Darkbloom Warns Participants of Declining Payouts
The network faces a shift in strategy as token prices for AI inference jobs fall significantly.
Updated on Oct. 30, 2026 in Data Centers
Live Poll
Is now a good time to rely on crowd-sourced computing networks for extra income?
Darkbloom, a software network allowing Mac users to rent GPU power for AI inference, has warned participants of decreasing daily earnings. The decline follows a 77 percent drop in token prices for Anthropic models since May.
Why it matters
The downturn highlights the competitive pressure from professional data centers, which currently offer more cost-effective computational power. Eigen Labs now plans to pivot the network to focus on distributed CI/CD jobs.
The Darkbloom network currently supports 800 participants processing AI prompts through the OpenRouter gateway. While output per NVIDIA cluster has increased 20 times this year, token prices have fallen from $2.65 to $0.62 since May 1, 2026.
The players
Eigen Labs
The Seattle-based organization behind the Darkbloom network is transitioning its focus toward distributed CI/CD jobs.
OpenRouter
This gateway service processes AI model prompts and is currently shifting its focus from price-based routing to speed-based decoding.
The details
The software enables owners of macOS devices to monetize their hardware by contributing to inference tasks. However, with professional data centers becoming more efficient, the platform is moving toward new compute work to sustain its infrastructure.
Timeline
May 1, 2026: Anthropic token prices reached $2.65 per million units.
October 22, 2026: Anthropic token prices fell to $0.62 per million units.
October 30, 2026: The current status of the network was reported.
The Tech Race
This transition reflects the broader evolution toward specialized, high-speed computational routing. By shifting away from consumer-led GPU aggregation, the project moves to mirror the efficiency of industrial-grade data centers.
Users currently participating in the network may see lower daily earnings as token prices remain suppressed. Those looking to contribute GPU power should anticipate a pivot toward different types of computing workloads beyond AI inference.
The takeaway
The move toward CI/CD jobs suggests that distributed networks must diversify their workload to compete with specialized data centers. Participants should monitor project updates regarding software requirements for the new compute tasks.
What happens next
Eigen Labs expects to begin testing sandboxing for CI/CD jobs in the near future.
Further reading
Learn more about the infrastructure behind regional computing on the Data Centers page.
Live Poll
Is now a good time to rely on crowd-sourced computing networks for extra income?
