo, Munich, and Augsburg.
No locations match these filters. Try turning off a product filter.
NVIDIA Preferred Partner
Polarise has achieved NVIDIA Preferred Partner status and is listed as official NVIDIA Cloud Service Provider (CSP), solidifying our position as a trusted leader in cloud innovation. This designation is reserved for select partners who operate large clusters built in coordination with NVIDIA, adhering to a tested and optimized reference architecture.
Let's start your AI project!
I'll help you find the perfect AI Infrastructure solution for your organization.
