Decentralized computing networks are the foundation of decentralized artificial intelligence (AI). They provide the distributed computing power needed to train and run AI models. This article explores three of the largest general-purpose decentralized computing networks and one decentralized AI project. We aim to help readers understand the similarities and differences between these projects.

1. Akash, Render Network and io.net are three of the largest decentralized computing networks on the market. Although they both provide decentralized computing services, each network has a different business focus.
3. Bittensor is a decentralized artificial intelligence project that uses distributed computing resources for machine learning. It aims to compete directly with centralized AI services like OpenAI.
5. On the supply side, Akash has a diverse hardware network including CPUs, GPUs, and storage, while Render has a large number of GPUs. io.net obtains a large number of GPUs from its network and other platforms.
7. A decentralized computing network is a two-sided market where each project’s token is used as a medium of exchange in their respective systems. The Render Network and Bittensor implement a token-burning mechanism to enhance value accumulation.

Different types of decentralized computing networks

How Akash differs from Render Network

Akash and Render Network are both decentralized computing networks that provide a platform where users can buy and sell computing resources for various tasks.

Akash operates as an open marketplace, allowing users to access CPU, GPU, and storage resources. It provides computing resources that can be used for various purposes, such as hosting game servers or running blockchain nodes. In the Akash marketplace, tenants deploying applications set the price and conditions for required deployments, while computing resource providers bid on those deployments, with the lowest bidder (the provider) winning the deployment. This reverse auction model gives users the power to set prices and conditions.

In contrast, Render uses a dynamic pricing algorithm to adjust task deployment pricing based on market conditions. Render Network focuses on GPU-based 3D rendering services and operates as a distributed GPU network. In this model, hardware providers provide computing resources and the Render network uses a multi-tiered pricing algorithm to determine prices and match users with buyers of services. Render does not operate as an open market where users can independently set prices or conditions.

Io.net - Focus on Artificial Intelligence and Machine Learning

io.net is a new decentralized computing network that sources GPU computing power from geographically distributed data centers, cryptocurrency miners, and decentralized storage providers to support machine learning and artificial intelligence computing. It also works with existing decentralized computing networks such as Render to leverage underutilized GPU computing resources on Render for AI and machine learning tasks.

There are two main differentiating factors for io.net: 1) focus on AI and machine learning tasks; 2) emphasis on GPU clusters. A GPU cluster refers to multiple GPUs working together as a unified system to handle computationally intensive tasks such as AI training and scientific simulations.

Bittensor - an AI-focused blockchain project

Unlike other decentralized computing networks, Bittensor is a decentralized artificial intelligence project aimed at creating a decentralized machine learning marketplace. This enables decentralized AI applications to be built and compete directly with centralized AI projects like OpenAI’s ChatGPT. The network consists of nodes (miners) that provide computational resources for training and running artificial intelligence models.

Bittensor utilizes a subnet structure, which is similar to a chain for a specific application. It currently has 32 subnets, each of which focuses on specific artificial intelligence-related tasks, including a decentralized text prompt AI network (Text Prompt AI refers to an AI application similar to ChatGPT), which can convert text prompts into Image-generating AI that translates into images, and an AI-based search engine.

Miners play a key role in the Bittensor ecosystem, providing computing resources and hosting machine learning models to perform off-chain AI task calculations and generate results. Anyone can join the network and become a miner with minimum hardware requirements. Miners compete with each other to provide the best results for users’ queries.

Network capacity and usage

Akash is initially focused on CPUs, and there are a lot of CPU resources within the network. With the rise of artificial intelligence, the demand for GPUs has increased dramatically, and Akash began adding GPU resources to its network in the third quarter of last year. However, Akash has a relatively small number of high-performance GPUs compared to other projects that focus on GPU resources. Render Network’s focus on providing decentralized GPU-based rendering solutions has allowed it to accumulate a large number of GPUs in its network.

