Supermicro unveils context memory storage server with NVIDIA STX architecture
Super Micro Computer Inc. (NASDAQ: SMCI) announced it has developed one of the first context memory storage servers built on NVIDIA's new STX reference architecture, according to a company statement released March 17.
The CMX storage server combines NVIDIA's Vera CPU and ConnectX-9 SuperNIC technologies. NVIDIA announced the STX modular reference architecture at its GTC 2026 conference, designed to accelerate AI applications.
The storage server addresses challenges with long-lived AI queries and multi-stage agentic workloads that require access to prior and intermediate tokens associated with user queries. The solution stores these tokens, called Key Value cache, which is managed by NVIDIA Dynamo, the company's inference orchestration layer.
"Building upon last year's introduction of the Petascale JBOF, where we proved the feasibility of a JBOF powered by NVIDIA BlueField-3 DPUs, we have developed the CMX storage server," said Charles Liang, president and CEO of Supermicro.
Supermicro said it will work with software partners on porting and validation as the STX solution reaches market. The company plans to leverage relationships with storage providers including Micron, Samsung, and Phison for testing specific to STX architecture requirements.
At GTC 2026, Supermicro also announced seven AI Data Platform solutions based on RTX PRO 6000 Blackwell Server Edition GPU with NVIDIA and storage partners including Cloudian, DDN, Everpure, IBM, Nutanix, VAST Data, and WEKA.
The CMX server is being displayed at Supermicro's booth at the NVIDIA GTC 2026 conference running March 16-19.
