Forward the Original Title ‘Foresight Ventures：去中心化云计算的革命才刚刚开始？’

With the long-term high-speed development of global technology, the market value of giant companies like OpenAI and NVIDIA has multiplied in the past two years. Crypto x AI has become the core narrative of this cycle, with the rising market sentiment and continuous inflow of capital proving that a strong consensus has been formed. In an environment where AI is the focus, decentralization as a powerful tool for AI development indeed holds great appeal and imagination. Although there is still a significant gap between decentralized and centralized models in terms of practical business implementation, leveraging the advantages of Web3 to expand the four core aspects of AI and continuously optimizing to unleash greater potential has become a common goal for Web3 participants.

1. Data
2. Model
3. Training
4. Inference

Currently, decentralization can provide support in the four areas mentioned above. First and foremost, data is the core element. Models, training, and inference are all ways of processing data, so data can be considered the raw material of AI technology, while the others are processing methods. Whether it is data labeling or data storage, decentralization plays a significant role and has immense value here.

If data is the raw material, then computing power is the tool for processing this raw material, used to maximize output efficiency. Moving directly to the theme of this article, we will briefly analyze the ecosystem framework and economic model of Crypto x AI x DePIN, focusing on “computing power.”

This article will mainly explain the ecosystem framework and market situation of “Crypto x AI x DePIN” to help readers understand the value and potential of decentralized computing power.

1. DePIN & Decentralized Computing Power Ecosystem Framework

Pain Point: High-quality computing power is essential for AI development, but this scarce resource has been monopolized by traditional giants, making it difficult for startups and individual users to purchase cost-effective computing power. The exorbitant prices are often unacceptable for most buyers.

Decentralized Solution: Currently, many projects in the DePIN sector adopt a peer-to-peer (P2P) economic model to provide high-quality resources to resource-demanding parties. This model allows every user to act as a provider of physical infrastructure resources while receiving token rewards in return.

With the explosive growth in demand for decentralized AI computing power, the development of a balanced and comprehensive framework for decentralized AI computing power supply ecology has emerged to better meet customer needs. Leading projects such as Io.net, Exabit, and PingPong play different vital roles in the ecosystem. The technological barriers and future development prospects of these projects in decentralized computing power are quite impressive.

The decentralized AI computing power ecosystem mainly consists of three parts, with each part serving as a resource agent, resource provider, and channel partner in the ecosystem, respectively:

Resource Agent - Io.net

Io.net is a decentralized computing network that acts as a computing power agent, providing high-quality AI computing power to customers at affordable prices. It has a global distribution of GPUs on the supply side and currently focuses on AI inference startups from seed to Series B rounds.

Recently, this DePIN project based on the Solana chain completed a $30 million Series A financing round, led by Hack VC, with participation from Multicoin Capital, Foresight Ventures, Solana Labs, and others.

As a top-tier AI computing power resource agent, Io.net aggregates 1,000,000 GPUs to form a massive DePIN computing power network, aiming to provide customers with lower-priced computing power. Users can manually contribute their idle GPU & CPU computing power to the Io.net platform to receive incentives in $IO tokens. The core goal is to provide high-quality AI computing power at a lower cost under decentralized price control, thereby helping AI startups reduce costs.

The computing service provided by Io.net is called IO Cloud. IO Cloud adopts a cluster-building module to keep all GPUs interconnected, enabling GPUs to coordinate on a large scale during training and inference processes. With coordinated GPU work, the computing power can be centralized to access larger databases and compute more complex models. AI startups can achieve their required computing needs while completing hardware deployments at one-tenth of the centralized price by using Io.net’s products. Moreover, Io.net focuses on aggregating machine learning computing power. It can help DePIN giants like Render Network and FileCoin format GPU supplies for machine learning, providing fundamental and direct resource support at the technical level.

Currently, Io.net has the largest number of aggregated GPU clusters in the industry. The number of GPUs available online exceeds 200,000, with GeForce RTX 4090 being the most available, with nearly 50,000 units, followed by GeForce RTX 3090 Ti with over 30,000 units.

