Aiconomy

Tensor Core

Specialized processing units within NVIDIA GPUs designed specifically for the matrix multiplication operations that dominate AI computation, delivering massive performance gains for AI workloads.

Tensor cores were introduced with NVIDIA's Volta architecture in 2017. The H100 contains 528 tensor cores capable of performing mixed-precision matrix operations at unprecedented speeds. The Blackwell B200 introduces FP4 tensor core operations, doubling throughput for inference workloads. Tensor cores can process matrix multiplications 8-16x faster than standard CUDA cores for AI workloads. They support multiple precision formats (FP64, TF32, FP16, INT8, FP4) allowing developers to trade precision for speed. Tensor cores are the key hardware innovation that made training trillion-parameter models feasible.

Explore the Data

AI Economy Pulse

Every Friday: the 3 AI data points that actually matter this week. Free, forever.

Built on data from Stanford HAI, IEA, OECD & IMF

Latest: “AI Investment Hits $42B in Q1 2026 — Here's Where It Went”

No spam, ever. Unsubscribe anytime.