Supermicro
GPU & AI Servers

From 8-GPU HGX H200 air-cooled flagships to liquid-cooled AI factory systems — Supermicro builds the broadest GPU server portfolio in the industry. The platform powering the world's largest AI training clusters.

Broadest GPU Portfolio — SXM to PCIe to Liquid-Cooled

Supermicro offers more GPU server configurations than any other OEM — from air-cooled 8U HGX SXM platforms to direct-to-chip liquid-cooled 4U systems to flexible PCIe multi-GPU servers supporting NVIDIA, AMD, and Intel GPU options. This breadth allows organisations to select the exact platform for their workload without over-engineering.

Liquid Cooling Leadership — D2C and Full-Immersion

Supermicro pioneered direct-to-chip (D2C) and full-immersion liquid cooling for GPU servers. Their complete rack solutions integrate CDUs, liquid manifolds, and servers as a single managed system — enabling 30–60kW/rack densities required for frontier AI training. Air-cooled variants (like the SYS-821GE-TNHR) extend liquid-cool-level density to standard DC environments.

Building Block Solutions® — Green Computing

Supermicro's Building Block Solutions® architecture separates server, storage, and networking components so customers purchase only what they need. Combined with industry-leading Titanium-level power supplies (96% efficiency) and intelligent fan management, Supermicro GPU servers achieve lower PUE and total cost of ownership than comparable platforms.

Supply Chain Security — Silicon Root of Trust

Supermicro implements platform security per NIST SP 800-193 — Silicon Root of Trust anchors cryptographic firmware attestation at the hardware level. Secure Boot, cryptographically signed firmware updates, automatic recovery, and supply chain remote attestation protect AI infrastructure from both software and hardware supply chain compromise.

JumpStart — Free Bare-Metal Server Access

Supermicro's JumpStart programme provides qualified customers with free, time-boxed bare-metal SSH/VNC/IPMI access to fully-configured production systems — including SYS-821GE-TNHR, SYS-421GE-TNHR2-LCC, and other AI systems. Customers can validate workloads, benchmark models, and gather performance data before purchase with no commitment.

NVIDIA Reference Architecture — Preferred AI Server Partner

Supermicro is a preferred NVIDIA AI server partner and contributes GPU server designs to NVIDIA's reference architectures including DGX SuperPOD-compatible configurations and NVIDIA AI Enterprise validated platforms. This partnership ensures the earliest access to new GPU generations and validated software stack compatibility.

• Deployable in any standard data centre
• No CDU, liquid manifold, or infrastructure modification
• Suitable for colocation (Tier 3/4) without facility changes
• Typical rack density: 10–30kW per rack
• Lower upfront infrastructure investment

• Higher sustained GPU clock rates — no thermal throttling
• Enables 30–60kW+ rack densities
• Lower PUE — 60–70% of heat removed via liquid
• Requires CDU + rack liquid cooling infrastructure
• Supermicro complete rack solutions available

Supermicro SYS-821GE-TNHR 8U Air-Cooled GPU SuperServer with NVIDIA HGX H100/H200 8-GPU

The SYS-821GE-TNHR is Supermicro's air-cooled 8U flagship GPU SuperServer — housing 8 NVIDIA HGX H100 (80GB) or H200 (141GB) SXM GPUs interconnected by NVLink and NVSwitch. Dual 5th/4th Gen Intel Xeon Scalable processors, up to 8TB DDR5, and Titanium-level red…