Conjure cash with old Macs by linking them to AI inference Borg
Eigen Labs, a Seattle-based startup, is inviting owners of Apple Silicon Macs to rent out their idle computing power to earn passive income through Darkbloom, a distributed AI inference network that recently became a paid provider on OpenRouter. The move matters because it offers a novel way for ordinary Mac owners to monetise underused hardware while helping power open-weight AI models, positioning itself as an alternative to costly frontier-lab data centres.
Darkbloom, which began as an experiment in April, now has more than 900 registered providers and has processed over 42 million inference requests, generating around $102,000 in annual recurring revenue from 4.5 billion tokens served. Eigen Labs estimates participants can earn between $120 and $200 a month on average, with annual earnings potentially ranging from $192 to $4,983, minus roughly $2 monthly in electricity costs. The system routes requests through a Go-based coordinator running in a confidential virtual machine, with inference performed locally on providers' Macs using Apple's MLX framework and hardened security measures designed to prevent access to prompt data, similar in principle to Apple's Private Cloud Compute.
- Darkbloom lets Mac owners earn money renting idle compute for AI inference.
- Network has 900+ providers, served 42 million requests, $102K in revenue.
- Average earnings $120-200 monthly per machine, using hardened privacy safeguards.
New here? Start with this
AI inference is the computing work involved in answering a user’s request with a trained AI model. It can run in large data centres, but some open-weight models—whose model files are available for others to use—can also run on individual computers.
Darkbloom is a network set up by Seattle startup Eigen Labs to share that work across participating Apple Silicon Macs. A coordinator sends jobs to the computers, which process them locally; the network is designed to keep prompt data from being exposed to the Mac owners.
The idea connects people with spare computing capacity to services that need it, including OpenRouter, which routes requests to AI models. It could give Mac owners a way to earn money from idle hardware and provide another source of computing for open-weight AI, though actual earnings depend on how much work the network sends each computer.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Supporters argue this represents a genuinely valuable opportunity for Mac owners to monetise idle hardware whilst contributing meaningfully to AI infrastructure. The reported earnings of £100-160 monthly constitute meaningful supplementary income for many households, and the distributed model reduces reliance on concentrated data-centre operators and their associated environmental costs. The security architecture, mirroring Apple's Private Cloud Compute principles, demonstrates serious consideration of privacy, and 900+ active providers suggest participants find the proposition worthwhile.
The case against
Sceptics raise legitimate concerns about whether the financial case truly holds for participants once electricity costs, potential hardware wear, and tax implications are properly accounted for. They note the inherent security risks of running third-party network software on personal machines, the untested sustainability of a single startup's network beyond its novelty phase, and the risk that terms of service or liability arrangements could shift unfavourably for participants. The company's profitability does not guarantee returns for contributors, and participants bear the technical and legal risks whilst Eigen Labs retains control.