Resource Provider - Exabit

As one of the most promising AI computing power providers, Exabits, as an AI computing power service node, can offer an abundance of chips for deep machine learning. The team at Exabits stands out in the traditional AI computing power resources and can be considered a unique presence. With a history as a primary distributor for AI giant NVIDIA, the team has access to hundreds of data centers, providing access to machines such as A/H100, RTX4090, and A6000, leveraging these technological resource barriers at the supply end.

Exabits provides massive machine learning computing power to web3 computing giants on the client side. Compared to Nebula Block, whose clients need to spend over $140,000 per month on cloud services, migrating to Exabits reduces monthly cloud service usage fees to around $40,000, reducing costs by over 70% while increasing efficiency by 30%.

Exabits aims to provide customers with the fastest, highest quality, and most reliable computing power through its unique supply channels. High-quality computing power not only saves user costs but also offers comprehensive service choices to customers.

The quality of AI computing power provided by Exabits has been recognized by multiple AI computing power agents. It has now partnered with giants like Renders Network and Io.net, contributing to machine learning through decentralization.

Resource Channel Partner (Uber) - PingPong

As a DePIN resource channel partner, PingPong matches demand with service providers by employing a platform-based open protocol. It aggregates resources at the base layer before offering services. PingPong aims to become a service aggregator for DePIN, akin to 1inch for DePIN, or an aggregated Uber.

Service Provision: PingPong provides SDKs by accessing various networks, strategies, resource situations, performance, stability, and other aspects through a control layer. It then employs routing algorithms to provide these SDKs to users.

Pain Point: Resources and services within various DePIN networks are limited, and the quality of service suffers due to centralized resource allocation in certain regions.

Solution: Routing algorithms gather basic information about data, networks, and machine resources to generate strategies. These algorithms match service provision according to customer requirements. The goal is to improve the quality and service of DePIN’s application layer and to find the most optimal computing power network at the best price when resources are insufficient.

2. Analyzing the Decentralized Computing Power Ecosystem

Io.net and Exabits have entered into a strategic partnership, with Exabits serving as the supply side with an extensive library of GPUs, committed to enhancing the speed and stability of the Io.net network. Io.net allows customers to directly purchase and lease the highest quality computing power provided by Exabits as an agent on the Io.net network. Both Io.net and Exabits agree that the success of the decentralized computing industry and the integration of web3 with AI can only be achieved through close cooperation among early industry leaders. With the growing demand for computing power, traditional cloud computing currently faces some issues:

• Limited Availability: Using cloud services like AWS, GCP, and Azure often requires several weeks to gain access to hardware, and the most commonly used GPU models are often unavailable.
• Limited Choices: Users are constrained in their selection of GPU hardware, location, security level, latency, etc.
• High Costs: The cost of selecting high-quality GPUs is expensive, and monthly expenses for projects during training and inference processes can easily reach hundreds of thousands of dollars.

The vision of decentralized computing is to provide an open, accessible, and affordable alternative that can address the core issues of centralized cloud service providers, including limited availability, restricted hardware choices, and high costs for training and inference. Given the current situation, challenging the dominant positions of major players in cloud computing still requires innovators to work together and support each other to take revolutionary steps forward.

Asset model

• Asset-heavy model

Exabits, backed by NVIDIA, holds an absolute advantage as a supplier in the form of a formidable barrier. The valuable machines for machine learning computing power include the A100, RTX4090, and H100, with each machine priced at around $300,000. These machines have become highly scarce resources monopolized by traditional AI giants. In this scenario, the resources that Exabits can access on the supply side are extremely valuable.

Given that individual GPU computing power shared by retail investors is insufficient to support the computation and processing of large-scale AI models, Exabits plays a crucial and irreplaceable role in the decentralized computing power ecosystem.

Exabits’ heavy asset model requires substantial fixed asset investment, making it difficult for startups to replicate. Therefore, if Exabits can collaborate with more decentralized computing power agents and continuously expand its supply side, providing the required computing power resources to the industry, it can easily achieve industry dominance and economies of scale in the B2B decentralized computing power field.

