SemiAnalysis rolls out AI cluster networking data model

Research shop SemiAnalysis has put the physical plumbing of the AI buildout into its own institutional data product. The AI Networking Model page describes it as "device-level tracking of the networks behind AI clusters: switches, transceivers, cables, and AECs across scale-up, scale-across, scale-out, front-end, and out-of-band fabrics."

The scope is broad. It ties together 80+ hyperscaler configuration panels for Microsoft, Google, Meta, Amazon, Oracle, X.AI and neoclouds, matched to specific accelerator SKUs — from H100 and H200 through GB200 NVL36/NVL72, B300, GB300 NVL and VR200, plus AMD's MI300, MI325 and MI350. On the supply side, 25+ vendors are tracked, including Nvidia, Arista, Broadcom, Cisco, Coherent and Lumentum, across transceiver speeds of 200G to 1.6T. Data runs 2023 through 2026.

Under the hood, the model is presented as four reconciled modules: cluster configurations across five network layers, bottom-up builds and spend, top-down market volumes and vendor share, and a master pricing and power table for devices and components.

Delivery is an Excel workbook with dashboard access. API access is reserved for institutional subscribers, and it is sold separately from SemiAnalysis's newsletter.