Inference is becoming the largest compute market and energy consumer in AI. Pearl turns inference CapEx of hyperscalers into a profit center: every LLM token produced by GPUs can simultaneously generate ¶PRL in parallel. That means users get 2-for-1 economics: useful inference