Maximum AI compute power, built on a foundation you can trust.
Fujitsu PRIMERGY GX2580 M8s brings 8× NVIDIA HGX B300 GPU acceleration, frontier NVLink bandwidth, and Kasashima-factory manufacturing quality to Saudi and GCC AI infrastructure.
AI moves faster than traditional infrastructure can support.
AI performance without compromise
8× NVIDIA HGX B300 GPUs, 5th Gen NVLink® at 1.8TB/s, dual Intel® Xeon® 6700P processors, and up to 8TB DDR5 memory.
Trust built in
Manufactured at Fujitsu's Kasashima factory, designed for EU regulatory adherence, NIS2 compliance efforts, and hyperscaler independence.
Enterprise lifecycle by design
Six redundant Titanium-level power supplies, full ISM management, and uSCALE-ready consumption.
Compact infrastructure. Frontier capability.
GX2580 M8s combines dense GPU compute, high-speed interconnect, and simplified expansion for LLM training, generative AI, deep learning, and HPC workloads.
8× NVIDIA HGX B300 GPUs
Exceptional compute density and memory bandwidth for large-batch distributed training.
5th Gen NVLink®
1.8TB/s GPU-to-GPU interconnect keeps every GPU fed at full throughput.
2× 25 Gbit/s SFP28 LAN
Reliable, high-speed connectivity for cluster and management networking.
Dual Intel® Xeon® 6700P
P-core processors with up to 8TB DDR5 memory across 32 DIMM slots.
A balanced path from workload to acceleration.
GX2580 M8s integrates compute, high-speed networking, and GPU acceleration into one enterprise platform, sized so no subsystem bottlenecks the others.
LLMs & generative AI
Training, fine-tuning, inference, deep learning, and HPC simulation.
High-speed access
2× 25 Gbit/s SFP28 LAN · up to 8× E1.S NVMe storage bays
GX2580 M8s
8× NVIDIA HGX B300 GPUs, 5th Gen NVLink®, 2.3TB HBM3e, dual Xeon 6700P.
Engineered for sustained full-load operation.
Stable performance under continuous full-load operation, supported by six redundant 6,600W Titanium-level (96%) power supplies — built for long-running training cycles rather than short bursts.
Trust starts at the factory floor.
Organizations need infrastructure they can trust — not just in performance, but in origin, process, and accountability.
Transparent supply chain
Manufactured and assembled at the Fujitsu Kasashima factory, Japan, with a traceable origin from build to deployment.
Japanese manufacturing quality
Meticulous manufacturing processes and enterprise-grade reliability standards on every system.
Regulatory adherence
Supports digital sovereignty strategies, NIS2 compliance efforts, and reduced dependency on opaque global supply chains.
Operational resilience
3-year onsite manufacturer warranty and at least 5 years of service lifecycle after shipment.
From experiment to enterprise AI.
Many organizations remain stuck in pilot mode. GX2580 M8s is engineered to close the gap between AI ambition and real-world execution.
Assessment
Understand your AI infrastructure needs
PoC & benchmarking
Validate performance against real workloads
Deployment
Bring GX2580 M8s into production
Scaling
Grow from a single node to a full cluster
Lifecycle
Ongoing management via ISM & PRS
Made for mission-critical AI workloads.
LLMs & Generative AI
Training, fine-tuning, and inference at scale.
Data Science
Computer vision, speech, and multimodal AI.
Scientific Computing
Simulation, physics, and climate modeling.
Enterprise Applications
Finance, healthcare, manufacturing, and research.
Explore the platform.
| Form factor | 8U GPU-accelerated dual-socket rack server |
| Mainboard type | X14DBG-LC |
| Processor | Intel® Xeon® 6700P series processors |
| Memory | Up to 8TB, 32 DIMM slots, DDR5, ECC protected |
| Weight / Rack dimensions | 119 kg · 449 x 950 x 356 mm (W x D x H) |
| GPUs | 8× NVIDIA HGX B300 |
| GPU interconnect | 5th Gen NVLink® — 1.8TB/s |
| GPU memory | Up to 2.3TB HBM3e per system |
| Architecture | NVIDIA Blackwell — optimized for trillion-parameter models |
| LAN controller | 2× 25 Gbit/s SFP28 |
| USB ports | 1 |
| Onboard accessible drives | 2× M.2 NVMe |
| Storage drive bays | Up to 8× E1.S NVMe (8 x 2.5-inch) |
| Power supply configuration | 6× 6,600W redundant (3+3), Titanium level (96%) |
| Server management | Infrastructure Manager (ISM) — mass Linux OS installation, node management, health monitoring, update management |
| DC infrastructure management | Remote management, update management, asset management, online diagnostics |
| Manufacturer warranty | 3 years onsite warranty |
| Service lifecycle | At least 5 years after shipment |
| Recommended service | 24x7 onsite service, 4h onsite response time |
Pay for AI capacity when growth happens.
Cloud-like consumption economics with infrastructure and data remaining inside your own data center.
Financial agility
Reduce upfront CAPEX and align spend with consumed GPU capacity.
Instant scalability
Pre-position buffer GPU capacity before it's needed.
On-premises sovereignty
Keep hardware and data securely within your own data center.
Build your enterprise AI foundation, with local expertise.
ComputingEra supports assessment, sizing, architecture, deployment, and ongoing lifecycle management for PRIMERGY GX2580 M8s across Saudi Arabia and the GCC.
