Could it handle projected “global AI traffic”? - If you req…

Could it handle projected “global AI traffic”?
- If you require 1 on‑chain tx per prompt: No. Even 1 billion prompts/day ≈ 11,600 TPS 24/7—already beyond what a single L2 sustainably offers today, and large multiples beyond the shared DA cap if everyone did it.
- With the recommended design (signed prompts off‑chain + batching): Yes. Example:
- Post 1 Merkle root per 10,000 prompts. If 10,000 agents each do 1M prompts/day, that’s 10,000 × (1,000,000 / 10,000) = 1,000,000 commits/day ≈ 11.6 TPS network‑wide. Spread across a few L2s, this is easy today.
- Session keys and revocations are infrequent (rotation, compromise), so their on‑chain footprint is tiny.