However, the greatest risk lies in the inability to sustainably provide resources to computing power agents after investing a significant amount of capital. Therefore, the profitability of the supply side heavily depends on whether computing power agents can continuously attract customers. Regardless of the identity of the computing power agent, as long as there are customers and demand, the value of Exabits as a supplier will grow with the increasing demand.

• Asset-light model

Io.net, as the leading computing power agent, relies on a vast decentralized computing network formed by GPUs distributed globally on the supply side. From a business perspective, Io.net adopts a light asset operating model, establishing a strong brand in the AI computing power agency sector through community operations and building high consensus.

The core business of Io.net:

1. Aggregating retail GPU computing power and rewarding tokens
2. Obtaining high-quality computing power from the supply side for sale to AI startups

Business perspective:

1. Buying low and selling high-quality computing power to end customers from the supply side
2. Assisting users in earning tokens by sharing idle GPU computing power
3. Providing customers with a computing power mining and staking platform, which requires an initial investment of around $4000 to generate good returns. Based on this, Exabits also offers fragmented H100 machines for rent to increase liquidity.

Customer perspective:

1. The computing power prices on the Io.net network are approximately 80% cheaper than other centralized cloud computing services.
2. Stake to earn & Share to earn.
3. Customers can compound their earnings after investing a certain amount of capital.

As a typical light asset model company, the greatest advantage lies in relatively low risk. The team does not need to invest heavily in machine costs to start, unlike on the supply side. With fewer capital inputs, it is easier for both the company and investors to achieve higher profit margins. Meanwhile, due to the low barriers to entry in the industry, the business model is susceptible to being copied or replicated, which is something long-term value investors need to consider carefully.

3. From 10 to 100?

If the collaboration between Exabit and Io.net could elevate the decentralized computing power ecosystem from 1 to 10, then bringing PingPong along might have the potential to advance it to 100.

PingPong aims to become the largest DePIN service aggregator, directly competing with Web2’s Uber. As a channel partner, it connects customers with the best-priced and highest-quality resources by aggregating real-time information on various resources. PingPong adopts a B2B2C light asset business model, where the first B-side is the supply side, connecting with the second B-side, which is the resource agent, while the C-side receives optimal resource selections through the information provided.

As a platform, if the channel partner can develop a platform that can issue assets as much as possible, it will make the product more valuable. PingPong utilizes the SDK provided by the routing algorithm to use computing resources to create its own AI agents, while dynamically assisting customers using applications with dynamic mining through the SDK, focusing on mining computing resources that are useful. This model is understood as “Assets on assets,” which can greatly enhance the liquidity of resources and funds.

PingPong hopes to see more suppliers and agents enter the decentralized computing power ecosystem to better highlight their advantages, expand longer business lines, and acquire more customers. Simply put, just as Baidu and Dianping are able to dominate the information field because there are more businesses and information uploaded to the Internet, this increases the demand for channel partners by customers.

4. The Future Is Promising

Decentralized cloud computing is still evolving step by step. While the ecosystem and models of decentralized cloud computing have become very clear, the leading roles in the ecosystem are fulfilling their responsibilities. However, it is still far from shaking the position of traditional cloud computing giants. When compared to traditional centralized cloud computing, decentralization can indeed conceptually solve many problems for customers. However, the overall resources and scale of this market are still relatively small. With the computing power resources driven by AI far from sufficient, the market needs another solution or model to overcome this dilemma. We can see that decentralized cloud computing can indeed meet some needs of startup AI companies. As for what lies ahead, let us together witness and participate in the evolution of this disruptive path as observers and participants of the revolution.

Disclaimer：

1. This article is reprinted from [Foresight Research]. Forward the Original Title ‘Foresight Ventures：去中心化云计算的革命才刚刚开始？’. All copyrights belong to the original author [David]. If there are objections to this reprint, please contact the Gate Learn team, and they will handle it promptly.
2. Liability Disclaimer: The views and opinions expressed in this article are solely those of the author 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.