Supermicro unveiled a desktop AI workstation powered by NVIDIA’s GB300 Grace Blackwell Ultra superchip at CES 2026, delivering over 5x the AI processing power of traditional PCIe GPU systems. The Super AI Station (ARS-511GD-NB-LCC) targets AI developers, startups, and research labs who need on-premises computing without relying on cloud infrastructure or server clusters. The announcement also includes new edge AI systems powered by Intel, AMD, and NVIDIA technologies.
What’s New
Supermicro’s Super AI Station brings server-grade NVIDIA GB300 Grace Blackwell Ultra Desktop superchip into a deskside liquid-cooled form factor. The system delivers more than 5 AI petaFLOPS of computing power with 775GB of coherent memory, enabling local training and inference for models up to 1 trillion parameters. This marks the first desktop implementation of NVIDIA’s GB300 architecture outside data center racks.
The company also introduced four additional product lines. The SYS-542T-2R workstation features Intel Xeon 6 SoC processors with dual 100GbE connectivity and hardware media transcoding for broadcast and streaming applications. A new AI PC (AS-C521D-11302U) uses AMD CPUs in a slim design optimized for office and personal AI workloads. Three edge AI systems based on AMD EPYC 4005 processors arrive in 1U, mini-1U, and slim tower configurations with up to 16 cores. Finally, a fanless compact system (SYS-E103-14P-H) powered by Intel Core Ultra Series 3 delivers 180 platform TOPS for robotics and edge deployments.
All systems were showcased in Las Vegas on January 6, 2026, as part of Supermicro’s CES presence. The products are available for order through Supermicro’s distribution channels, though specific pricing and regional availability were not disclosed in the announcement.
Why It Matters
Desktop AI computing has been constrained by power, cooling, and memory limitations compared to cloud or data center alternatives. Supermicro’s liquid-cooled approach addresses these barriers by packaging data center-class hardware into a workstation form factor. Organizations concerned about data privacy, cloud costs, or network latency can now run frontier AI models locally without building traditional server infrastructure.
The 775GB coherent memory capacity supports massive language models and multimodal AI applications that previously required distributed systems. Research institutions and startups lacking access to hyperscale infrastructure gain a competitive option for AI development and prototyping. Supermicro CEO Charles Liang stated the products “deliver unprecedented performance and energy efficiency, empowering the next generation of users including creators and developers”.
Technical Breakdown
Super AI Station Core Specs:
- NVIDIA GB300 Grace Blackwell Ultra Desktop superchip
- 5x AI petaFLOPS vs. PCIe GPU workstations
- 775GB coherent memory
- Liquid-cooled, self-contained design
- Supports trillion-parameter models locally
Edge and Client Systems:
- Intel Xeon 6 workstation: Media transcoding, 100GbE networking
- AMD AI PC: Slim form factor, minimalist design
- AMD EPYC 4005 edge: 1U/mini-1U/tower options, 16 cores
- Intel Core Ultra edge: Fanless, 180 TOPS, 12 Xe GPU cores
The GB300 superchip uses NVIDIA’s N4P process with 208 billion transistors and 20,480 CUDA cores. It supports FP4 precision compute, enabling efficient inference for quantized models. The NVLink interface provides 10TB/sec bandwidth between dual GPU dies, while HBM3e memory delivers 8TB/sec per GPU.
What’s Next
Supermicro has not announced specific ship dates beyond “available for order” status. The company is simultaneously expanding its NVIDIA Blackwell portfolio with liquid-cooled HGX B300 systems for hyperscale deployments. Analysts project 65% revenue growth for Supermicro in fiscal 2026, partially driven by AI infrastructure demand.
The desktop AI market remains fluid as Intel, AMD, and NVIDIA compete across performance tiers and price points. Supermicro’s strategy focuses on multi-vendor support, offering NVIDIA, Intel, and AMD configurations across its new product lines. Future updates may include next-generation processors as chip vendors release 2026 roadmap products, though no commitments were disclosed.
Featured Snippet Boxes
How much does the Supermicro Super AI Station cost?
Supermicro has not disclosed pricing for the Super AI Station (ARS-511GD-NB-LCC). Enterprise AI workstations with similar liquid-cooled designs and data center GPUs typically range from $50,000 to $150,000. Contact Supermicro or authorized distributors for quotes.
Can the Super AI Station run trillion-parameter AI models?
Yes. The 775GB coherent memory and NVIDIA GB300 superchip with FP4 precision support models up to 1 trillion parameters locally. This includes frontier models like DeepSeek-V3.2, Meta Llama 4, and Mistral Large 3.
What makes it 5x faster than traditional GPU workstations?
The NVIDIA GB300 Grace Blackwell Ultra superchip integrates GPU and CPU on a unified architecture with 10TB/sec NVLink bandwidth and 8TB/sec memory bandwidth. Traditional PCIe workstations use separate GPUs with lower interconnect speeds, creating bottlenecks for large AI models.
Who should buy the Supermicro AI Station vs cloud services?
The system targets organizations with data privacy requirements, high cloud costs, or latency-sensitive AI applications. Startups, research labs, and universities without server infrastructure also benefit. Cloud services remain cost-effective for intermittent workloads or teams requiring elastic scaling.
