Highlights 

  • Supermicro unveils next-generation AI Factory and HPC systems powered by NVIDIA GB300 and GB200 architectures. 
  • New liquid-cooled rack-scale platforms aim to boost performance while lowering data center energy use. 
  • Advanced cooling innovations include rear-door heat exchangers and 200kW sidecar CDUs. 
  • Expanded SuperBlade, FlexTwin, and BigTwin families target high-density HPC and enterprise AI workloads. 
  • Supermicro reinforces its focus on end-to-end, energy-efficient IT infrastructure from desk-side to hyperscale. 

Super Micro Computer, Inc. (NASDAQ:SMCI) used the global stage of Supercomputing 2025 in St. Louis to spotlight what it sees as the next leap inAIand high-performance computing. The company’s latest Data Center Building Block Solutions® showcased how modern organizations can scale from workstation development environments to full rack-level AI clusters built on NVIDIA’s newest architectures. 
At the center of the display was the NVIDIA GB300 NVL72, a liquid-cooled rack-scale platform integrating 72 NVIDIA Blackwell Ultra GPUs and 36 Grace CPUs per rack, each GPU paired with 279GB of HBM3e memory. Supermicro also highlighted its upcoming 4U HGX B300 liquid-cooled server and the compact 1U GB200 NVL4 system designed specifically for large-scale model training and scientific workloads. 

Liquid Cooling and DCBBS: Reducing Time-to-Online 

Energy-efficient cooling was a major theme for Supermicro at SC25. The company presented an expanded lineup of liquid-ready enclosures, including 8U 20-node and 6U 10-node SuperBlade platforms, as well as the FlexTwin multi-node architecture supporting AMD EPYC 9005 and Intel Xeon 6900 CPUs up to 500W. 
To simplify deployment, Supermicro’s DCBBS approach integrates compute, networking, storage, and cooling into modular, ready-to-deploy clusters. New rear-door heat exchangers offer cooling capacity options of 50kW or 80kW, while the liquid-to-air Sidecar CDU can handle up to 200kW with no external plant infrastructure. These solutions target enterprises looking to speed up cluster commissioning while reducing operational energy use. 

Optimized Product Families for Every Intensive Workload 

Supermicro emphasized how its system families cater to HPC, cloud AI, financial analytics, manufacturing simulation, climate modeling, and semiconductor design. 
The X14 SuperBlade continues to serve as a premium choice for GPU-dense and CPU-heavy workloads with integrated networking options. FlexTwin, meanwhile, enables up to 24,576 performance cores in a single 48U rack for institutions requiring compact but high-capacity compute clusters. 
For cost-optimized density, the MicroBlade and MicroCloud families support up to 40 nodes in 6U or 10 nodes in 3U, enabling efficient scaling for x86 and GPU-augmented workloads. Supermicro also showcased petascale flash storage platforms and rackmount workstations for organizations seeking centralized resources with secure access and high throughput. 

Conclusion 

Supermicro’s SC25 portfolio makes clear the company’s intention to lead the AI and HPC infrastructure market with modular, cooling-efficient, and scalable designs. By pairing NVIDIA’s latest Blackwell-generation hardware with ready-to-deploy DCBBS architectures and advanced liquid-cooling options, Supermicro continues to help enterprises reduce time-to-online while supporting increasingly compute-intensive scientific and AI workloads. 

Also read - Sohu.com (NASDAQ:SOHU) Reports Q3 Results Driven by Online Game Upswing