作者：MIIX Capital 来源：medium

MIIX Capital发布io.net项目研究报告

1、项目情况

1.1 业务概要

io.net是去中心化的GPU网络，旨在为机器学习提供计算。通过组装来自独立数据中心、加密货币矿工和Filecoin或Render等项目的100万多个GPU来获取计算能力。

它的目标是将100万个GPU组合到DePIN（去中心化物理基础设施网络）中，打造一个企业级、去中心化的分布式计算网络，通过汇聚全球闲散的网络计算资源（目前主要是GPU），为人工智能工程师们提供价格更低、更易获得、更灵活适配的网络计算资源服务。

对于用户来说，它就相当于一个去中心全球闲散GPU资源的集市，让人工智能工程师或团队可以在这里按照他们的需求定制化搭配和购买所需的GPU计算服务。

1.2 团队背景

创始人兼首席执行官Ahmad Shadid是WhalesTrader量化系统工程师。

首席战略官兼首席营销官Garrison Yang是Ava Labs增长与战略副总裁。

首席运营官Tory Green是Hum Capital首席运营官、Fox Mobile Group企业发展与战略总监。

商务拓展副总裁Angela Yi毕业于美国哈佛大学，负责规划并执行销售、伙伴关系和供应商管理等关键战略。

2020年，Ahmad Shadid为机器学习量化交易公司Dark Tick构建GPU计算网络时，因为交易策略接近于高频交易，所以需要大量的算力，云服务厂商高昂的GPU服务费用成为了他们的难题。

对算力的巨大需求以及所面临的高昂成本促使他们决定去做去中心化分布式计算资源这件事，随后又在Austin Solana Hacker House获得关注度。因此，io.net属于该团队从自身面临的痛点出发，提出解决方案并进行业务落地和拓展。

1.3 产品/技术

市场用户面临的问题：

可用性有限，使用AWS、GCP或Azure等云服务访问硬件通常需要数周时间，而且市场上流行的GPU模型通常不可用。

选择余地很少，如在GPU硬件、位置、安全级别、延迟等方面用户几乎没有选择余地。

成本较高：获得优质GPU非常昂贵，每月很花费数十万美元用于训练和推理。

解决方案：

通过聚合未充分利用（例如独立数据中心、加密矿工以及Filecoin、Render等加密项目）的GPU，把这些资源整合到DePIN中，使工程师能够在系统中获得大量计算能力。它允许ML团队跨分布式GPU网络构建推理和模型服务工作流

The author released the project research report from the source. The business summary of the project is a decentralized network, which aims to provide computing for machine learning. By assembling more than 10,000 cryptocurrency miners and or other projects from independent data centers, the computing power is obtained. Its goal is to combine 10,000 into a decentralized physical infrastructure network to build an enterprise-level decentralized distributed computing network. At present, it mainly provides artificial intelligence engineers with lower prices by gathering idle network computing resources around the world. It is easier to obtain more flexible and adaptive network computing resource services. For users, it is equivalent to a market that goes to the center of global idle resources, where artificial intelligence engineers or teams can customize and purchase the required computing service team background according to their needs. The founder and CEO of the project research report are quantitative system engineers, chief strategy officers and chief marketing officers, vice president of growth and strategy, chief operating officer and director of enterprise development and strategy, Business Extension. Vice President Zhan graduated from Harvard University in the United States and is responsible for planning and implementing key strategies such as sales partnership and supplier management. When building a computing network for machine learning quantitative trading companies in, because the trading strategy is close to high-frequency trading, it needs a lot of computing power. The high service cost has become their problem. The huge demand for computing power and the high cost they face prompted them to decide to decentralize distributed computing resources, and then it gained attention, so it belongs to. The team puts forward solutions from the pain points they face, and carries out business landing and product technology project research report. The problems faced by market users are limited in availability. It usually takes several weeks to access hardware by using or waiting for cloud services, and the popular models in the market are usually unavailable. For example, users have little choice in terms of hardware location, security level delay, etc. The cost is high and it is very expensive to obtain high quality. Hundreds of thousands of dollars are spent on training and reasoning solutions every month. Integrating these resources into the system by aggregating encryption projects such as independent data center encryption miners and so on, enables engineers to gain a lot of computing power in the system. It allows teams to build reasoning and model service workflows across distributed networks. 比特币今日价格行情网_okx交易所app_永续合约_比特币怎么买卖交易_虚拟币交易所平台

注册有任何问题请添加 微信：MVIP619 拉你进入群

文章转载注明 网址：https://netpsp.com/?id=63726