What it means
Reasoning models are trained to produce intermediate steps, or a chain of thought, before committing to an answer. Instead of responding in a single pass, they generate many internal tokens that explore, check and refine a solution, a pattern often called test-time compute. This makes them stronger on hard tasks like math, coding and multi-step logic, but each query now costs far more tokens and far more accelerator time than a simple completion. In the AI supply chain they sit in the workloads layer, and their rise shifts demand from one-time training toward continuous inference. That reshapes what data centers are optimized for: serving long reasoning traces cheaply becomes as important as raw model quality, putting pressure on throughput, memory and cost per token.
Why it matters to investors
Reasoning models change the balance between training and inference demand, pushing more of the compute bill into everyday serving rather than one-off training runs. That benefits inference-heavy infrastructure and raises the stakes on cost per token for model builders like DeepSeek, xAI and Moonshot AI.
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 Reasoning model 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.