Supermicro and Arm advance compute for the agentic AI era

At COMPUTEX, Supermicro announced a new class of servers designed to meet the rapidly growing compute demands of the Agentic AI era. Powered by Arm’s recently introduced AGI CPU, these systems deliver industry-leading compute density and power efficiency for next-generation AI inference and agentic workloads.

AI infrastructure is entering its inference era

Since the launch of ChatGPT in late 2022, AI infrastructure conversations have largely centered around GPUs. Datacenter expansion over the past several years has been driven by the race to deploy more accelerated compute for large-scale model training. However, the AI landscape is evolving quickly. Unlike first-generation AI deployments that focused primarily on model training, agentic AI workloads are persistent, distributed, and inference-driven. They require systems capable of handling orchestration, retrieval, reasoning, and real-time decision making at scale.

This shift is driving a new wave of infrastructure requirements where efficient CPU compute plays a foundational role in maximizing overall AI system performance. As workloads shift from training to inference — and increasingly toward autonomous, multi-step agentic AI systems — CPUs are becoming a critical component of modern AI infrastructure. 

Agentic AI introduces a fundamentally different compute profile. Unlike traditional chatbot-style interactions, agentic systems continuously orchestrate reasoning, memory access, retrieval, planning, and communication across multiple services and models. These workflows generate massive demand for highly efficient general-purpose compute, memory bandwidth, and I/O scalability alongside GPU acceleration.

To address this shift, Arm introduced the AGI CPU in March 2026. Built with up to 136 Arm Neoverse V3 cores, 12 DDR5 memory channels at up to 8800 MT/s, and PCIe Gen6 connectivity within a 300W power envelope, the AGI CPU is designed to deliver exceptional compute density and energy efficiency for AI-first data centers. Arm AGI CPU with leading performance per core combined with high core density, high memory bandwidth per core and industry leading power efficiency enables up to 2x higher performance per rack to comparable x86-based solutions, according to Arm estimates.

Purpose-built infrastructure for next-generation AI workloads

Supermicro’s new server and rack-scale portfolio brings the AGI CPU capabilities to market across cloud, enterprise, and edge deployments. 

For hyperscale and neocloud AI infrastructure, Supermicro unveiled the liquid-cooled Open Rack Wide (ORW) platform, the ARS-142TP-QNR-LCC. A fully populated ORW rack can support up to 336 AGI CPUs, enabling massive compute density for cloud-scale agentic AI and inference workloads.

For customers adopting Open Rack V3 (ORV3) environments, Supermicro also introduced the liquid-cooled 2U4N ORV3 ARS-242TP-QNR-LCC server, enabling up to 168 AGI CPUs per rack while maintaining deployment flexibility for modern datacenters. Both the ORW and ORV3 systems are targeted for sampling in Q1 2027, with production availability in Q2 2027.

Supermicro is also extending AGI CPU support into air-cooled environments. For edge deployments with constrained power and space requirements, the single-socket ARS-212HE-FNR short depth server provides an optimized platform for distributed AI inference and edge computing applications. The system is targeted to sample in Q4 2026 and reach production in Q1 2027.          

For general-purpose compute workloads, the dual-socket 2U ARS-222H-NR server supports up to 8 NVMe drives and additional accelerator expansion in a standard 19-inch form factor. These servers are ideally suited for a wide variety of data center workloads such as web and application serving, databases and analytics, virtualization and cloud infrastructure, and media and content processing applications.

Meanwhile, the 5U ARS-522GP-NR platform targets high-performance AI inference deployments with support for up to eight accelerator cards alongside dual AGI CPUs and high-density NVMe storage. These platforms are targeted to sample during Q3 ’26 and released to production in Q1 ’27.

Together, these platforms highlight an important industry transition: the future of AI infrastructure will not be defined by GPU performance alone. As agentic AI scales across enterprises and cloud providers, balanced architectures that combine high-performance CPUs, accelerators, memory bandwidth, and efficient system design will become essential.

At the same time, power efficiency and datacenter scalability are becoming increasingly critical. As enterprises look to deploy AI broadly across cloud, enterprise, and edge environments, infrastructure must deliver higher compute density without unsustainable increases in power and cooling requirements. This is where platforms built around the AGI CPU can provide a significant advantage by enabling scalable AI compute with improved performance-per-watt.

With this portfolio based on the AGI CPU, Supermicro is helping customers build AI infrastructure optimized for the realities of agentic computing — from hyperscale inference clusters to enterprise and edge deployments. As the industry moves toward AI systems that can autonomously reason, collaborate, and act, the combination of efficient CPU compute and accelerated AI infrastructure will form the backbone of the next generation of datacenters.