TLDR
- Supermicro launched new server systems supporting NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs for enterprise and edge data centers.
- Systems come in 1U and 2U form factors, designed to drop into existing racks without changes to power or cooling infrastructure.
- The portfolio spans three tiers: large-scale AI (up to 8 GPUs), enterprise AI (up to 6 GPUs), and compact edge AI (up to 4 GPUs).
- All systems are NVIDIA-Certified and support workloads including AI inference, virtualization, media transcoding, and cloud gaming.
- The announcement adds to a busy product stretch for SMCI, which has also recently launched AMD EPYC-based blade servers and NVIDIA Vera Rubin platform systems.
Super Micro Computer (SMCI) announced new server systems on Wednesday built around NVIDIA’s RTX PRO 4500 Blackwell Server Edition GPUs, targeting enterprise data centers and edge deployments where space and power are tight.
Super Micro $SMCI unveils the CMX storage server at GTC 2026.
Built on Nvidia $NVDA STX architecture with Vera CPUs, it enables pod level Key Value cache sharing for agentic AI, delivering 5x token throughput for long context and multi turn inference. https://t.co/Vh1Xjf8uEV pic.twitter.com/peU3qkbuCL
— Ticker Wire (@Tickerwire) March 17, 2026
The timing is notable. Supermicro has been on a product push, and this latest launch continues that momentum ā even as the stock has fallen around 31% over the past six months.
The new systems are designed for environments that can’t easily accommodate high-density compute hardware. The RTX PRO 4500 is a single-slot, power-efficient GPU that can operate as low as 165 watts ā low enough for edge locations that previously couldn’t run this class of hardware.
Super Micro Computer, Inc., SMCI
One of the practical selling points here is rack compatibility. The 1U and 2U form factors are built to replace existing CPU-only servers without needing any modifications to rack, power, or cooling setup. That lowers the barrier for enterprises looking to upgrade without a major overhaul.
The portfolio is split into three tiers. Large-scale AI solutions in 4U and 5U form factors support up to 8 GPUs per node. Enterprise AI and data center systems in 1U and 2U formats support up to 6 GPUs. Compact edge AI solutions in short-depth chassis support up to 4 air-cooled GPUs.
All systems are NVIDIA-Certified, meaning they’ve been tested for compatibility with NVIDIA RTX PRO Blackwell GPUs, NVIDIA networking, and NVIDIA software including AI Enterprise and Omniverse libraries.
Target Workloads
The systems are aimed at a broad range of workloads: AI inference, virtualization, data analytics, media transcoding, and cloud gaming. Supermicro is also offering storage solutions that use GPU acceleration for data vectorization and vector database searching.
CEO Charles Liang said the flexible, modular architecture is aimed at helping enterprises “shorten Time-to-Online” ā essentially getting new compute capacity up and running faster.
Supermicro designs and manufactures its products in the United States, Taiwan, and the Netherlands.
A Busy Product Stretch
This launch is part of a broader product push. In recent weeks, Supermicro also launched a MicroBlade platform using AMD EPYC 4005 processors, supporting up to 320 nodes in a 48U rack.
The company revealed one of the first context memory storage servers based on NVIDIA’s STX architecture. It also introduced seven integrated AI Data Platform solutions developed with partners including Cloudian, DDN, and IBM.
SMCI also announced systems for the NVIDIA Vera Rubin platform, built on its liquid-cooling stack, and a collaboration with One Blockchain LLC on AI infrastructure platforms.
On the financial side, SMCI reported revenue growth of 35% over the last twelve months. The company’s market cap sits at around $18.92 billion.
The new Blackwell GPU-based systems are available now.





