jakiro
Source: Trustless Labs
background
As OpenAI launches GPT4 LLM, the potential of various AI Text-To-IMAGE models has been witnessed. The increasing application of mature AI models is increasing, and the demand for computing power resources such as GPU has risen.
An article in GPU Utils discussed the supply and demand of NVIDIA H100 GPUs, stating that large enterprises involved in AI business have strong demand for GPUs. Technology giants such as META, TESLA and Google have purchased a large number of NVIDIA GPUs to build facing AI for AI.Data center.Meta has about 21,000 A100 GPUs, TESLA has about 7,000 A100, and Google’s data center also has a large amount of GPU investment, although no specific number is provided.The needs of training large -scale language models (LLM) and other AI applications have continued to grow for GPU (especially H100).
At the same time, according to Statista’s data, the AI market has increased from 134.8 Billion in 2022 to 241.8 Billion in 2023, and is expected to reach 738.7 Billion in 2030, and the market value of cloud services will increase by approximately 14%from 633 Billion.Some parts of the AI market have increased rapidly for the GPU computing power.
For the fast -growing and huge potential AI market, what perspective can we deconstruct and dig out the entry point related to it?According to an IBM report, we summarize the infrastructure required for creating and deploying artificial intelligence applications and solutions.It can be said that the AI infrastructure is mainly to deal with and optimize the large amount of data sets and computing power resources relied on the training model. From the aspects of hardware and software, the data set processing efficiency, model reliability and application scalability are solvedEssence
>
A large amount of computing power resources needed in the AI training model and applications, the low -delayed cloud environment and the computing power of the GPU, also include a distributed computing platform (Apache Spark/Hadoop) in terms of software stack.Spark has a built -in parallel mechanism and fault -tolerant design by dispersing the workflows that need to be processed into various large computing clusters.The natural decentralized design idea of the blockchain makes distributed nodes the norm, and the POW consensus mechanism created by BTC has established the result of the miners who need to use computing power (workload) competition. This is the same as AI.There are similar workflows to generate model/reasoning problems.So traditional cloud server manufacturers began to expand new business models to rent graphics cards like rental servers, and selling computing power.In the idea of imitating the blockchain, the AI computing power uses a distributed system design, which can use idle GPU resources to reduce the computing power cost of startup companies.
>
IO.NET project profile
IO.NET is a distributed computing power provider that combines Solana blockchain to use distributed computing power resources (GPU & AMP; CPU) to solve the challenges of computing needs in the field of AI and machine learning.IO gathered more than 1 million GPU resources to solve the problem of AI computing resources by integrating idle graphics cards from independent data centers and cryptocurrency miners, and combined with Filecoin/Render and other Crypto projects.
At the technical level, IO.NET is built on Ray.io, a machine learning framework of distributed computing, and provides distributed computing resources for AI applications from enhanced learning, deep learning to model tuning, and model operation.Anyone can add IO’s computing power network with the role of Worker or DEVELOPER without additional permission. At the same time, the network will adjust the computing power price according to the complexity, emergency and computing power supply of computing power resources according to the market dynamic pricing.Based on the characteristics of computing power distribution, the back end of IO will also match the GPU provider and developers according to the GPU demand type, current available amount, and the position and reputation of the request.
$ IO is the native currency of the IO.NET system. It acts as a trading medium between computing providers and computing power service buyers. Compared with $ IO and $ USDC, the order handling fee can be reduced by%2.At the same time, $ IO also plays an important incentive role to ensure the normal operation of the network: the $ IO token holder can pledge a certain number of $ IO to the node, and the node operation also needs to be obtained with the $ IO token pledge before obtainingCorresponding income during the free period of machine.
The current market value of $ IO tokens is about 360 million US dollars, and FDV is about $ 3 billion.
$ IO token economics
The maximum total supply of $ IO is 800 Million, of which the 500 Million is allocated by all parties when the token TGE.About 12%of the year).The current IO circulation is 95 Million, where the mining reward composition of 75 Million and the Launchpool 20 Million unlocked by ecological R & D and community construction in TEGE.
The reward distribution of computing power providers during the IO test network is as follows:
-
Season 1 (as of April 25) -17,500,000 IO
-
Season 2 (May 1 -May 31) -7,500,000 IO
-
Season 3 (June 1 -June 30) -5,000,000 IO
In addition to testing the computing power reward, IO also gave some airdrops participating in the construction of the community:
-
(First round) Community / content creator / Galxe / Discord -7,500,000 IO
-
Season 3 (June 1 -June 30) Discord and Galxe participants -2,500,000 IO
Among them, the first quarter test network computing reward and the first round of community creation/Galxe awards have been completed in the airdrop.
