Is the windfall a false proposition? What is the value of the AI + Web3 track?

Written by: TinTinLand

As we enter 2025, the narrative of "AI + Web3" remains strong. According to the latest report released by Grayscale in May 2025, the overall market value of the AI Crypto sector has reached $21 billion, achieving a nearly fivefold increase compared to $4.5 billion in the first quarter of 2023.

Behind this wave, is it a true integration of technology or just another concept packaging?

From a macro perspective, the traditional AI ecosystem has increasingly revealed structural problems: high barriers to model training, lack of data privacy protection, highly monopolized computing power, black-box inference processes, and imbalanced incentive mechanisms... These pain points are precisely aligned with the native advantages of Web3: decentralization, open market mechanisms, on-chain verifiability, and user data sovereignty.

The combination of AI and Web3 is not just the stacking of two popular terms, but a structural technological complement. Let's start from the major core pain points currently faced by AI, and delve into those Web3 projects that are effectively solving problems, helping you to see the value and direction of the AI Crypto track.

The access threshold for AI services is too high and the costs are expensive.

Currently, AI services are often costly, and obtaining training resources is difficult, creating a high barrier to entry for small and medium-sized enterprises and individual developers. Additionally, these services are often technically complex and require a professional background to get started. The AI service market is highly concentrated, with users lacking diverse options, opaque calling costs, unpredictable budgets, and even facing issues of computing power monopolies.

The solution of Web3 is to break down platform barriers through decentralization, building an open GPU market and model service network, supporting flexible scheduling of idle resources, and incentivizing more participants to contribute computing power and models through on-chain task scheduling and a transparent economic mechanism, thus reducing overall costs and improving the accessibility of services.

Representative project

Render Network: Focused on decentralized GPU rendering, also supports AI inference and training, adopting a "pay-as-you-go" model to help developers access image generation and AI services at a low cost.

Gensyn: Building a decentralized deep learning training network that uses the Proof-of-Compute mechanism to verify training results, promoting the shift of AI training from platform centralization to open collaboration.

Akash Network: A decentralized cloud computing platform based on blockchain technology, where developers can rent GPU resources on demand for deploying and running AI applications, considered the "decentralized version of cloud computing."

0G Labs: A decentralized AI-native Layer-1 that significantly reduces the cost and complexity of running AI models on-chain through an innovative storage and computing separation architecture.

Data contributors lack incentives

High-quality data is the core fuel for AI models, but under traditional models, data contributors find it difficult to receive rewards. The lack of transparency in data sources, high redundancy, and the lack of feedback on usage methods lead to a long-term inefficient operation of the data ecosystem.

Web3 provides a new paradigm of solutions: through cryptographic signatures, on-chain verification of rights, and composable economic mechanisms, it creates a clear collaboration and incentive loop among data contributors, model developers, and users.

Representative Project

OpenLedger: Innovatively proposes the concept of "Payable AI", combining data contribution, model invocation, and economic incentives to promote the formation of an AI collaborative data economy network on the blockchain.

Bittensor: A complete incentive system centered around TAO rewards, Yuma consensus mechanism, subnet precise incentives, and knowledge collaboration, directly linking data contributions with model operation results to enhance overall value contribution.

Grass: An AI data network that collects user browsing behavior data through plugins, contributing to the training of a blockchain search engine. Users are rewarded based on data quality, creating a community-driven data sharing mechanism.

Model black-boxing, AI inference cannot be verified.

The inference process of current mainstream AI models is highly opaque, making it difficult for users to verify the accuracy and reliability of the results, especially in high-risk areas such as finance and healthcare where issues are more pronounced. In addition, models may be subject to tampering, poisoning, and other attacks, making traceability or auditing challenging.

To this end, Web3 projects are attempting to introduce Zero-Knowledge Proofs (ZK), Fully Homomorphic Encryption (FHE), and Trusted Execution Environments (TEE) to ensure the model inference process has verifiability and auditability, thereby enhancing the interpretability and trust basis of AI systems.

Representative project

Sentient: Ensures that invocation behavior is traceable through innovative model fingerprinting technology, enhancing model usage transparency and tamper-resistance.

Modulus Labs: Utilizing ZK technology for encrypted verification of model inference processes, achieving a new paradigm of "Trustworthy AI".

Giza: Utilizing zero-knowledge cryptography to bring machine learning inference calculations on-chain, thereby enhancing the transparency and trust of AI model deployment.

Privacy and security risks

The AI training process often involves a large amount of sensitive data, facing risks such as privacy breaches, model misuse or attacks, and lack of decision-making transparency. At the same time, the ownership of data and models is ambiguously defined, further exacerbating security concerns.

