IBM · News · 20261105N

AI Inference Partnership

INTERNATIONAL BUSINESS MACHINES CORP · 2026-11-05 · Importance null · Surprise null

IBM partnered with Supermicro and NVIDIA on a key-value cache storage solution for AI inference workloads. The solution combines Supermicro Petascale storage servers, IBM Storage Scale software, NVIDIA Dynamo inference software, and NVIDIA Spectrum-X Ethernet switches. Reported testing showed reduced time-to-first-token for long-context windows compared with systems without KV cache; the companies said it can improve GPU utilization, throughput, and infrastructure scalability.

Key facts

• IBM, Supermicro, and NVIDIA announced a KV cache storage solution that combines Supermicro Petascale storage servers, IBM Storage Scale software, NVIDIA Dynamo inference software, and NVIDIA Spectrum-X Ethernet switches. source
• IBM, Supermicro, and NVIDIA introduced a high-performance KV cache solution to scale AI inference workloads for enterprise users. source
• A news headline described the IBM and Supermicro AI deal as delivering a 56x speed boost. source
• Performance testing showed the KV cache solution reduces Time-to-First-Token for long-context windows compared with systems without KV cache. source