The Fusion of AI and DePIN: Exploring a New Pattern of Computing Resources

Since 2023, AI and DePIN have received significant attention in the Web3 space, with market capitalizations reaching 30 billion dollars and 23 billion dollars, respectively. This article focuses on the development of their intersection.

In the AI technology stack, the DePIN network empowers AI by providing computing resources. The high demand for GPUs from large tech companies has led to shortages, putting other AI model developers in a predicament of insufficient computing resources. Traditional solutions such as choosing centralized cloud service providers face issues of inflexibility and high costs.

The DePIN network provides a more flexible and cost-effective alternative. It integrates individual GPU resources into a unified supply through a token incentive mechanism, offering customized and on-demand computing power to demand-side users while generating additional income for GPU resource owners with idle resources.

Various AI DePIN networks are emerging in the market. Below, we will explore the characteristics and development status of several typical projects.

AI DePIN Network Overview

Render

Render is a pioneer in the P2P GPU computing network, initially focusing on graphics rendering for content creation, and later expanding to AI computing tasks.

Main features:

• Founded by the Oscar-winning cloud graphics company OTOY
• Adoption by entertainment industry giants such as Paramount Pictures and PUBG
• Collaborate with Stability AI and integrate AI models with 3D rendering workflows
• Support multiple computing clients, integrating more DePIN network GPU resources.

Akash

Akash is positioned as a "supercloud" platform that supports storage, GPU, and CPU computing, serving as an alternative to traditional cloud services.

Main features:

• A wide range of computing tasks from general computing to web hosting
• AkashML supports running over 15,000 models on Hugging Face.
• Multiple well-known AI applications have been hosted, such as Mistral AI's LLM chatbot.
• Metaverse, AI deployment, and federated learning platforms are using its services.

io.net

io.net provides distributed GPU cloud clusters, focusing on AI and ML application scenarios.

Main features:

• IO-SDK is compatible with frameworks such as PyTorch and Tensorflow.
• Supports the creation of 3 different types of clusters, which can be launched within 2 minutes.
• Actively integrate GPU resources from other DePIN networks such as Render and Filecoin.

Gensyn

Gensyn focuses on GPU computing power for machine learning and deep learning.

Main features:

• The equivalent cost of V100 GPU is approximately $0.40 per hour, significantly reducing costs.
• Support fine-tuning of pre-trained base models
• Decentralized global shared foundation model

Aethir

Aethir focuses on enterprise-level GPUs, specializing in computation-intensive fields such as AI, machine learning, and cloud gaming.

Main Features:

• Expand to cloud phone services, collaborating with APhone to launch a decentralized cloud smart phone.
• Establish extensive cooperation with Web2 giants such as NVIDIA and Super Micro
• Collaborating with multiple Web3 projects such as CARV, Magic Eden

Phala Network

Phala Network serves as the execution layer for Web3 AI solutions, addressing privacy issues through a trusted execution environment (TEE).

Main Features:

• As a co-processor protocol for verifiable computation, supporting AI agents for on-chain resource calls.
• AI proxy contracts can be integrated with top language models such as OpenAI and Llama through Redpill.
• In the future, it will support zk-proofs, multi-party computation, fully homomorphic encryption, and other multi-proof systems.
• 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 | Artificial Intelligence, Cloud Gaming and Telecommunications | On-chain AI Execution | | AI Task Type | Inference | Both | Both | 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 staking amount | | Security | Render Proof | Proof of Stake | Proof of Work | 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 |

Importance Analysis

Availability of cluster and parallel computing

The distributed computing framework implements GPU clusters, improving training efficiency and scalability without affecting model accuracy. Complex AI model training requires powerful computing capabilities, often relying on distributed computing. Most projects have now integrated clusters for parallel computing. io.net has integrated GPU resources with multiple partners and has deployed over 3,800 clusters in the first quarter of 2024.

Data Privacy

AI model development requires a large dataset, which may involve sensitive personal information. Various projects commonly use data encryption to protect privacy. io.net has collaborated with Mind Network to launch fully homomorphic encryption (FHE), allowing data to be processed in an encrypted state. Phala Network introduces a trusted execution environment (TEE), isolating to prevent external access or modification of data.

Completion certificate and quality inspection

Different projects use various methods to verify the completion and quality of calculations. Gensyn and Aethir generate completion proofs and perform quality checks. The proofs from io.net indicate that GPU performance is fully utilized and without issues. Render recommends using a dispute resolution process. Phala generates TEE proofs to ensure that AI agents perform the required operations.

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 ( Estimated ) | $0.33 ( Estimated ) | - |

Requirements for high-performance GPUs

AI model training tends to use high-performance GPUs such as Nvidia A100 and H100. Decentralized GPU market providers need to offer a sufficient number of high-performance hardware to meet market demand. io.net and Aethir have over 2000 H100/A100 units, which are more suitable for large model computations.

The costs of these decentralized GPU services have fallen below those of centralized services. Gensyn and Aethir claim that A100-level hardware can be rented for less than $1 per hour.

Provide consumer-grade GPU/CPU

The CPU also plays an important role in AI model training. Consumer-grade GPUs can be used for fine-tuning or small-scale model training. Projects like Render, Akash, and io.net can serve this market, providing options for computing needs of different scales.

Conclusion

The AI DePIN field is still in its early stages and faces numerous challenges. However, the significant increase in the tasks and hardware executed by these decentralized GPU networks highlights the demand for alternatives to Web2 cloud services.

The future AI market is expected to grow into a trillion-dollar scale, and these decentralized GPU networks are likely to play a key role in providing developers with cost-effective computing alternatives. By continuously bridging the gap between demand and supply, these networks will make significant contributions to the future landscape of AI and computing infrastructure.