Render Network and Akash are more mature projects, with network usage growing steadily year over year. Particularly, Akash has seen a significant increase in quarterly active leases after expanding its focus to include GPUs.

io.net is a new decentralized computing network that launched its public testnet in November 2023. Despite its shorter history, io.net has accumulated a significant number of GPUs by integrating resources from Render, Filecoin, and its network. io.net recently announced support for Apple Silicon chip clusters, allowing Apple users to allocate their unused computing power to the network, further increasing the hardware count in its network. Additionally, io.net has not yet launched its protocol token, and many hardware providers may be hoping to join the network as providers to potentially receive token airdrop opportunities.

Bittensor is a decentralized artificial intelligence network where miners contribute computational resources to the network. Miners can either invest in hardware setups themselves or simply utilize computing resources provided by cloud services. In terms of hardware count, Bittensor cannot be directly compared to typical decentralized computing networks, as it currently boasts over 7,000 miners.

Token economy

Decentralized computing platforms act as two-sided markets, with users paying fees to computing resource providers. Akash, Render Network, and Bittensor have all issued their respective tokens as a medium for exchanging value within their ecosystems. Render and Bittensor implement a token-burning mechanism to enhance token value accumulation.

Akash

Akash is an independent PoS blockchain and $AKT is its native token used for staking to ensure the security of the network and pay for network fees. The token also serves as a medium of exchange in the ecosystem, with $AKT being the primary unit of pricing when users trade or lease on Akash. As a PoS chain, Akash needs to generate block rewards for validator nodes by issuing $AKT, and the current inflation rate is about 14%.

Akash currently charges 4% on fees paid in AKT, or 20% if paid in USDC, which will flow into the community pool. Specific uses for community pool funds have yet to be determined, but potential uses could include public funding, incentives, or simply burning the tokens.

Render Network

Render Network has migrated from Ethereum to Solana, and its protocol token RNDR is used for value exchange within the Render ecosystem, with creators and users using the token to pay for rendering jobs.

To balance the dynamic relationship between supply and demand of computing resources, Render implements a Burning and Minting Balance (BME) mechanism. When demand (i.e. rendering jobs) exceeds the supply of computing resources, RNDR tokens will be burned, creating a deflationary effect. Conversely, if the supply of computing resources exceeds demand, more RNDR tokens will be minted, causing inflation. The RNDR token is inflated due to the current lack of computing demand.

Bittensor

Bittensor’s native token $TAO is used to access network services and serves as a medium for the core reward mechanism. The maximum supply of $TAO is 21 million, and 7,200 tokens are generated daily as rewards to miners and validator nodes. Bittensor implements a token issuance halving mechanism, which means that when half of the total supply is distributed, the issuance rate will be halved. After the first halving, subsequent halvings will occur after half of the remaining token supply is distributed until the maximum supply of 21 million is reached.

Although the issuance rate of 7,200 TAO per day is fixed during the current period, the time of the next halving is not predetermined due to the token recycling mechanism. This recycling mechanism burns issued TAO tokens, effectively delaying the point at which half of the total supply is distributed. Miners and verification nodes need to recycle (i.e. burn) TAO tokens to register into the network. These burned tokens will be deducted from the circulating supply and can be mined again. The network regularly deregisters miners and validator nodes that cannot provide sufficiently competitive AI tasks, and miners need to pay/burn TAO again when they re-enter the network, making registration a recurring cost. This dynamic burning mechanism creates a constant demand for TAO.

The first planned halving date was originally planned for January 2025, but the current halving date has been postponed to October 2025. It shows that a large number of TAO tokens have been burned.

Statement:

1. This article originally titled “DePIN x AI - Overview of Four Major Decentralized Computing Networks” is reproduced from [tokeninsigh]. All copyrights belong to the original author [0xEdwardyw]. If you have any objection to the reprint, please contact the Gate Learn team, the team will handle it as soon as possible.

2. Disclaimer: The views and opinions expressed in this article represent only the author’s personal views and do not constitute any investment advice.

3. Translations of the article into other languages are done by the Gate Learn team. Unless mentioned, copying, distributing, or plagiarizing the translated articles is prohibited.