Decoder. plain-English AI glossary

AI chip

● CoreAI chip

A specialized processor (GPU, TPU, NPU) optimized for training and inferencing neural networks—faster and cheaper than CPUs for AI.

Think of it like

A engine built specifically for speed; a car engine, not a truck engine.

Example

NVIDIA's H100, Google's TPU, Apple's Neural Engine. Each optimized for matrix operations that power neural nets.

How it actually works

AI chips are the bottleneck. Demand >> supply. NVIDIA dominates (80%+ market share). Alternatives exist (TPU, Cerebras, Graphcore) but are niche or closed. Everyone wants to design their own chip for cost/advantage; success is rare.

For product teams

Chip availability and cost directly limit model size and deployment scale.

For engineers

Matrix-optimized; high bandwidth memory; interconnect is critical for distributed training.

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open DecoderAdd to Chrome