Back

Darkbloom reassesses business as token prices fall
SiTech AI Team3 min read

Darkbloom reassesses business as token prices fall

Eigen Labs' Darkbloom distributed inference network is reassessing plans after falling token prices cut earnings for Mac providers and changed OpenRouter routing.

Darkbloom, a distributed inference network from Seattle-based Eigen Labs, is reassessing its business plans after declining token prices reduced earnings for participants who rent macOS GPU capacity. The network has about 800 participants, and has processed its trillionth token, according to researcher Kydo. Earnings are low and are not expected to recover soon, Kydo said in a Slack message to participants.

Falling prices hit provider income

Darkbloom lets people with sufficiently capable macOS devices rent their GPUs through its network for inference jobs and get paid. Participants handle AI model prompts submitted through OpenRouter. In theory, a Mac mini with an M5 Pro chip and 64 GB of RAM could earn as much as $192 per month at full 24/7 utilization on a sought-after model such as Qwen3.6 35B A3B, which sells for $0.065 per million input tokens and $0.75 per million output tokens. The median earnings for the 50 people on the leaderboard are about $5.26 per day, $159.86 per month, or $1,918.50 per year.

Token prices have declined, reportedly down 54 percent since May. Ramp's AI Token Price Index, which tracks Anthropic and OpenAI token prices, shows similar slippage: Anthropic's price per million tokens fell from $2.65 on May 1, 2026 to $0.62 on October 22, 2026, a decline of 77 percent. Kydo said data centers have become cheaper, output per NVIDIA cluster is up about 20x this year, and margins are wide enough to cut prices 90 percent and still profit at 50 percent. He said the flood of tokens washed away Darkbloom's price advantage and that the market caught up to its price.

Routing changes and older hardware

The inference routing market is also changing, with OpenRouter moving away from price-based routing and putting more emphasis on decode speed, quantization, and prefill speed. For network participants with older or less powerful Mac hardware, those without top of the line Max or Ultra Apple Silicon, Kydo said inference will not be worth running. He also said Darkbloom could go lower on price, but then most providers would earn even less.

Adapting to commoditized tokens

Darkbloom aims to adapt by analyzing model demand to steer participants toward more sought-after models and by looking at ways to provide other kinds of compute work, such as distributed CI/CD jobs. For Mac hardware, the focus now appears to be on clusters of high-end Mac Studios, now available with up to 512 GB of RAM, using Apple's Remote Direct Memory Access over Thunderbolt. The CI/CD work is based on the observation that AI agents now make pull requests faster than CI/CD pipelines can handle them, and it will require sandboxing, something the business expects to begin testing soon. Darkbloom also aims to improve network performance in terms of time to first token and to look for other token buyers beyond OpenRouter.

Team says it will not give up

Kydo concluded that the next stretch will be a fight and that he expects many machines will leave and Slack will get quieter. He said those who stay will be first in line when new earning opportunities come along. "We are a driven team and well capitalized. We will not give up," he said.

Sources: The Register

SiTech

SiTech — AI-powered web development

We build fast, modern websites and bring AI into real business workflows. Have a project or a question? We'd love to help.