People keep talking about token moves, but the actual bottleneck here is prover throughput, since rollups only get cheap once proof generation stops eating GPU-hours. Before ASICs, most teams were living on GPUs and a lot of hand-tuned math, which works until the circuit mix changes and the economics get ugly.

No tape-out date, and somehow we are already pricing the chip like it shipped. That is a very crypto way to do hardware. I still want the first post that compares porver economics before and after an ASIC, not the token chart.