A new research study claims the Pearl blockchain network performs random matrix calculations rather than genuine artificial intelligence workloads, despite marketing itself as a Proof-of-Useful-Work system.

Key Points

• Researchers estimate the Pearl network consumes 112 megawatts of power while producing zero verifiable AI computation.
• The study found that Pearl’s protocol verifies correct matrix multiplication but fails to confirm if the input data originates from actual AI training or inference tasks.
• Mining activity on the network contributed to a 38% increase in budget GPU rental prices on the marketplace vast.ai since May.
• Analysis of 8,012 mining workers revealed no identifiable machine-learning framework code within the dominant mining binary.
• The research demonstrated that random, non-AI data can successfully earn mining rewards, incentivizing participants to bypass useful work entirely.

Why it Matters

The findings suggest that Pearl functions primarily as a cryptocurrency mining operation rather than a legitimate distributed AI compute platform. This discrepancy creates significant market inefficiencies, as the network's demand for hardware drives up rental costs for independent researchers without providing the promised utility to the AI industry.