SMCI partners with Arm to launch AI servers with 2x rack performance

Super Micro Computer's new Arm AGI CPU servers promise more than 2x the performance per rack, threatening to reshape the economics of enterprise AI data centers.

Super Micro Computer Inc. on Tuesday unveiled rack-scale AI servers built around Arm Holdings' new AGI CPU, claiming more than 2x the performance per rack and as much as $10 billion in potential capital expenditure savings per gigawatt of AI capacity. The announcement at Computex 2026 in Taipei sent SMCI shares up 4.6% in pre-market trading.

"Supermicro continues to lead the industry when it comes to deploying new and innovative rack-scale solutions that maximize performance and efficiency," Charles Liang, chief executive officer of Supermicro, said. The company's Data Center Building Block Solutions technology stack, combined with the Arm AGI CPU microarchitecture, helps enterprises realize significant total cost of ownership savings on agentic AI infrastructure investments, he said.

The Arm AGI CPU features a 136-core microarchitecture with 6GB of memory bandwidth per core and support for up to 6TB of DDR5-8800 RDIMMs per node. A single air-cooled rack can support more than 6,000 cores, while the company's OCP ORW rack configuration packs 336 Arm AGI CPUs and 45,696 cores into a 48U form factor. Supermicro introduced five server models: a 2U Hyper server for cloud and memory-intensive workloads, a 5U GPU server supporting up to eight double-width GPUs, a 2U4N liquid-cooled server for OCP environments, a 2U Hyper-E edge server, and a 1U 4N configuration for maximum compute density. Arm did not disclose the test conditions for the 2x performance comparison against traditional architectures.

The partnership positions Supermicro to capture a larger share of the fast-growing enterprise AI infrastructure market, where hyperscalers and cloud providers are racing to deploy energy-efficient computing. SMCI shares have surged more than 100% year-to-date, outpacing the broader Zacks Computer & Technology sector's 17.5% gain. The company faces competition from Hewlett Packard Enterprise, which recently launched its Compute Scale-up Server 3250 for business-critical workloads, and from its own AMD-based Helios platform also announced at Computex. But the Arm partnership opens a new front in the battle for data center efficiency, targeting enterprises that prioritize performance per watt over raw compute.

The new servers target agentic AI workloads — systems that can make autonomous decisions and orchestrate thousands of parallel tasks across distributed infrastructure. Mohamed Awad, executive vice president of the cloud AI business unit at Arm, said the combination of Arm AGI CPUs with Supermicro's rack-scale expertise enables "higher AI throughput, maximum compute density, and improved data center economics at scale."

Supermicro's timing aligns with a broader industry shift toward energy-efficient AI infrastructure. Vertiv, a provider of data center power and cooling equipment, reported first-quarter organic sales growth of 44% in the Americas and 12% in Asia-Pacific, reflecting surging demand for data center buildouts. Vertiv expects high-30s organic growth in the Americas and mid-20s in Asia-Pacific for the full year.

For investors, the question is whether Supermicro can maintain its momentum against well-funded rivals. The company's shares trade at roughly 25x forward earnings, a premium to the broader tech sector, reflecting expectations that AI infrastructure spending will remain elevated. Supermicro guided second-quarter revenue between $3.25 billion and $3.45 billion, representing organic growth of 20% to 24%. The Arm-based lineup gives the company a differentiated offering in a market where hyperscalers are increasingly seeking alternatives to traditional x86 architectures to manage power constraints and total cost of ownership.

This article is for informational purposes only and does not constitute investment advice.