Skip to main content
Top10Grid
Google Cloud TPU v6e (Trillium) key art
GADGET PROFILE

Google Cloud TPU v6e (Trillium)

Purpose-built for Google's AI Hypercomputer stack, Cloud TPU v6e (Trillium) delivers 918 TFLOPS of BF16/FP8 compute and 1,638 GB/s of HBM bandwidth per chip while scaling across 256-chip 2D torus pods (234.9 PFLOPS BF16, 102.4 TB/s all-reduce) with 99% multislice scaling efficiency. TPU v6e takes #2 ahead of AMD Instinct MI355X (#3) as the world's most battle-tested non-NVIDIA AI silicon—powering Gemini training and serving with up to 2.5x better training performance-per-dollar than TPU v5p. It trails Blackwell (#1) only because TPUs are exclusive to Google Cloud rather than sold on-premise.

Gadget
Peak Community Rank
#1
Top 3 in 7
Avg Position
#1.6
7 Lists
Community Signal
Community Darling
Prudential & Structural Ledger

Official Stats & Benchmarks

Site Extent
32 GB HBM · 1,638 GB/s per Chip
Site Extent
918 TFLOPS BF16/FP8 · 1,836 TOPS INT8
Site Extent
800 GB/s ICI · 256-Chip 2D Torus Pod
Peer Benchmark Matrix & 1-Tap Head-to-Head Showdown

Peer Contenders Ranked Alongside Google Cloud TPU v6e (Trillium)

Google Cloud TPU v6e (Trillium)

Gadget · Official Profile

Category / TypeGadget
Shared Rankings7 Active Lists
Subject Benchmark
Category / TypeGADGET
Shared Rankings4 Shared Lists
Category / TypeGADGET
Shared Rankings4 Shared Lists
Category / TypeGADGET
Shared Rankings4 Shared Lists
Category / TypeGADGET
Shared Rankings4 Shared Lists
Category / TypeGADGET
Shared Rankings4 Shared Lists
Direct Leaderboard Ballot

Vote on Rankings Featuring Google Cloud TPU v6e (Trillium)

4 Leaderboards · Live Community Order
Community Weigh-In & Debate

Should Google Cloud TPU v6e (Trillium) rank Top 3 in NVIDIA Blackwell vs. Google TPU v6e? Top 10 AI Datacenter Chips & Accelerators?

Ranked #1 on "NVIDIA Blackwell vs. Google TPU v6e? Top 10 AI Datacenter Chips & Accelerators in 2026". Cast your vote to back or challenge this spot.

Be the first to vote on this ranking.