The economics of running distributed AI inference on consumer hardware are facing severe headwinds. Darkbloom, a network that allows Mac owners to monetize their idle Apple Silicon machines by lending out computing power for AI workloads, has reported a drastic plunge in daily earnings alongside a steep collapse in market token costs.
For everyday users hoping to offset the cost of owning premium Apple hardware through decentralized AI sharing, the financial outlook has darkened considerably. As inference pricing plummets across the broader industry, participating in these peer-to-peer compute networks yields a fraction of what it once did.
Darkbloom Earnings Decline
According to the headline claim from Tom's Hardware, the Darkbloom network has seen earnings drop as Anthropic tokens fall 77 percent. The sharp drop highlights the brutal economic realities facing distributed compute platforms as the market for running artificial intelligence models matures and scales.
The contraction is directly tied to broader pricing trends across major AI providers. As token costs fall globally, they drag down the value of shared GPU compute power.
The Pressures of Commodity AI Inference
The core business model of decentralized inference relies on underutilized hardware—such as the unified memory and capable neural engines found in modern Macs—providing a cheaper alternative to centralized data centers. However, as hyperscalers and specialized providers optimize their infrastructure, the market value of inference tokens has experienced a race to the bottom.
When token costs drop so precipitously, the compensation available for individual hardware contributors shrinks accordingly. What was once viewed as a lucrative way for desktop owners to generate passive income has rapidly transformed into a marginal return that barely covers the electricity required to keep the system powered on during heavy workloads.
Industry-Wide Shifts in Distributed Compute
The phenomenon of leveraging idle macOS hardware for machine learning is not isolated to external networks. Even tech giants like Apple have explored internal distributed GPU systems—such as its "Minions" project, which aggregates idle Apple Silicon across corporate devices to reduce internal queue times for machine learning tasks.
However, while enterprise environments leverage internal pooling to solve localized compute bottlenecks, public decentralized networks must contend with wide-open market forces. As open-source models proliferate and API pricing collapses, platforms like Darkbloom face ongoing challenges in keeping provider payouts attractive enough to maintain a robust hardware pool.
Ultimately, the steep drop in Darkbloom earnings serves as a cautionary tale for the decentralized AI infrastructure space. While consumer hardware holds immense theoretical computing capacity, intense deflation in AI inference pricing threatens the viability of crowd-sourced monetization models.
