IT Home, October 13: On October 12 local time, AMD unveiled its next-generation Verano series EPYC data center processors for the first time at the 2026 OCP Global Summit.
The series of processors, scheduled for release in 2027, will feature the Zen 6 architecture and, for the first time in the EPYC lineup, support LPDDR5X memory via SOCAMM2 memory modules to deliver edge computing capabilities for agent-based AI data centers.
AMD CEO Lisa Su referred to Verano as AMD’s next step in its CPU product strategy for the AI era. The processor is primarily designed for AI accelerator host nodes and is planned to be paired with future generations of AMD Instinct GPUs to meet the host-side computing and memory system demands of agent-based AI workloads.
In terms of memory, Verano will adopt a solution combining LPDDR5X with the SOCAMM2 module. AMD describes SOCAMM2 as a modular memory specification designed for server environments, offering replaceability and ease of maintenance while retaining the low power consumption and high bandwidth advantages of LPDDR5X. AMD believes this approach helps improve the performance-per-watt ratio of AI data centers and provides greater design flexibility for rack-level systems.
ComputerBase reported that the Verano processor demonstrated by AMD at the OCP summit features a configuration of up to 192 cores. However, AMD’s previously released official documentation has not confirmed this core count, so the specification remains subject to AMD’s final, complete product specifications.
Tom's Hardware previously reported that Verano's configuration could reach up to 72 Zen 6 cores with a 24-channel LPDDR5X memory subsystem; there is currently a discrepancy between these two reports.
In terms of interconnectivity, Verano offers up to 112 Gbps xGMI connectivity for linking Instinct MI500 series GPUs. However, the final core count, interconnect specifications, and full product configuration remain to be officially disclosed.
Suzanne also stated during her speech that AMD expects the server CPU market to exceed $200 billion by 2030 (IT Home note: approximately RMB 1.34 trillion at current exchange rates), with an annual compound growth rate of about 50% over the coming years. She linked this growth projection to the development of agent-based AI workloads, noting that such applications will further increase demand for server CPU computing resources.