Maximum AI compute power, built on a foundation you can trust.
Fujitsu PRIMERGY GX2580 M8s brings 8× NVIDIA HGX B300 GPU acceleration, frontier NVLink bandwidth, and Kasashima-factory manufacturing quality to Saudi and GCC AI infrastructure.
AI moves faster than traditional infrastructure can support.
AI performance without compromise
8× NVIDIA HGX B300 GPUs, 5th Gen NVLink® at 1.8TB/s, dual Intel® Xeon® 6700P processors, and up to 8TB DDR5 memory.
Trust built in
Manufactured at Fujitsu's Kasashima factory, designed for EU regulatory adherence, NIS2 compliance efforts, and hyperscaler independence.
Enterprise lifecycle by design
Six redundant Titanium-level power supplies, full ISM management, and uSCALE-ready consumption.
Compact infrastructure. Frontier capability.
GX2580 M8s combines dense GPU compute, high-speed interconnect, and simplified expansion for LLM training, generative AI, deep learning, and HPC workloads.
8× NVIDIA HGX B300 GPUs
Exceptional compute density and memory bandwidth for large-batch distributed training.
5th Gen NVLink®
1.8TB/s GPU-to-GPU interconnect keeps every GPU fed at full throughput.
2× 25 Gbit/s SFP28 LAN
Reliable, high-speed connectivity for cluster and management networking.
Dual Intel® Xeon® 6700P
P-core processors with up to 8TB DDR5 memory across 32 DIMM slots.
A balanced path from workload to acceleration.
GX2580 M8s integrates compute, high-speed networking, and GPU acceleration into one enterprise platform, sized so no subsystem bottlenecks the others.
LLMs & generative AI
Training, fine-tuning, inference, deep learning, and HPC simulation.
High-speed access
2× 25 Gbit/s SFP28 LAN · up to 8× E1.S NVMe storage bays
GX2580 M8s
8× NVIDIA HGX B300 GPUs, 5th Gen NVLink®, 2.3TB HBM3e, dual Xeon 6700P.
Engineered for sustained full-load operation.
Stable performance under continuous full-load operation, supported by six redundant 6,600W Titanium-level (96%) power supplies — built for long-running training cycles rather than short bursts.
Trust starts at the factory floor.
Organizations need infrastructure they can trust — not just in performance, but in origin, process, and accountability.
Transparent supply chain
Manufactured and assembled at the Fujitsu Kasashima factory, Japan, with a traceable origin from build to deployment.
Japanese manufacturing quality
Meticulous manufacturing processes and enterprise-grade reliability standards on every system.
Regulatory adherence
Supports digital sovereignty strategies, NIS2 compliance efforts, and reduced dependency on opaque global supply chains.
Operational resilience
3-year onsite manufacturer warranty and at least 5 years of service lifecycle after shipment.
From experiment to enterprise AI.
Many organizations remain stuck in pilot mode. GX2580 M8s is engineered to close the gap between AI ambition and real-world execution.
Assessment
Understand your AI infrastructure needs
PoC & benchmarking
Validate performance against real workloads
Deployment
Bring GX2580 M8s into production
Scaling
Grow from a single node to a full cluster
Lifecycle
Ongoing management via ISM & PRS
Made for mission-critical AI workloads.
LLMs & Generative AI
Training, fine-tuning, and inference at scale.
Data Science
Computer vision, speech, and multimodal AI.
Scientific Computing
Simulation, physics, and climate modeling.
Enterprise Applications
Finance, healthcare, manufacturing, and research.
Explore the platform.
| Form factor | 8U GPU-accelerated dual-socket rack server |
| Mainboard type | X14DBG-LC |
| Processor | Intel® Xeon® 6700P series processors |
| Memory | Up to 8TB, 32 DIMM slots, DDR5, ECC protected |
| Weight / Rack dimensions | 119 kg · 449 x 950 x 356 mm (W x D x H) |
| GPUs | 8× NVIDIA HGX B300 |
| GPU interconnect | 5th Gen NVLink® — 1.8TB/s |
| GPU memory | Up to 2.3TB HBM3e per system |
| Architecture | NVIDIA Blackwell — optimized for trillion-parameter models |
| LAN controller | 2× 25 Gbit/s SFP28 |
| USB ports | 1 |
| Onboard accessible drives | 2× M.2 NVMe |
| Storage drive bays | Up to 8× E1.S NVMe (8 x 2.5-inch) |
| Power supply configuration | 6× 6,600W redundant (3+3), Titanium level (96%) |
| Server management | Infrastructure Manager (ISM) — mass Linux OS installation, node management, health monitoring, update management |
| DC infrastructure management | Remote management, update management, asset management, online diagnostics |
| Manufacturer warranty | 3 years onsite warranty |
| Service lifecycle | At least 5 years after shipment |
| Recommended service | 24x7 onsite service, 4h onsite response time |
Pay for AI capacity when growth happens.
Cloud-like consumption economics with infrastructure and data remaining inside your own data center.
Financial agility
Reduce upfront CAPEX and align spend with consumed GPU capacity.
Instant scalability
Pre-position buffer GPU capacity before it's needed.
On-premises sovereignty
Keep hardware and data securely within your own data center.
Build your enterprise AI foundation, with local expertise.
ComputingEra supports assessment, sizing, architecture, deployment, and ongoing lifecycle management for PRIMERGY GX2580 M8s across Saudi Arabia and the GCC.