Idle Mac Rental Network Warns Earnings Will Fall to $1.23 a Day

Darkbloom, a decentralized network that rents out idle Macs for AI inference, has told its members to expect revenue to fall sharply as token prices keep dropping. The price per million tokens for Anthropic fell 77 percent in under five months, from $2.65 in May to $0.62 in October 2026, according to The Register. The network estimates that a top operator's monthly earnings could shrink to roughly $36.77 — about $1.23 a day.

Why are payouts falling?

The core reason is price competition from conventional data centers. Kydo, a researcher at Eigen Labs, the developer of the compute platform, explained the mechanism plainly.

"Data centers have become cheaper. Output per NVIDIA cluster is up ~20x this year, and their margins are wide enough to cut prices 90 percent and still profit at 50 percent. The market caught up to our price. We could go lower, but then most of you would earn even less [than] nothing."

In other words, hyperscale GPU capacity has become so productive and so cheap to run that centralized providers can undercut distributed Mac fleets while keeping healthy margins. Darkbloom's pricing power has evaporated from the supply side, not from falling demand.

The numbers show how steep the slide is. A best-case estimate puts an M5 Pro Mac Mini with 64GB of Unified Memory at as much as $192 per month. The median among the top 50 operators on Darkbloom's leaderboard is lower: about $159.86 per month, or roughly $5.26 a day and $1,918.50 a year. Applying Anthropic's 77 percent token price drop, that median figure collapses to about $36.77 a month, or $1.23 per day.

The takeaway: the same hardware earning almost $160 a month could soon earn barely a quarter of that, purely because of market pricing.

Are token prices falling faster than chips improve?

Yes, and by a wide margin. Some reports cited alongside the Darkbloom warning say AI inference is becoming 50 percent cheaper every quarter and up to 13 times more affordable each year — a pace that outruns Moore's Law by several multiples.

That collapse does not mean AI budgets are shrinking. Companies have been reported to burn through annual AI spending in months, but the cause is volume, not unit price. Agentic AI workflows generate enormous numbers of tokens, and that growth in consumption has outpaced the per-token price drops.

Cheap tokens therefore expand total spending even as each individual token earns less for whoever supplies the compute.

What changed at OpenRouter?

Darkbloom's primary client, OpenRouter, has also shifted what it values. The routing platform now prioritizes decode speed, quantization, and prefill speed over raw cost. That change favors machines with top-of-the-line Max or Ultra Apple silicon and puts operators with lesser chips at a structural disadvantage.

The commercial effect is twofold. Falling token prices cut revenue per unit of work, while the new speed requirements shrink the pool of Macs that qualify for the best-paying jobs in the first place.

How does Darkbloom plan to respond?

The network says it will not sit idle while members lose income. Its response, as outlined to members, has four strands:

• Analyzing model demand to advise members on which models are currently in highest demand
• Exploring workloads beyond inference, including distributed CI/CD jobs
• Improving network performance to compete better against other compute providers and aggregators
• Finding new clients beyond OpenRouter

Each move addresses a different weakness. Demand analysis helps members pick profitable models today; new workload types and clients reduce dependence on a single buyer; performance work targets the speed criteria OpenRouter now rewards.

The takeaway in plain words: Darkbloom is trying to diversify before its single largest customer and a collapsing price curve squeeze its suppliers out of the market. Whether distributed Apple-silicon compute can hold a niche against 20x-cheaper GPU clusters will decide if $1.23 a day is a floor or a waypoint.