Infini View | The Genesis Coupling of AI and Crypto

——Exploring the Future Landscape and Investment Opportunities of Decentralized Intelligence

Introduction — The Convergence of Paradigm Revolutions

Research Background: The Inevitable Encounter of Two Technological Waves

The modern world is being swept by two powerful technological waves: the explosive rise of Artificial Intelligence (AI) and the ongoing evolution of Cryptography-based technologies (Crypto). AI, as the next-generation productivity tool, is reshaping industries at unprecedented speed; Crypto, meanwhile, is constructing a decentralized global value network. Though their trajectories and application domains seem divergent, the evolution of their underlying logics is propelling them toward a historic point of convergence.

The Centralization Dilemma of AI:

Current AI development heavily depends on a handful of tech giants, forming a closed loop of “data–models–compute.” This leads to issues such as data monopolies, opaque algorithms, soaring compute costs, and single points of failure. Such centralization restricts innovation while raising concerns about privacy, bias, and concentrated power.

Crypto’s Value Proposition:

At its core, Crypto is a “trust machine” built on cryptography, characterized by decentralization, transparency, censorship resistance, and programmable value. These traits offer novel solutions to AI’s centralization crisis.

Core Thesis: The fusion of AI and Crypto is not accidental but inevitable. Crypto is not meant to replace AI’s “forces of production” (compute, algorithms, data), but to provide a new “mode of production relations”—a decentralized, trustless, and incentive-aligned network and economic model. With these new production relations, AI’s productive capacity can be more fairly and widely unleashed and combined.

Core Principles The synergy of AI and Crypto is built upon deep complementarity, rooted in rethinking power structures and value distribution in the digital age.

• Mutual Infrastructure: AI enhances the intelligence, adaptability, and automation of Crypto applications—such as smarter DeFi strategies, more immersive GameFi experiences, and efficient DAO governance. Conversely, Crypto provides AI with a trustless, open, censorship-resistant environment, liberating models and data from the control of single entities.

• Power Balance: This convergence represents a contest between “bottom-up” decentralized forces and “top-down” centralized control. Crypto’s distributed networks counteract AI’s centralization trend, fostering a more open and pluralistic intelligence ecosystem.

• Return of Digital Ownership: In Web2, user data was captured freely by platforms. Through tokenization, Crypto can transform data and AI models into ownable, tradable digital assets (NFTs or tokens), allowing creators and contributors to capture the value they generate.

• Verifiable Intelligence: Blockchain’s immutability and transparency, combined with cryptographic tools like zero-knowledge proofs, make AI’s decision-making traceable and verifiable. This establishes the foundation for trustworthy AI, addressing the “black-box” problem.

Deep-Dive into the Technology

Overview of the Tech Stack The integration of AI and Crypto is built upon a multi-layered technology stack, combining the underlying decentralized physical infrastructure (DePIN) with upper-layer intelligent applications and economic models. This stack aims to systematically resolve the centralization bottlenecks in the current development of AI, providing a complete solution for realizing decentralized intelligence. Its core components include: compute, data, models, applications, and verifiability. Each layer is driven by innovative protocols and economic incentive mechanisms.

Decentralized Compute

Core Logic: Decentralized compute aims to break the monopoly of the compute market held by a few cloud service giants (such as AWS, GCP, Azure). Its basic principle is, through token economic incentive mechanisms, to aggregate a vast amount of idle computing resources (mainly GPUs) across the globe, forming a peer-to-peer, permissionless compute market. In this market, demanders of compute (such as AI developers, rendering artists) can directly rent compute from suppliers (such as miners, data centers, or even individual users), thereby potentially obtaining computational power at a lower cost.

Technology Layered Architecture:

• Bare Metal Layer: This is the lowest physical resource layer, consisting of globally distributed and heterogeneous computing devices, including server clusters in professional data centers, mining machines of miners, and consumer-grade graphics cards from individual users.

• Orchestration Layer: This layer is responsible for standardization, scheduling, load balancing, and task allocation of the underlying compute. It is the brain of the decentralized compute network, ensuring that suitable tasks are matched to suitable computing nodes, while guaranteeing service stability.

• Aggregation Layer: In order to capture larger network effects, some projects focus on aggregating resources from different decentralized compute networks (such as Render, Filecoin) as well as from traditional cloud services, to provide users with a unified and optimized compute access point.

Current Challenges: Despite broad prospects, decentralized compute still faces serious challenges. First, in scenarios with high requirements such as AI training, network latency and bandwidth between distributed nodes are huge bottlenecks, making it difficult to match the high-speed internal networks of centralized data centers. Second, when considering hidden costs such as networking, storage, and operations, its price advantage is not always significant. Finally, market demand for high-end GPUs (such as H100/A100) far exceeds supply, while the large amount of consumer-grade GPU compute in the network faces insufficient demand.

