Darkbloom – Private inference on idle Macs (darkbloom.dev)

Darkbloom has launched a decentralized inference network that utilizes idle Apple Silicon Macs to provide private AI inference at significantly lower costs. Unlike the traditional multi-layered architecture where costs accumulate as AI compute passes through various providers—from GPU manufacturers to hyperscalers—Darkbloom connects end users directly with the idle computing power of over 100 million Apple machines. This innovation allows operators to earn 95% of the generated revenue, as they utilize their existing hardware, while end users benefit from up to 70% savings compared to centralized services. The significance of Darkbloom lies in its solution to the marketplace inefficiencies in AI compute, akin to how Airbnb and Uber converted idle assets into income-generating resources. Darkbloom ensures user privacy through a rigorously secure architecture: inference data is end-to-end encrypted, operators cannot view user data, and every response is verified to be traceable to specific hardware via a tamper-resistant secure enclave. This three-part verification process enhances trust—crucial for enterprises that may be hesitant to share sensitive information on machines they do not own. Overall, Darkbloom highlights a shift towards more decentralized, cost-effective, and privacy-centric approaches to AI inference.