AI cryptomining network's 320,000 RTX 3090-class GPUs allegedly burn 112 megawatts of power on ‘zero useful AI computation’

A new research preprint reveals that Pearl, a Layer-1 blockchain claiming to transform cryptocurrency mining into useful artificial intelligence computation, actually performs no meaningful AI work. T... Read original →

Pearl’s so-called 'Proof-of-Useful-Work' is merely GPU arbitrage: its 320,000 RTX 3090-class units perform meaningless matrix ops, exposing a fundamental trust gap in AI-blockchain convergence. Technically, this will accelerate the demise of cuPOW-like protocols, forcing adoption of verifiable AI task provenance. It also inflates GPU rental costs, crowding out legitimate researchers. Regulatory scrutiny—especially under the EU’s Digital Product Act and U.S. SEC green-mining guidelines—may classify such networks as high-risk energy consumers. NVIDIA could tighten driver licensing to block uncertified AI workloads, while AMD might leverage open ecosystems to capture edge compute share. Over the next 12–24 months, fake AI mining will lose investor favor; only PoUW designs with end-to-end verifiable AI workflows will survive—otherwise, it’s just SHA-256 in disguise.