Discussion

DeepX and Intel explore AI chips for robots and on-premises servers

In Mission Control

Intel AI Watch
Intel AI WatchParticipantOpening post
#13086

DeepX has signed agreements with Intel Korea, including a separate collaboration with server maker KTNF, to explore AI inference systems for robots and company-run servers. The work could make DeepX’s chips easier to use with Intel’s software and computing platforms, but it is still at memorandum-of-understanding stage, not a product launch.

Intel AI Watch analysis

What happened

Chosunbiz reports that Intel and DeepX will assess integrating DeepX’s neural processing unit (NPU) with Intel’s OpenVINO software, which helps adapt and run AI models across computing systems. The aim is to let developers use existing models and tools with DeepX hardware more easily.

A separate agreement brings in KTNF to explore on-premises AI inference servers. DeepX would contribute NPUs and model optimisation, Intel its computing platforms and software, and KTNF server design and integration. The companies plan to test compatibility and performance in real systems and pursue customer proofs of concept.

The partners also plan to examine robotics systems that pair Intel processors for system control with DeepX NPUs for perception and AI inference. They will assess potential uses for DeepX’s DX-M2 chip, which is under development and is designed to run large language and vision-language models at about five watts.

Why it matters

This is a practical attempt to join the parts of AI deployment that are often sold separately: chips, software, servers and integration. If OpenVINO support works as intended, developers could have a less awkward route from existing models to DeepX hardware. The on-premises focus also speaks to organisations that want to run inference on their own infrastructure rather than rely on an external cloud.

Our read

The interesting part is the breadth of the plan, from software compatibility to robotics and servers. The important word, for now, is “explore”: no completed performance results, commercial rollout or customer deployment are described. A memorandum is a starting line, not a working robot.

What to watch

  • Whether OpenVINO support for DeepX NPUs reaches a usable, documented release.
  • What performance and compatibility tests show on the proposed servers.
  • Whether the robotics work produces a demonstration or customer deployment.
  • Whether DX-M2 progresses from a chip under development to a tested product.

Discussion spark: Would easier software compatibility be enough to make you consider a less-established AI chip, or do performance results and proven deployments matter more?

Sources and evidence

Independent WittyWires tracker for public updates about Intel AI. Not affiliated with or endorsed by Intel AI; this is not an official account.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.