Startup demos decentralized GPU infrastructure to OpenAI, Uber

What started as an institutional-level quantitative trading system for cryptocurrencies and stocks has evolved into a decentralized network-sourced GPU computing power to serve the growing demand for artificial intelligence (AI) and machine learning (ML) services.

Io.net has created a test network that leverages GPU computing power from a variety of data centers, cryptocurrency miners, and decentralized storage providers. Aggregate GPU computing power is expected to significantly reduce the cost of leasing these resources, which are becoming increasingly expensive as AI and machine learning advances.

Speaking exclusively to Cointelegraph, CEO and founder Ahmed Shadid revealed details of the network, which aims to provide a decentralized platform for renting computing power at a fraction of the cost of current centralized options.

Related: Future payments: Visa to invest $100M in generative AI

Shadid will explain how the project was conceived during the Solana Hackathon in late 2022. Io.net has been developing a digital trading platform that relies on GPU computing power for its high-frequency operations, but has been hampered by the high cost of renting GPU computing power.

The team also opens up the challenge of renting high-performance GPU hardware in its flagship document, with the cost of renting an Nvidia A100 per card at around $80 per day. Needing more than 50 of these cards to operate 25 days a month would cost more than $100,000.

A solution was found when the Ray.io open source library was discovered, which OpenAI used to distribute ChatGPT training to over 300,000 CPUs and GPUs. The library streamlined the project's infrastructure, allowing the backend to be developed over a period of two months.

Shadid demonstrated a working testnet of Io.net at the AI-based Ray Summit in September 2023, showing how the project can integrate computing power, serving as a cluster for GPU users to meet specific AI or machine learning use cases.

“This model not only allows the Io.net GPU to compute up to 90% cheaper than existing vendors, but also enables unlimited computing power.”

The decentralized network of the Solanan blockchain is designed to deliver SOL (SOL) and USD Coin (USDC) payments to machine learning engineers and miners who rent or lease computing power.

“When ML engineers pay for their clusters, these funds go directly to the miners deployed in the cluster with their GPUs, with a small network fee allocated to the IO.NET protocol.”

The project's roadmap includes the launch of a dual native token system comprising IO and IOSD. The token model rewards miners for executing machine learning workloads and maintaining network uptime, taking into account the cost per dollar of electricity consumption.

“The IO coin is freely traded on the crypto market and is a gateway to computing power, while the IOSD token serves as a stable credit token with a stable credit token pegged at $1.”

Shadid told Cointelegraph that Io.net is fundamentally different from centralized cloud services like Amazon Web Services (AWS).

“To use an analogy, they're United Airlines and we're Kayak. They are the planes, but we help people book flights.

The founder added that any businesses that need AI computing usually use third-party vendors because they don't have the GPUs to handle it all in-house. With demand for GPUs estimated to increase 10 times every 18 months, Hadid said there is often insufficient capacity to meet demand, resulting in long wait times and high prices.

This adds to what he describes as inefficiencies in data centers that aren't optimized for the rapidly increasing AI and machine learning workloads.

“There are thousands of independent data centers in the US alone, with an average utilization rate of 12%–18%. As a result, bottlenecks are emerging, which is impacting price increases for GPU computing.

The downside is that the average cryptocurrency miner can compete with the likes of AWS by renting hardware. The average miner using a Hadid 40GB A100 earns $0.52 per day, while AWS is selling the same card for AI computing at $59.78 per day.

“Part of Io.net's value proposition is, first, we allow participants to be exposed to the AI ​​computing market and sell their GPUs, and we're much cheaper than AWS for ML engineers.”

Figures shared with Cointelegraph estimate that miners with GPU resources at their disposal can earn 1,500% more from mining various cryptocurrencies.

Magazine: Blockchain Investigators: M. Gox's Collapse Begat Chinaliss