Darkbloom: Harnessing Apple Silicon Macs for Decentralized AI Inference and Earnings
Clip title: This feels illegal… Author / channel: Matthew Berman URL: https://www.youtube.com/watch?v=z1ez0yWu1P4
Summary
The video introduces Darkbloom, a novel project aiming to revolutionize AI inference by creating a distributed, peer-to-peer network utilizing the idle computing power of individual users’ Apple Silicon Macs. The core concept is to decentralize the traditional data center model, allowing applications to send encrypted requests to a network of verified user Macs, which then process the requests privately and cost-effectively, returning a private result. This “distributed data center” model offers a compelling alternative to massive, centralized data centers that often draw criticism for their scale and energy consumption.
For individual users, Darkbloom presents an opportunity to earn passive income by sharing their Mac’s unused compute resources. The video demonstrates an “earnings estimate” calculator, showing that a high-end Mac Studio could potentially earn around $37 per month, framing this as “basically free money” given the minimal incremental electricity cost due to Apple Silicon’s efficiency. This distributed approach also significantly reduces the cost of AI inference for developers, being reportedly 50% cheaper than other providers on platforms like OpenRouter, where Darkbloom is currently operational and has already served billions of tokens.
Technically, Darkbloom prioritizes privacy and security. It runs inference engines directly within a single, hardened Swift process on the Apple Silicon GPU, with the macOS kernel blocking all external access to the process’s memory. This design prevents the Mac owner from observing the user’s prompts or the model’s responses, addressing critical data privacy concerns. The project, backed by Eigen Labs, provides a technical white paper for transparency. While an AI audit found no hidden malware or credential theft, it did note that the “100% revenue” claim, while mostly true (currently a 3% platform fee, but with admin override potential), could technically be altered in the future, and aspects like money routing are highly centralized.
The presenter highlights the project’s early-stage but rapidly growing nature, with over 250 Macs already online. While the current minimum RAM requirement for providers is 48GB (due to high demand), this may be adjusted as the system matures. The overarching takeaway is the profound potential of distributed compute in AI. It democratizes access to powerful AI models, reduces reliance on a few concentrated entities, and empowers individuals to contribute to and benefit from the AI landscape, fundamentally shifting the paradigm of AI infrastructure towards a more decentralized and transparent future.
Video Description & Links
Description
My Links 🔗
Links: https://www.darkbloom.dev/
Tags
ai, llm, artificial intelligence, large language model, openai, mistral, chatgpt, ai news, claude, anthropic, apple ai, apple intelligence, llama, meta ai, google ai
URLs
Related Concepts
- decentralized AI inference
- peer-to-peer network — Wikipedia
- Apple Silicon — Wikipedia
- distributed computing — Wikipedia
- encrypted requests
- idle computing power
- data center model
- Darkbloom
- passive income — Wikipedia
- data privacy — Wikipedia
- OpenRouter — Wikipedia