Preprint Alleges Pearl's AI Network Consumes 112 MW with No Useful AI Computation
A recent preprint research paper claims that Pearl’s AI mining network is utilizing an equivalent of 320,000 RTX 3090-class GPUs and consuming 112 megawatts of power. The study alleges that this substantial infrastructure produces no verified useful artificial intelligence computation. Instead, the GPUs are reportedly performing random matrix math, and the research also notes a 38% increase in GPU rental costs.
A new preprint study alleges that Pearl's AI mining network consumes significant power and computing resources without performing useful artificial intelligence computations. The research claims the network utilizes the equivalent of 320,000 RTX 3090-class GPUs.
This extensive GPU infrastructure is reportedly consuming 112 megawatts of power. Despite this considerable energy expenditure, the preprint asserts that the GPUs are engaged in "random matrix math" and are producing "zero useful AI computation."
The study also indicates a 38% increase in GPU rental costs.
(Source: Tom's Hardware)
Advertisement
AdSense slot • inline
