What it means
Quantization is a technique for shrinking an AI model by storing and computing its numbers at lower numerical precision, for example moving weights and activations from 16-bit down to 8-bit formats like FP8 or INT8, or lower. Fewer bits per value means the model uses less memory, moves less data, and runs faster, usually with only a small loss of accuracy. It is central to inference serving, the recurring workload of actually running trained models to answer users, because it directly cuts the memory footprint and cost per token while raising throughput on the same hardware. In the AI supply chain it sits in the inference and model-optimization layer, and it is a key lever for making large models economical to deploy at scale. Newer accelerators add native support for low-precision formats, tightening the link between quantization and chip design.
Why it matters to investors
Quantization lowers the memory and cost of serving models, so it directly improves the economics of inference, the workload that monetizes the stack beneath it. It benefits inference providers and the accelerator vendors whose chips add native low-precision support, with players like Together AI, Microsoft, and Google optimizing serving this way.
Companies on this part of the chain
Named to show where the term sits in the AI supply chain — research, not advice, and never a recommendation to buy or sell.
Related terms
See Quantization in the live AI chain.
THE ENTITY maps every constraint onto one live model — which part is tight now, who owns it, and who gets squeezed when it moves. Plain-English reads you can check.
THE ENTITY is an educational read on the AI supply chain — research, not investment advice. It explains how the chain works and who sits where, never price targets or buy/sell calls.