By leveraging the immutability of blockchain, cryptographic computing technologies (such as ZK, FHE), and trusted execution environments, ensure the security and controllability of AI systems' data and models throughout the entire process of training, storage, and invocation.

Representative project

Phala Network: Provides Trusted Execution Environment (TEE) support, encapsulating critical computations within secure hardware to prevent data leakage and model theft.

ZAMA: Focuses on Fully Homomorphic Encryption (FHE) technology, allowing model training and inference to be conducted in an encrypted state, achieving "calculations without needing plaintext."

Mind Network: Build a decentralized AI data sharing and inference platform that supports privacy protection, achieving secure data sharing and privacy computing through cutting-edge cryptographic technologies (such as homomorphic encryption, zero-knowledge proofs, etc.).

Vana: An AI identity generation application designed to empower users with ownership and control over their own data, ensuring the privacy and security of that data.

Copyright and intellectual property disputes of AI models

Current AI models heavily rely on internet data for training, but often use copyrighted content without authorization, leading to frequent legal disputes. At the same time, the copyright ownership of AI-generated content is unclear, and there is a lack of transparent mechanisms for distributing rights among creators, model developers, and users. Cases of models being maliciously copied and misappropriated are also common, making intellectual property protection difficult.

Web3 uses on-chain certification mechanisms to store evidence of the model's creation time, training data sources, contributor information, etc., and employs tools such as NFTs and smart contracts to identify the copyright ownership of the model or content.

Representative project

Story Protocol: Build on-chain intellectual property agreements that allow for the modular authentication, combination, and licensing of AI content, code, models, etc., implementing a mechanism of "creation equals authentication, usage equals payment."

Alethea AI: A generative AI model (such as characters, voices, etc.) linked to on-chain identities and NFTs, where each AI character has clear creator and copyright information, preventing abuse and plagiarism.

Lack of decentralized AI governance

The development and evolution of current AI models heavily rely on large tech companies or closed teams, with an opaque model update rhythm and difficult-to-correct value biases, which can lead to algorithmic bias, misuse, and a trend of "technological feudalism." Communities and users often cannot intervene in the model's update paths, parameter adjustments, or behavioral boundaries, lacking mechanisms for effective supervision and correction of AI systems.

The advantages of Web3 lie in programmable governance and open collaboration mechanisms. With on-chain governance, DAO mechanisms, and incentive structures, key aspects such as the design, training objectives, and parameter updates of AI models can gradually incorporate community consensus, enhancing the democracy, transparency, and diversity of model development.

Representative project

Fetch.ai: Introduces Autonomous Economic Agents (AEA) and an open governance mechanism, allowing the behavior of AI agents to be constrained by community rules and coordinating cooperation between agents through economic incentives.

SingularityNET: Encapsulating AI services into composable on-chain modules, allowing users to choose or substitute models in an open market, while the platform governance mechanism supports consensus evaluation and improvement proposals for model quality and services.

Cross-chain AI collaboration issues

In a multi-chain environment, AI agents and models may be distributed across different blockchains, making it difficult to unify state, context, or invocation logic, leading to a fragmented user experience, complex development, and challenging data synchronization.

Some projects are exploring the "Multi-Chain AI Protocol", attempting to promote the continuity and consistency of AI agents operating across chains through shared context, cross-chain communication, and state synchronization mechanisms.

Representative Project

OpenPond: Utilizes the MCP cross-chain protocol to connect AI models and agents across different chains, achieving synchronization of call status and context sharing, simplifying multi-chain collaboration scenarios.

Lava Network: Provides cross-chain RPC and data bridge services to facilitate underlying communication channels for multi-chain AI systems, supporting agent data synchronization and unified task execution.

Virtuals Protocol: Through the ACP (Agent Commerce Protocol) smart collaboration protocol, it supports cross-agent requests, negotiations, execution, and settlement processes. Its "Parallel Hypersynchronicity" parallel synchronization technology enables AI agents to run concurrently across platforms, synchronizing behaviors and memories in real-time.

Conclusion

The rise of AI Crypto is not an empty talk, but a bottom-up system reconstruction: it breaks the shackles of centralization in the era of large models, and gradually builds a new AI paradigm that is participatory, transparent, trustworthy, and collaboration-driven across dimensions such as computing power, data, incentives, security, and governance.

The field has now transitioned from the conceptual stage to the substantive product implementation phase. It is believed that AI Crypto projects that can genuinely create actual value and address core pain points will have the opportunity to lead the next wave of development in the AI era, promoting the advancement of artificial intelligence technology towards a more open, fair, and trustworthy direction.