Conjure cash with old Macs by linking them to AI inference Borg

Eigen Labs, a Seattle-based tech company, estimates Apple Silicon device owners can earn $120 to $200 monthly by renting out idle compute power for AI inference. Gajesh Naik, an engineer at Eigen Labs, announced the Darkbloom project, a network of Apple Silicon devices, as a paid inference provider on OpenRouter. Currently, 250 Mac machines are online in the network, with users earning $120-200 per month per machine. Suitable devices include Mac mini, M1 MacBook Pro, and other Apple Silicon-equipped macOS devices.

The Darkbloom software is installed on a personal device, bypassing privacy risks associated with AI inference on someone else's hardware. A Swift process handles the inference engine, running on the Apple Silicon GPU via mlx-swift-lm, without subprocesses, local servers, or inter-process communication. The process is hardened, preventing external access to its memory and ensuring confidentiality.

Inference requests are sent over HTTPS to a Go-based coordinator in a Confidential VM, routing the request to an inference provider running the Swift CLI. The provider decrypts the inference job within a Secure Enclave, runs the inference on the Apple Silicon GPU, and re-encrypts the plaintext for the selected provider. The privacy model is akin to Apple's Private Cloud Compute, but applied to a distributed network.

Initially launched as a free service on OpenRouter, Darkbloom now charges for its services. The top-paying model is Qwen 3.6 35B A3B at $399 per month, followed by GPT-OSS 20B at $52, and Gemma 4 26B at $27. Earnings vary, with top earners averaging $26/day, $19/day, and $15/day, respectively. Network participants can expect to earn between $192 and $4,983 annually, factoring in estimated electricity costs.