What it means
FLOPs — floating-point operations per second — is a raw measure of how much arithmetic a chip or cluster can do. It is the headline number used to compare the brute compute of accelerators and is a rough proxy for how big a model can be trained and how fast. But FLOPs alone are misleading: a chip rated for huge throughput still stalls if it cannot be fed data quickly enough or linked efficiently to its neighbours. Real-world AI performance is usually gated by memory bandwidth and networking long before it is gated by raw FLOPs.
Why it matters to investors
Spec-sheet FLOPs make for easy comparisons and loud marketing, but the binding limit in practice is often bandwidth or interconnect, not arithmetic. Investors who look only at peak FLOPs can misjudge where the real value and the real bottlenecks sit — usually one layer over, in memory and networking.
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 FLOPs 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.