TLDR
- Supermicro debuts industry-first CMX storage server at NVIDIA GTC 2026.
- New server reduces recompute, speeds AI data access for enterprise workloads.
- Built on NVIDIA STX, CMX integrates storage and compute for faster queries.
- Partnerships with Micron, Samsung, Phison ensure top-tier flash performance.
- Full AI platform portfolio spans cloud, on-prem, and hybrid deployments.
Super Micro Computer, Inc. (SMCI) stock closed at 31.51, down 1.10%, as the company announced a new storage server built on NVIDIA’s STX architecture. The company revealed the context memory (CMX) storage server at NVIDIA GTC 2026, marking a major product milestone in AI infrastructure. This launch highlights Supermicro’s focus on high-performance storage technologies designed to support evolving AI data demands.
Super Micro Computer, Inc., SMCI
Industry’s First CMX Storage Server Integrated with NVIDIA’s STX Rack‑Scale Storage Architecture
Super Micro Computer introduced the industry’s first context memory storage server using the NVIDIA STX rack‑scale architecture at GTC 2026. This product integrates storage and processing capabilities, and it targets workloads that require fast access to intermediate data. The server aims to reduce recompute power needs, while maintaining fast query turnaround for long‑lived AI tasks.
The CMX server design stems from growing enterprise demand for rapid AI data access and efficient storage utilization. Supermicro engineered this system to retain prior tokens and intermediate results, and thereby cut unnecessary recomputation cycles. Moreover, the architecture supports the key value (KV) cache, which NVIDIA’s Dynamo inference layer manages to streamline data flows.
Supermicro’s storage solution is part of the broader STX rack‑scale initiative that NVIDIA launched at the same event. The STX design aims to modularize AI storage and computation layers to accelerate varied data workloads. As a result, systems built on STX offer enhanced flexibility for enterprises building AI factories.
Supermicro plans extensive testing and validation efforts with key software partners to optimize the CMX server. Partnerships span multiple independent software vendors working to ensure compatibility and performance. Concurrently, Supermicro will collaborate with SSD manufacturers to meet the rigorous storage requirements of the STX architecture.
The company’s engagement with SSD suppliers such as Micron, Samsung, and Phison underscores its supply chain strength. These suppliers bring high‑performance flash storage that aligns with demanding AI workloads. Thus, Supermicro aims to deliver robust solutions that scale across enterprise environments.
Broader Platform Announcements at NVIDIA GTC 2026
At GTC 2026, Supermicro also showcased seven new AI data platform solutions using the RTX PRO 6000 Blackwell Server Edition GPU. These platforms include joint efforts with storage and AI partners, like IBM, Nutanix, and VAST Data. Each solution addresses specific data processing needs for enterprise AI pipelines, and they span cloud, on‑premises, and hybrid deployments.
The expanded portfolio shows Supermicro’s commitment to building full stack solutions that integrate compute, storage, and networking. The offerings deliver optimized performance for diverse AI training, inference, and data pre‑processing tasks. Consequently, Supermicro positions itself as a supplier of complete infrastructure for AI workloads.
Supermicro displayed the new CMX server at booth #1113 and at the NVIDIA exhibit during GTC 2026. Visitors could experience live demonstrations that highlight performance gains compared to legacy storage systems. The showcase emphasized real‑world use cases where rapid data access and lower energy use matter.
Super Micro Computer continues to align product innovation with the accelerating needs of data‑driven industries. The combined strengths of STX architecture and context memory storage seek to set new performance benchmarks. As enterprises adopt more complex AI applications, Supermicro’s advancements aim to support scalable, efficient compute and storage ecosystems.


