< Back to all clusters
[TECHNOLOGY] · United States · 13 sources

started · updated

Intel unveils new AI architectures for data centers and edge devices

At the Hot Chips 2026 conference, Intel unveiled three major architectures designed to support the rise of agentic AI across cloud, enterprise, and edge environments.

For high-performance enterprise workloads, Intel introduced Diamond Rapids, a next-generation Xeon processor. Built on the Intel 18A-P process, it features up to 256 cores, 1.28 GB of last-level cache, and supports PCIe Gen6 and CXL 3.0. The architecture utilizes Foveros Direct 3D packaging and UCIe-S interconnects to manage complex AI agent tasks.

To optimize real-time AI inference in data centers, Intel presented Crescent Island, a dedicated GPU. This 350W air-cooled PCIe card is based on the Xe3P architecture and includes 32 Xe cores and 256 XMX engines. It offers high memory capacity, with configurations reaching up to 480 GB of LPDDR5X memory, aiming to improve the economic efficiency of large-scale AI model deployment.

For the client and edge markets, Intel announced Wildcat Lake, part of the Core Series 3. This SoC is designed for cost-effective AI performance in laptops and edge devices. It utilizes the Intel 18A process and is the company's first processor to implement the UCIe interconnect standard. It includes an NPU providing 17 TOPS of AI capability.

Separately, AMD announced internal projections stating that its rack-scale AI systems will achieve a fourfold increase in energy efficiency per watt by 2026 compared to 2024 levels.

Entities

AMD · Crescent Island · Diamond Rapids · Hot Chips 2026 · Intel · Nvidia · Pushkar Ranade · Wildcat Lake

Claims

What the coverage asserts, and how many sources carry each claim.