Verified, then logged.
Sun 22 Nov 2026 · 00:20 KST / 15:20 UTC한국어 | ENSearch /

08 / Industry · Analysis

Intel sets out mixed-computing strategy for agentic AI

Intel proposed combining CPUs, GPUs and dedicated AI accelerators, arguing that GPU-centred infrastructure alone cannot efficiently handle complex agent tasks.

Half-installed server racks line a nearly finished data center hall, with transmission towers on the Central Asian steppe visible through an open loading bay. AI-generated image

Intel outlined a data centre strategy combining central processing units (CPUs), graphics processing units (GPUs) and dedicated AI accelerators for agentic AI, according to AI Times, a Korean AI news publication.

AI Times framed the strategy around a shift from generative AI that answers questions towards agentic AI that plans and carries out tasks. That framing provides the context for Intel’s proposal.

According to the report, Intel argued that GPU-centred infrastructure alone is insufficient to handle complex agent tasks efficiently. Its proposal links that efficiency to the mix of computing devices in a data centre.

Combining different computing devices

AI Times reported that Intel proposed using CPUs, GPUs and dedicated AI accelerators together. The combination forms the core of its infrastructure strategy for agent workloads.

In the explanation relayed by AI Times, Intel treated the choice of computing devices as part of the efficiency challenge posed by agentic AI. The argument concerns how those devices work together to handle complex tasks.

An inference accelerator unveiled in Seoul

AI Times reported that Anil Nanduri, a vice-president in Intel’s Data Center AI Group, unveiled a next-generation inference accelerator at an event at COEX in Samseong-dong, Seoul, on 12 Nov. Inference is the process of running a trained AI model to produce an output.

The infrastructure direction Intel set out, according to AI Times, brings CPUs, GPUs and dedicated AI accelerators together for agent workloads. Its proposal places the combination of computing devices at the centre of how data centres handle those tasks.