Eigen Labs, the company behind Ethereum's EigenLayer restaking protocol, launched Darkbloom this week. It's a decentralized network designed for running LLMs on idle Apple Silicon Macs. Over 100 million Macs sit unused most of the day, and their unified memory architecture happens to be well-suited for running large language models. Users get an OpenAI-compatible API at roughly half the cost of centralized providers. Mac owners get paid for hardware they already own.
Sending proprietary data to a stranger's MacBook sounds like a non-starter for any serious company. Darkbloom tackles this with hardware-bound encryption tied to Apple's secure enclave, end-to-end encryption, and a hardened runtime that blocks debugger attachment and memory inspection. The attestation chain traces back to Apple's root certificate authority. Operators cannot see your prompts or responses, according to Eigen Labs. That's the claim, anyway.
There are real friction points. The service requires installing what amounts to MDM software on your machine, which will make security-conscious users nervous regardless of the technical guarantees. Early reports from users suggest demand is thin and there are still technical issues to work through. The model catalog is small: Gemma 4 26B, Qwen 3.5 variants, MiniMax M2.5, and Cohere Transcribe for speech-to-text. These are capable models, but they're not what most developers are reaching for first.
The economics could work if the network achieves critical mass. Eigen Labs says operators keep 100% of inference revenue, with electricity costs running $0.01 to $0.03 per hour on Apple Silicon. The rest is profit. But Darkbloom faces a familiar cold-start problem: the network needs enough supply to attract demand, and enough demand to attract supply. Sreeram Kannan, who founded Eigen Labs and leads the project, knows this dynamic well from building EigenLayer. Whether the crypto playbook translates to physical compute infrastructure remains an open question.