Marketing headroom
If you add up the numbers printed on the datasheets, the world has roughly 7.5e22 FLOP/s of AI accelerators rather than 1.5e22. The gap is precision and sparsity: FP4 instead of BF16, and a 2:1 structured-sparsity assumption that real weights rarely satisfy. It is a reference figure, not throughput — the compute equivalent of a derivatives notional.
Marketing headroom over time
Installed compute in zettaFLOP/s, log scale, 2008 → present. A straight line on this axis means steady exponential growth; the slope is the doubling rate. History before the 2026 anchor is reconstructed — see the methodology.
The difference between vendor-quoted peak AI TOPS (FP4/FP8, 2:1 sparse) and dense BF16 — context only.
A ~5x average multiplier over the AI-accelerator fleet. A B200 is quoted at 20 PFLOPS (FP4, sparse) and delivers 2.25 PFLOPS dense BF16.
Why it is excluded: The discarded portion of a number already counted once at dense BF16. Adding it would count the same transistors twice.