The Fusion of AI and DePIN: The Rise of Distributed GPU Networks

Since 2023, AI and DePIN have become hot trends in the Web3 space, with market values reaching $30 billion and $23 billion respectively. This article focuses on the intersection of the two and explores the development of this emerging field.

In the AI technology stack, the DePIN network empowers AI by providing computing resources. The demand for GPUs from large tech companies has led to supply shortages, making it difficult for other developers to obtain enough resources to train their own models. Traditional centralized cloud services often require signing inflexible long-term contracts, which are inefficient. The DePIN network offers a more flexible and cost-effective alternative by aggregating decentralized GPU resources through token incentives, providing users with a unified supply. This not only allows developers to access customizable computing power on demand but also creates additional revenue for users with idle GPUs.

AI DePIN Network Overview

Render

Render is a pioneer in the P2P GPU computing network, originally focused on content creation graphics rendering, and later expanded to a wide range of AI computing tasks including generative AI.

Highlights:

• Founded by a company with Oscar-winning technology
• Used by entertainment giants such as Paramount Pictures
• Collaborate with companies like Stability AI to integrate AI models with 3D content rendering.
• Supports multiple computing clients and integrates more DePIN network GPUs.

Akash

Akash is positioned as a "super cloud" alternative to traditional cloud platforms, supporting storage, GPU, and CPU computing. Its container platform and Kubernetes-managed compute nodes can seamlessly deploy cloud-native applications.

Highlights:

• Covers a wide range of tasks from general computing to web hosting
• AkashML supports running over 15,000 models on Hugging Face.
• Well-known applications such as the LLM chatbot from Mistral AI have been hosted.
• Supports platforms such as metaverse, AI deployment, and federated learning.

io.net

io.net provides a distributed GPU cloud cluster, focusing on AI and ML use cases. It aggregates GPU resources from multiple sources such as data centers and crypto miners.

Highlights:

• IO-SDK is compatible with frameworks such as PyTorch and can be dynamically expanded according to needs.
• Support creating 3 different types of clusters, start within 2 minutes.
• Collaborate with Render, Filecoin, and others to integrate more GPU resources.

Gensyn

Gensyn focuses on machine learning and deep learning computations. It improves verification efficiency through mechanisms such as proof of learning, graph-based protocols, and staking incentives.

Highlights:

• The cost of V100 GPU is about $0.40 per hour, significantly saving costs.
• Fine-tune the pre-trained base model to complete specific tasks.
• Provide a decentralized, globally shared foundational model

Aethir

Aethir focuses on enterprise-level GPUs, serving computation-intensive fields such as AI, ML, and cloud gaming. It utilizes container technology to shift workloads from local to cloud, achieving low-latency experiences.

Highlights:

• Expand to cloud phone services and launch decentralized cloud smartphones in cooperation with APhone.
• Establish extensive cooperation with Web2 giants such as NVIDIA and Foxconn.
• Collaborating with multiple parties such as CARV and Magic Eden in the Web3 space

Phala Network

Phala Network, as the execution layer of Web3 AI solutions, addresses privacy issues using Trusted Execution Environment (TEE) design. Its execution layer allows AI agents to be controlled by on-chain smart contracts.

Highlights:

• As a co-processor protocol for verifiable computing, empowering AI agents with on-chain resources.
• AI agent contracts can access top LLMs like OpenAI through Redpill
• The future will include multiple proof systems such as zk-proofs, MPC, and FHE.
• Plan to support H100 and other TEE GPUs to enhance computing power.

Project Comparison

| | Render | Akash | io.net | Gensyn | Aethir | Phala | |--------|-------------|------------------|---------------------|---------|---------------|----------| | Hardware | GPU & CPU | GPU & CPU | GPU & CPU | GPU | GPU | CPU | | Business Focus | Graphics Rendering and AI | Cloud Computing, Rendering, and AI | AI | AI | AI, Cloud Gaming, and Telecommunications | On-chain AI Execution | | AI Task Type | Inference | Inference and Training | Inference and Training | Training | Training | Execution | | Work Pricing | Performance-Based Pricing | Reverse Auction | Market Pricing | Market Pricing | Bidding System | Equity Calculation | | Blockchain | Solana | Cosmos | Solana | Gensyn | Arbitrum | Polkadot | | Data Privacy | Encryption& Hashing | mTLS Authentication | Data Encryption | Secure Mapping | Encryption | TEE | | Work Fees | 0.5-5% per job | 20% USDC, 4% AKT | 2% USDC, 0.25% reserve fee | Low fees | 20% per session | Proportional to the staked amount | | Security | Render Proof | Proof of Stake | Proof of Calculation | Proof of Stake | Render Capability Proof | Inherited from Relay Chain | | Completion Proof | - | - | Time-Lock Proof | Learning Proof | Rendering Work Proof | TEE Proof | | Quality Assurance | Dispute | - | - | Verifier and Reporter | Checker Node | Remote Proof | | GPU Cluster | No | Yes | Yes | Yes | Yes | No |

Key Features Comparison

Cluster and Parallel Computing

The distributed computing framework implements GPU clusters to improve training efficiency and scalability. Most projects have integrated cluster support for parallel computing to meet the demands of complex AI models. io.net has successfully deployed over 3,800 clusters. Although Render does not support clusters, it can break tasks down to multiple nodes for simultaneous processing. Phala supports CPU worker clustering.

Data Privacy

Protecting sensitive datasets is crucial for AI development. Most projects use data encryption to safeguard privacy. io.net introduces fully homomorphic encryption (FHE), which allows data to be processed in an encrypted state. Phala Network uses a Trusted Execution Environment (TEE) to prevent external access or modification of data.

Completion Certificate and Quality Inspection

To ensure service quality, most projects adopt completion certificates and quality inspection mechanisms. Gensyn and Aethir generate work completion certificates and conduct quality inspections. io.net certifies that the rental GPU performance is fully utilized. Render recommends using a dispute resolution process to handle problem nodes. Phala generates TEE certificates to ensure correct execution.

Hardware Statistics

| | Render | Akash | io.net | Gensyn | Aethir | Phala | |-------------|--------|-------|--------|------------|------------|--------| | Number of GPUs | 5600 | 384 | 38177 | - | 40000+ | - | | Number of CPUs | 114 | 14672 | 5433 | - | - | 30000+ | | H100/A100 Quantity | - | 157 | 2330 | - | 2000+ | - | | H100 Cost/Hour | - | $1.46 | $1.19 | - | - | - | | A100 Cost/Hour | - | $1.37 | $1.50 | $0.55 ( expected ) | $0.33 ( expected ) | - |

High-performance GPU Demand

AI model training requires the best-performing GPUs, such as the NVIDIA A100 and H100. The decentralized GPU market needs to provide a sufficient number of high-performance hardware to meet demand. io.net and Aethir each have over 2000 H100/A100 units, making them more suitable for large model computations. The rental costs of GPUs in these networks are already far lower than centralized services.

Consumer-grade GPU/CPU supply

In addition to enterprise-level GPUs, some projects like Render, Akash, and io.net also cater to the consumer-grade GPU market. This can leverage a large amount of idle consumer GPU resources to develop specific market segments.

Conclusion

The AI DePIN field is still in its early stages and faces numerous challenges. However, the number of tasks executed on these networks and the quantity of hardware have significantly increased, highlighting the demand for alternatives to traditional cloud services. In the future, as the AI market continues to grow, these distributed GPU networks are expected to play a key role in providing developers with cost-effective computing resources, making important contributions to the future landscape of AI and computing infrastructure.