Penguin Solutions (PENG) will showcase its AI infrastructure and memory solutions at AI Infra Summit 2026 from September 15-17 at the Santa Clara Convention Center. The company will highlight its MemoryAI KV Cache Server, which uses Compute Express Link (CXL) based memory to help address the AI inferencing “memory wall.”Penguin states that MemoryAI KV Cache Server can deliver up to 11 TB of CXL-based memory to optimize enterprise-scale inference workloads, including agentic AI and large language model inference. At booth #520, Penguin will also demonstrate CXL-based memory expansion servers, add-in cards, the LIQID EX-5410C CXL memory expansion chassis, and an advanced memory technology preview developed with Marvell. A speaking session titled “Scaling the AI Memory Wall – Where Model Architecture Meets Hardware” is scheduled for September 16 at 4:10 p.m. PT, featuring Andy Mills, Vice President of Advanced Product Development.

ペンギンソリューションズ（2026年4月に日本ストラタステクノロジーから社名変更）は2026年6月24日、大規模AI推論用の「MemoryAI KVキャッシュサーバー」を発表した。4Uラックマウントの筐体に11TBのメモリーを積んだストレージキャッシュ装置であり、LLMが推論時に生成するトークンのKVキャッシュを外部ストレージ階層にオフロードする際のI/O性能を高められる。2026年第4四半期に提供を開始する。

ペンギンソリューションズは、AIインフラ向け新製品として、AIファクトリープラットフォーム管理ソフトウェア「ClusterWareAI」、データセンターおよびAI推論向けに特別に設計されたメモリーソリューション「MemoryAI KVキャッシュサーバー」、AIワークロード向けに最適化された高度なコンピューティングシステム「ComputeAI」を日本市場で提供開始する。

Penguin Solutions has announced the launch of innovative AI factory products in Japan, aiming to support rapid deployment and operational stability in large-scale AI implementations.

Penguin Solutions (NASDAQ: PENG) expanded its OriginAI portfolio on March 16, 2026 to optimize enterprise AI inference by adding large-memory appliances and KV cache technology for NVIDIA RTX PRO 6000 and B300 GPU architectures. OriginAI targets lower latency, higher throughput, improved GPU utilization, and scalable deployment with MemoryAI KV cache and ICE ClusterWare software.The offering supports mid-sized and frontier models, long-context assistants, and multi-tenant security, and will be showcased at NVIDIA GTC March 16–19, 2026.

Penguin Solutions (Nasdaq: PENG) introduced the MemoryAI KV cache server, a production-ready CXL-based KV cache appliance delivering up to 11 TB of memory to reduce inference latency and improve GPU throughput for enterprise AI. The server combines 3 TB DDR5 main memory with up to eight 1 TB CXL add-in cards to expand cluster memory, support larger context windows and increase token throughput while claiming ~10x faster access than NVMe-based approaches.The product targets latency-sensitive, memory-bound inference workloads and is compatible with NVIDIA Dynamo; Penguin is showcasing it at NVIDIA GTC, March 16–19, 2026.