A new academic preprint is leveling a striking accusation at Pearl, an AI-focused cryptocurrency mining network: that its sprawling fleet of graphics cards is consuming enormous amounts of electricity without producing any verified artificial intelligence work.
According to Tom's Hardware, the study claims Pearl's network operates the equivalent of 320,000 RTX 3090-class GPUs and draws roughly 112 megawatts of power continuously — enough to supply tens of thousands of homes. The researchers allege the cards are performing "random matrix math" rather than any genuine AI computation, essentially mimicking the appearance of AI workloads without delivering results anyone can verify or use.
The timing is notable. GPU rental prices have already jumped 38%, according to the report, a squeeze felt by researchers, startups, and developers who legitimately need computing power to train and run AI models. If networks like Pearl are soaking up a significant share of available GPU supply while doing nothing productive, that cost pressure lands on real users.
Pearl has not, based on the sources available, publicly responded to the preprint's allegations. The study is a preprint, meaning it has not yet been peer-reviewed, and its findings should be treated as preliminary.
The story matters because it exposes a potential dark side of the GPU gold rush: bad actors may be exploiting AI's credibility to run what amounts to a dressed-up crypto scheme, crowding out legitimate AI work and driving up costs for everyone else in the process.