ROOT / COMPUTE // AI ACCELERATORS
// Compute · decoded · publication-grade pass

AI Accelerators, decoded.

Every frontier training and inference accelerator, cross-checked against vendor disclosures and third-party teardowns.

CURATED INDEX◊ VERIFIED · multi-source
288 GB
Rubin HBM4
432 GB
AMD MI455X (most)
23.3 TB/s
top bandwidth
~90%
NVIDIA share
// the signal · what the data says
NVIDIA holds ~90% of data-center AI silicon and set the 2026 bar with Rubin (R200): a dual-die, 336-billion-transistor GPU (1.6x Blackwell's 208B) carrying 288 GB of HBM4 across 8 stacks at 22 TB/s, 35 PFLOPS dense FP4 (50 'effective' with sparsity), and NVLink 6 at 3.6 TB/s per GPU — double Blackwell. It ships H2 2026 on every major cloud. AMD's answer, MI455X (MI400 series, Helios rack), actually beats Rubin on memory: 432 GB HBM4 and 23.3 TB/s, with up to 40 PFLOPS FP4. Google (TPU v7 Ironwood) and Amazon (Trainium2) are the credible custom-silicon escapes from NVIDIA margins.
// comparison · every entry, benchmarked & sourced
NameMakerMemory (GB)TypeBandwidth (TB/s)FP4 (PFLOPS)TDP (W)Node (nm)
Rubin R200NVIDIA288HBM4225018003
◊ 336B transistors dual-die; 35 PFLOPS dense / 50 effective FP4; NVLink 6 @3.6 TB/s; ships H2 2026 (TDP est.)
MI455XAMD432HBM423.34015003
◊ MI400 series, Helios rack; most memory + bandwidth of any 2026 GPU
Blackwell Ultra GB300NVIDIA288HBM3e81514004
◊ Mid-cycle refresh of GB200; higher FP4 and memory
GB200 (Blackwell)NVIDIA192HBM3e81012004
◊ Per-GPU; NVL72 wires 72 into one 1.4-EF FP4 accelerator
MI355XAMD288HBM3e81014003
◊ CDNA4, 256 CUs; 10.1 PFLOPS MXFP4; 1400W TBP
MI325XAMD256HBM3e6510005
◊ CDNA3, 2024 gen
TPU v7 IronwoodGoogle192HBM3e7.496003
◊ 9,216-chip pods (~42.5 EF FP8/pod); inference-optimized
Trainium2Amazon96HBM32.965005
◊ Backs Project Rainier for Anthropic; UltraServer of 64
Gaudi 3Intel128HBM2e3.779005
◊ Ethernet-native scale-out; value play
MTIA v2Meta128HBM32.654505
◊ Meta internal ranking/inference
// leaderboard · Memory (GB)
01MI455X432
02Rubin R200288
03Blackwell Ultra GB300288
04MI355X288
05MI325X256
06GB200 (Blackwell)192
07TPU v7 Ironwood192
08Gaudi 3128
09MTIA v2128
10Trainium296
// sources & where to verify
// full index, per-field sourcing & CSV export   [ AUTHENTICATE → ]
◊ Publication-grade pass, mid-2026. Specs cross-checked against the multiple sources listed. Green = best value in column. Footnotes: FP4 figures are dense per-GPU unless noted; 'effective' inference PFLOPS assume Transformer-Engine sparsity. TDP is board/module typical.
← Autonomous Vessels◊ indexAI Training Clusters →