According to the official document, the overall allocation of $ IO is as follows:
>
>
>
$ IO token destruction mechanism
IO.NET performs the repurchase and destruction of the $ IO tokens based on a fixed set of preset programs. The specific repurchase and destruction quantity depends on the $ IO price during execution.The funds used to repurchase $ IO are from the operating income of IOG (The Internet of GPUS -GPU Internet). The computing power buyers and computing power providers in IOG each charge 0.25%of the order schedules, and the $ USDC2%of the fee for computing power.
>
>
Competing product analysis
Projects similar to IO.NET include Akash, NOSANA, Octaspace, CLORE.AI, etc. focusing on the decentralized computing power market focusing on the calculation of AI models.
-
Akash Network uses a decentralized market model to use idle distributed computing resources to collect and rent excess computing capabilities, and respond to imbalances of supply and demand through dynamic discounts and incentive mechanisms.So as to provide safe, economical and efficient and decentralized cloud computing services.It allows Ethereum miners and other users with unused GPU resources to lease these resources out to create a cloud service market.In this market, the pricing of services is carried out by the reverse auction mechanism. Buyers can rent these resources to promote the decrease in price competitiveness.
-
NoSana is a decentralized computing market project in the Solana ecosystem. The main intention is to form a GPU grid with the help of idle computing power resources to meet the calculation needs of the AI reasoning link.This project is defined by the program on Solana to define its computing power market operation and ensure that the GPU node participating in the network completes the task reasonably.In addition to the test network operation in the second stage, it provides computing power services for LLAMA 2 and Stable Diffusion model reasoning processes.
-
Octaspace is an open source and scalable distributed distributed computing cloud node infrastructure, allowing access to distributed computing, data storage, service, VPN, etc.OctASPACE includes CPU and GPU computing power, service space for ML tasks, AI tools, image processing, and using Blender to rendering scenarios.Octaspace was launched in 2022 and runs on its own Layer 1 EVM compatible blockchain.The blockchain adopts a dual -chain system, combining the workload certificate (POW) and authoritative proof (POA) consensus mechanism.
-
CLORE.AI is a distributed GPU supercomputing platform that allows users to obtain high -end GPU computing power resources from nodes that provide computing power worldwide.It supports AI training, cryptocurrency mining and film rendering.The platform provides low -cost, high -performance GPU services, and users can get the CLORE token reward through the lease GPU.CLORE.AI focuses on security, abide by European law, and provide strong API to achieve seamless integration.In terms of project quality, CLORE.AI’s webpage is relatively rough, without detailed technical document verification of the authenticity and data authenticity of the self -introduction of the project. We have maintained doubts about the graphics card resources and real participation of the project.
Compared with other products in the decentralized computing power market, IO.NET is currently the only one who can join the project without admission to provide computing power resources. Users can use the minimum 30 series consumer GPU to participate in the networkThe contribution of computing power also has Apple chip resources such as MacBook M2, Mac mini.More sufficient GPU and CPU resources and rich API construction enable IO to support various AI computing needs, such as batch reasoning, parallel training, super -digital adjustment, and strengthening learning.The terminal infrastructure is composed of a series of modular layers, which can achieve effective management and automated pricing of resources.Other distributed computing power market projects are mostly cooperated with enterprise graphics resources, and there are certain thresholds for users to participate in.Therefore, IO may have the ability to use the encrypted flywheel to use token economics to pry more graphics card resources.
The following is the comparison of the current market value/FDV of IO.NET and competing products:
>
Review and Conclusion
The $ IO online Binance can be described as the start of the start. The hot network of the test network and the heavy project that gradually attacked and questioned the opaque of integration rules during the actual expansion of the test network and drew a match ending.The tokens were launched during the market recovery period, opened low, and eventually returned to the valuation range of relatively rational rationality.However, for the test network participants who came from IO.NET’s strong investment lineup, several people who were happy and happy, most of them leased GPUs but did not insist on participating in each quarter test network.Instead, he faces the reality of “anti -现”.During the test network, IO.NET divided each issue of the prize pools into two POOLs of GPU and high -performance CPU to calculate separately. Season 1 was delayed due to the announcement of hackers’ points.90: 1. The cost of users participating in GPUs in major cloud platform manufacturers far exceeds the income of airdrops.During the Season 2 period, the official POW verification mechanism was officially fulfilled. Nearly 3W GPU devices successfully participated and passed POW verification. The final score exchange ratio was 100: 1.
After the highly anticipated start, whether IO.NT can achieve the goals of providing various links for AI applications, how much real requirements remain after the test network, maybe only time can give the best proof.
