What initially began as an institutional-grade cryptocurrency and stock quantitative trading system has evolved into a decentralized network designed to source GPU computing power, addressing the growing demand for artificial intelligence (AI) and machine learning (ML) services. Io.net has developed a test network that aggregates GPU computing power from different sources, including data centers, cryptocurrency miners, and decentralized storage providers. The primary goal is to significantly reduce the cost of renting these resources, which have become increasingly expensive due to advances in AI and ML.
In an exclusive interview , Io.net CEO and co-founder Ahmad Shadid revealed the project's aim to create a decentralized platform for renting computing power at a fraction of the cost compared to existing centralized alternatives. The concept emerged during the Solana Hackathon in late 2022 when Io.net was developing a quantitative trading platform heavily reliant on GPU computing power for high-frequency operations. However, the high cost of renting GPU computing power became a significant hurdle.
The challenges of renting high-performance GPU hardware are outlined in the project's core documentation, with costs averaging around $80 per card per day for renting a single Nvidia A100. For those requiring more than 50 cards per month for 25 days, the expenses could exceed $100,000. The solution was found in Ray.io, an open-source library used by OpenAI for ChatGPT training, which streamlined the project's infrastructure and was developed within just two months.
Io.net demonstrated its working testnet at the Artificial Intelligence Ray Summit in September 2023, illustrating how the project aggregates computing power into a cluster for GPU consumers to address specific AI or machine learning needs. This model allows Io.net to deliver GPU computing that is 90% cheaper than existing vendors while providing virtually unlimited computing power.
The decentralized network leverages Solana's blockchain to deliver SOL and USDC for fees paid to machine learning engineers and miners who rent or provide computing power. The project's roadmap includes the launch of a dual native token system with IO and IOSD capabilities, where IO serves as a gateway to access computing power, and IOSD acts as a stable credit token algorithmically pegged to $1. Io.net differs fundamentally from centralized cloud services like Amazon Web Services (AWS) as it empowers participants to access the AI computing market and resell their GPUs at a significantly lower cost. With rapidly growing demand for GPUs in AI computing, Shadid believes the capacity will be insufficient, leading to long wait times and high prices.
He also points out that data centers are underutilized, with average utilization rates of 12 to 18 percent, creating bottlenecks and driving up GPU computing prices. Regular cryptocurrency miners can potentially profit by renting their hardware to compete with giants like AWS, with miners using the 40GB A100 card earning an average of $0.52 per day, while AWS charges $59.78 per day for the same card in AI computing.
The value proposition of Io.net lies in enabling participation in the AI computing market and offering GPUs at more affordable rates than major providers like AWS, thus potentially allowing miners to earn significantly higher revenues from GPU resources than from various cryptocurrencies,
