Io.net has created a physically decentralized infrastructure network with the aim of sourcing GPU resources for AI and machine learning. 

The project was initially conceived as a quantitative trading system catering to institutional clients in the cryptocurrency and stock markets. However, the development took a different direction due to the exorbitant costs associated with renting GPU computing capacity.

By centralizing the sourcing of GPU computing power from various cloud mining data centers, Io.net aims to reduce the expenses involved, thus making AI and machine learning operations more accessible and affordable.

The project was officially launched during a Solana hackathon in 2022, where it quickly became apparent that the rental costs for high-performance GPU hardware could reach up to $80 per day per card.

To streamline the project's infrastructure, the team utilized the open-source library Ray.io, which proved to be instrumental in simplifying operations. The backend of the project was fully developed within a span of just two months.

By consolidating GPU resources and leveraging a decentralized approach, Io.net is looking to revolutionize the accessibility and affordability of AI and machine learning operations in the broader market.