Authored by: Shayon Sengupta, MultICOin Capital
Translated by: JIN, Techub News
On June 6, 2024, Binance announced that Launchpool would be launching io.net. Users could deposit BNB and FDUSD into the IO mining pool on the Launchpool website starting from 8:00 AM Hong Kong time on June 7th to earn IO rewards. The mining of IO tokens would be available for a total of 4 days. The website was expected to be updated approximately five hours before the mining activity opens.
Furthermore, Binance would list the IO token on June 11th at 8:00 PM Hong Kong time on io.net. Trading pairs such as IO/BTC, IO/USDT, IO/BNB, IO/FDUSD, and IO/TRY would be available on the exchange.
IO Token Unlock and Rewards
According to the official documentation of io.net, the total supply of IO tokens is 8 billion. Initially, 5 billion IO tokens would be released at launch, with an additional 3 billion IO tokens gradually issued over the next 20 years until reaching the limit of 8 billion tokens. The initial 5 billion supply would be unlocked and allocated as shown in the infographic below, divided into five categories: seed investors, Series A investors, core contributors, research and ecosystem, and community.
IO Token Estimated Allocation
Seed Investors: 12.5%
Series A Investors: 10.2%
Core Contributors: 11.3%
Research: 16%
Ecosystem and Community: 50%
The following information about io.net was introduced by Multicoin Capital, one of the investors in the $30 million Series A funding round of io.net:
“We are pleased to announce our investment in io.net, a distributed network that provides AI computing power rental services. We not only led the seed round but also participated in the Series A funding. io.net raised a total of $30 million, with investors including Multicoin, Hack VC, 6th Man Ventures, Modular Capital, and a consortium of angel investors, aiming to build an on-demand, readily available AI computing power marketplace.”
I first met the founder of io.net, Ahmad Shadid, at the Solana hackathon event in Austin Hacker House in April 2023, and was immediately drawn to his unique insights into decentralized AI computing infrastructure.
Since then, the io.net team has demonstrated strong execution capabilities. Today, the network has aggregated tens of thousands of distributed GPUs and provided over 57,000 hours of computing time to AI enterprises. We are excited to collaborate with them and support the AI renaissance in the coming decade.
1. Global Computing Power Shortage
The demand for AI computing is growing at an astonishing rate, exceeding the current supply. In 2023, data centers providing computing power for AI generated revenue exceeding $100 billion, yet even in the most conservative estimates, the demand for AI surpasses chip supplies.
During periods of high interest rates and cash flow constraints, new data centers capable of accommodating such hardware require significant initial investments. The core issue lies in the limitations on the production of advanced chips like NVidia A100 and H100. While GPU performance continues to improve and costs steadily decrease, the manufacturing process cannot be accelerated due to shortages of raw materials, components, and production capacity.
Despite the vast potential of AI, the physical footprint required to support its operations is increasing daily, leading to a significant rise in the demand for space, power, and cutting-edge equipment. io.net has opened a path for us where computing power is no longer constrained by these limitations.
io.net is a classic example of applying DePIN in real-world scenarios: structurally reducing the cost of acquiring supply-side resources through token incentives, lowering costs for end users of GPU computing power. By aggregating idle GPU resources from around the world into a shared pool for AI developers and companies to use, the network is supported by thousands of GPUs from data centers, mining farms, and consumer-grade devices.
While valuable resources can be integrated, they do not automatically scale to a distributed network. In the history of cryptocurrency technology, there have been several attempts to build distributed GPU computing networks, all of which failed because they did not meet the demands of the end-users.
Coordinating and scheduling computing tasks on heterogeneous hardware with different memory, bandwidth, and storage configurations is a key step in achieving a distributed GPU network. We believe that the io.net team has the most practical solution in the market today, enabling the aggregation of hardware to be useful to end customers and cost-effective.
2. Paving the Way for Clusters
In the history of computer development, software frameworks and design patterns adjust themselves around the available hardware configurations in the market. Most frameworks and libraries used for AI development heavily rely on centralized hardware resources, but in the past decade, distributed computing infrastructure has made significant progress in practical applications.
io.net leverages existing idle hardware resources to interconnect them by deploying custom networks and orchestration layers, creating an ultra-scalable GPU internet. This network utilizes Ray, Ludwig, Kubernetes, and various other open-source distributed computing frameworks, enabling machine learning engineering and operations teams to scale their workloads on existing GPU networks.
ML teams can parallelize workloads on io.net GPUs by launching clusters of computing devices and utilizing these libraries for orchestration, scheduling, fault tolerance, and scaling. For example, if a group of graphic designers contribute their GPUs at home to the network, io.net can build a cluster designed to allow image model developers worldwide to rent collective computing resources.
BC8.ai is an example, a fine-tuned stable diffusion variant model trained entirely on the io.net network.
The io.net browser displays real-time inferences and incentives for network contributors.
Each generated image information is recorded on-chain. All fees are paid to a 6 RTX 4090 cluster, which consists of consumer-grade GPUs used for gaming.
Today, there are thousands of devices on the network, spread across mining farms, underutilized data centers, and consumer nodes of Render Network. Besides creating new GPU supply, io.net can also compete in costs with traditional cloud service providers, often offering cheaper resources.
They achieve cost reduction by outsourcing GPU coordination and operations to decentralized protocols. On the other hand, cloud service providers mark up their products due to employee expenses, hardware maintenance, and data center operational costs. The cost of consumer-grade GPU clusters and mining farms is much lower than the costs that hyperscalers are willing to bear for large-scale computing centers, providing a structural arbitrage that dynamically prices resources on io.net below the continuously rising cloud service rates.
3. Building the GPU Internet
io.net has a unique advantage of maintaining lightweight asset operations, reducing marginal costs to almost zero in servicing any specific customer, and establishing direct relationships with market demands and supply. It can cater to thousands of individuals needing access to GPUs to build competitive AI products, paving the way for everyone to interact with it in the future.
