Darkbloom Warns Mac Users of Declining GPU Rental Payouts
Distributed inference network faces market headwinds as AI token prices plummet significantly.
1Network Revenue Model Stalls
Darkbloom, a Seattle-based project by Eigen Labs, allows Mac users to rent out GPU capacity for AI inference tasks. While participants previously earned income by processing prompts via OpenRouter, declining token prices have severely impacted profitability. The platform now warns users that current earnings are low and unlikely to recover soon.
2Earnings Reality Versus Estimates
Theoretical earnings for high-end M5 Pro Mac mini users were estimated at $192 monthly under full 24/7 utilization. However, the median earnings for top leaderboard participants currently sit at approximately $5.26 per day. This disparity highlights the challenge of maintaining consistent income as market conditions for AI compute work shift rapidly.
3Market-Wide Token Price Collapse
The AI token market has experienced a significant downturn, with prices dropping 54 percent since May. Data from the Ramp AI Token Price Index confirms this trend, showing Anthropic token prices falling 77 percent between May and October 2026. This commoditization has eroded the competitive price advantage Darkbloom previously held for users.
4Structural Shifts in Compute Costs
Researcher Kydo noted that data centers have become significantly cheaper, with NVIDIA cluster output increasing 20x this year. These facilities operate with margins wide enough to slash prices by 90 percent while remaining profitable. Consequently, the market has caught up to Darkbloom’s pricing, making it difficult for individual providers to compete.
5Changing Routing and Hardware Needs
OpenRouter is shifting its focus from price-based routing toward metrics like decode speed, quantization, and prefill speed. This transition renders older or less powerful Mac hardware obsolete for inference tasks. Kydo explicitly warned that for users without high-end Apple Silicon, running inference jobs will no longer be a financially viable activity.
6Strategic Pivot to New Workloads
To adapt, Darkbloom is exploring distributed CI/CD jobs, noting that AI agents generate pull requests faster than current pipelines handle them. The team plans to implement sandboxing for these tasks soon. They are also analyzing model demand to steer participants toward more sought-after models to maximize potential earnings for remaining users.
7Focus on High-End Hardware
The project is shifting its hardware focus toward clusters of high-end Mac Studios, which support up to 512 GB of RAM. By utilizing Apple's Remote Direct Memory Access over Thunderbolt, the team aims to improve network performance. This shift prioritizes technical efficiency to remain relevant in a highly competitive, commoditized compute market.
8Future Outlook for Participants
Darkbloom management acknowledges the upcoming period will be difficult, expecting many machines to leave the network. Despite the bleak short-term outlook, the team remains well-capitalized and committed to finding new token buyers. They emphasize that those who stay will be positioned to capture future earning opportunities as the platform evolves.