---
title: "Discussion on \"64×800G vs. 128×400G: Designing a 51.2T Network for AI Clusters\" | Hashnode"
description: "Discussion on \"64×800G vs. 128×400G: Designing a 51.2T Network for AI Clusters\". Learn how to evaluate 64×800G and 128×400G 51.2T switches for AI clusters, including topology, optics, breakout compatibility, and deployment costs."
url: "https://hashnode.com/posts/64-800g-vs-128-400g-designing-a-51-2t-network-for-ai-clusters/6ad9cef4153ceb6522d9d845"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [64×800G vs. 128×400G: Designing a 51.2T Network for AI Clusters](https://aicplight.hashnode.dev/64-800g-vs-128-400g-designing-a-51-2t-network-for-ai-clusters?utm_source=hashnode&utm_medium=feed)

Learn how to evaluate 64×800G and 128×400G 51.2T switches for AI clusters, including topology, optics, breakout compatibility, and deployment costs.

Oct 22 · 8 min read · aicplight.hashnode.dev  
Read the article

-   [#networking](https://hashnode.com/tag/networking)
-   [#ai](https://hashnode.com/tag/ai)
-   [#data-center](https://hashnode.com/tag/data-center)
-   [#ethernet](https://hashnode.com/tag/ethernet)

## Written by

[Weiwei Zhang](https://hashnode.com/@aicplight)

@aicplight

  
[View profile](https://hashnode.com/@aicplight)

## Responses

Markdown supported · ⌘↵ to post

No responses yet.
