Purpose-built for the pinnacle of AI performance
Engineered to be the heart of large-scale AI infrastructure, Cisco AI GPU servers enable training, fine-tuning, and deployment of massive models fast, efficiently, and at unprecedented scale.
Largest-model optimization
Get exceptional GPU parallelism and bandwidth for cutting-edge use cases, from LLMs, deep learning, and generative AI to scientific simulations.
Architectural choice
Select NVIDIA HGX for NVLink-connected GPU meshes or choose AMD OAM MI300/MI350X accelerators for modular GPU density—both fully integrated into enterprise-ready Cisco platforms.
Rack-scale efficiency
Maximize performance per square foot of data center space with server designs that carefully balance power, cooling, and form factor.
Unified operations
Cisco dense AI GPU servers are integrated with Cisco Intersight for cloud-based lifecycle management, monitoring, and optimization across your entire AI cluster.
Choose your dense AI GPU platform
C880A M8 with NVIDIA HGX B300
AI super computing platform
Accelerate AI reasoning and training with the Blackwell Ultra-powered HGX B300 in the Cisco UCS C880A M8.
- Built on dual-reticle Blackwell Ultra architecture
- Up to 11x inference and 4x training speed versus previous generation
- NVLink-5 / 1.8 TBps GPU interconnect for massive scale
- Up to 288 GB HBM3E per GPU for trillion-plus-parameter models
- Next-gen precisions (NVFP4, FP8/FP6) and transformer engine enhancements
C885A M8 with NVIDIA HGX
NVLink-enabled GPU mesh
Ultra-dense AI server with NVIDIA HGX and NVLink for maximum GPU-to-GPU bandwidth in large-scale AI training.
- Supports NVIDIA H100/H200 HGX air-cooled baseboard
- NVLink high-bandwidth GPU interconnect
- Optimized for LLM and generative AI training
- High-density NVMe storage options
- Full Cisco Intersight integration
C885A M8 with AMD OAM
Modular GPU density
High-density AI server with AMD OAM MI300/MI350X GPUs for scalable, modular GPU compute in AI and HPC workloads.
- Supports AMD OAM MI300/MI350X OAM GPUs
- High-throughput GPU-to-GPU interconnect
- Designed for AI, HPC, and simulation
- Balanced cooling and power for dense loads
- Full Cisco Intersight integration
Cisco dense AI GPU servers support two leading accelerator frameworks: NVIDIA HGX for ultra-high bandwidth GPU meshes and AMD OAM MI300/MI350X for flexible AMD GPU clusters. Both are designed for tight integration, performance, and enterprise-grade management.
- NVIDIA HGX – GPU mesh excellence: NVLink-enabled HGX modules deliver unmatched inter-GPU throughput—ideal for training large-scale models with full-bandwidth interconnect.
- AMD OAM MI300/MI350X – Modular GPU compute: Open Accelerator Modules from AMD give you modular, interchangeable GPU units with dedicated NVLink-equivalent bandwidth—optimized for dense AI workloads on AMD.
- Cisco integration – Ready for the rack: Engineered with power, cooling, and Intersight management to ensure seamless deployment from bare metal to full AI rack pod.
Discover the Total Economic Impact™ of Cisco Intersight
Hear how Cisco Intersight improved IT productivity and provided a 192% ROI. Explore the full Forrester TEI findings.
