Renting out your Mac to AI companies now pays far less than hoped

A project that paid people to lend their Macs to AI companies is warning that earnings have dropped and may not recover soon. The project is called Darkbloom, from Seattle-based Eigen Labs. About 800 people take part. Its own researcher says that for most of them, older Macs included, running the software likely will not be worth it.

How it works

Darkbloom's software lets anyone with a capable Mac rent out its graphics chip, the part that does heavy computing. The Mac answers AI requests, a job called inference. Those requests come through OpenRouter, a gateway that connects people to many AI models. Participants earn from a few cents to a few dollars per day.

What people actually earn

Darkbloom estimates that a new Mac mini with an M5 Pro chip and 64 GB of memory could earn up to $192 per month. That assumes the machine works nonstop on a popular model. Among the 50 people on the leaderboard, the median is about $5.26 per day, or $159.86 per month. That is the top group, not the typical participant.

Why pay is falling

The price of tokens, the unit of AI work, has dropped. It is reportedly down 54 percent since May. Ramp's index shows Anthropic's price per million tokens falling from $2.65 on May 1, 2026 to $0.62 on October 22, 2026, a 77 percent drop.

Eigen Labs researcher Kydo told participants that data centers have become cheaper. He said output per NVIDIA cluster is up about 20 times this year. He also said the market has caught up to Darkbloom's prices. Cutting them further would leave people earning even less. OpenRouter is also putting more weight on speed than on price.

What Darkbloom plans next

• Steer participants toward the AI models in highest demand.
• Offer other computing work, such as software testing jobs known as CI/CD.
• Focus on clusters of high-end Mac Studios, now available with up to 512 GB of memory.
• Speed up the network and find buyers beyond OpenRouter.

The testing jobs will need sandboxing, a way to keep outside code walled off from the host machine. The company expects to begin testing that soon.

What to keep in mind

Kydo said that for owners of Macs without Max or Ultra chips, inference will not be worth running. He also expects many machines to leave the network. Those who stay, he said, will be first in line when new earning chances arrive. The company says it is well capitalized and will not give up. Whether new work will pay better is not yet known.