Within the plane, each switch is connected in full mesh using 800G 2xFR4 OSFPs in 8x100G breakout mode on the L1 layer a…

SemiAnalysis Twitter · SemiAnalysis (@SemiAnalysis_) · 2026-06-03

SemiAnalysis reveals that Google's AI cluster networking uses 800G 2xFR4 OSFPs in breakout mode for L1 full-mesh connections and 400G FR4 QSFPs for L2 links, with doubled inter-layer links to preserve a non-blocking topology.

Open original ↗

Appears in

Google Reveals TPUv8t and Virgo 134K-Chip Scale-Out Interconnect

Extraction

Topics: ai-infrastructurenetwork-topologygoogle-infrastructureoptical-networking

Claims

Google connects L1 switches in full mesh using 800G 2xFR4 OSFPs operating in 8x100G breakout mode.
Google uses 400G FR4 QSFPs in 4x100G breakout mode on the L2 layer.
Google doubles some links between L1 and L2 layers to maintain a non-blocking network topology.

Key quotes

Within the plane, each switch is connected in full mesh using 800G 2xFR4 OSFPs in 8x100G breakout mode on the L1 layer and 400G FR4 QSFPs in 4x100G breakout mode on the L2 layer.

To maintain a non-blocking topology, Google is doubling some links between the L1 and the L2 layers.