Verifiable Compute and ZKML (Zero-Knowledge Machine Learning)

Core Logic: Zero-Knowledge Machine Learning (ZKML) is a frontier application in the field of verifiable compute. Its goal is to, under the premise of not disclosing any sensitive information (such as input data or internal model parameters), prove through cryptographic methods (Zero-Knowledge Proof, ZKP) that the inference process of an AI model is complete and correct. This is equivalent to providing a trustworthy mathematical receipt for AI’s “black box” operations.

Core Value:

• Privacy Protection: When handling sensitive data such as medical, financial, or personal identity information, ZKML can achieve “usable but invisible,” that is, completing AI inference without exposing raw data, thereby fully protecting user privacy.

• Computation Integrity and Trust: In on-chain environments, ZKML can prove to smart contracts that a particular AI decision is derived from a specific model and input, without any tampering. This is critical for scenarios such as on-chain AI agents, decentralized identity verification (e.g., Worldcoin).

• Preventing Cheating: In decentralized AI networks, ZKML can be used to verify whether contributors (such as model operators) honestly performed compute tasks, preventing them from submitting incorrect or forged results to fraudulently obtain rewards.

Current Bottlenecks: ZKML is one of the most cutting-edge but also most immature technologies in this field. Its biggest bottleneck lies in the enormous computational overhead: the time and compute cost required to generate a proof may be hundreds or even thousands of times that of the original computation. In addition, converting complex AI models with floating-point numbers into mathematical circuits suitable for ZKP leads to accuracy loss, and current technology mostly supports only small-scale models. ZK proofs of large-scale models on-chain are still at a very early exploratory stage.

Decentralized Data and Storage

Core Logic: The intelligence level of AI is highly dependent on the quality and quantity of its training data. Decentralized data and storage aim to solve the problems of data monopolies and data silos in the Web2 world, building an open, trustable, and value-fairly-distributed data ecosystem. The core is to encourage users and organizations to contribute, verify, and share data through incentive mechanisms.

Key Technologies and Paradigms:

• Decentralized Storage Networks (DSN): Represented by Filecoin and Arweave, providing underlying infrastructure for long-term, censorship-resistant storage of massive data.

• Data Collection and Verification: Projects such as Grass incentivize users to contribute idle network bandwidth, building a distributed web crawler to capture public web data for AI training, aiming to provide a more decentralized and transparent data source for AI models.

• Data Ownership and Value Capture: By tokenizing data ownership (Data NFTs), data contributors can truly own their data, and trade or license it through data markets or data liquidity pools, gaining economic returns. This gives rise to a new paradigm of “data as an asset.”

Application Directions: In addition to providing data for large-model training, decentralized data networks also show great potential in synthetic data generation, data annotation (such as RLHF), and building verifiable data sources (ensuring the authenticity of off-chain data through technologies like TLS Notary).

AI Agent Economic Models

Core Logic: If the previous technology stack lays the infrastructure for AI, then AI Agents are the native inhabitants of this new continent. The core idea of on-chain AI Agents is to grant AI models independent economic identities (such as a crypto wallet address), enabling them to autonomously own assets, sign transactions, execute smart contracts, and interact with other agents or protocols, thus becoming native economic entities on-chain.

Key Paradigms:

• Model as Node: In some networks, the AI model itself is a first-class citizen in the network, able to participate in network consensus and governance. Its contributions (such as providing high-quality inference) are rewarded with tokens.

• Intent-Centric: This is the core paradigm of AI agent interaction. Users no longer need to execute complex multi-step operations, but instead express their ultimate goal (“intent”) in natural language. For example, “deposit ETH into Aave, and use the obtained aETH as collateral to borrow USDC.” The AI agent will automatically parse this intent, then plan and execute a series of optimal on-chain transactions to achieve it.

• Autonomous Economic Activity: AI Agents can autonomously manage funds, execute complex DeFi arbitrage strategies, participate in DAO voting, act as intelligent NPCs in on-chain games, and even call other AI services to complete more complex tasks, thereby forming a service economy among machines.

Application Scenarios: The application prospects of AI agents are extremely broad, covering automated trading (MEV capture and protection), intelligent investment advisory and asset management, task automation, decentralized oracle networks, and the construction of more complex autonomous economic systems. Projects like Fetch.ai and Olas (Autonolas) are providing core technology stacks and protocol frameworks for building such agent economies.

Ecosystem Panorama

Panoramic Sorting of Tracks

The AI + Crypto ecosystem is rapidly expanding, forming a multi-layered structure that covers infrastructure, middleware, and application layers. This emerging landscape is full of vitality and innovation, with different projects attempting to cut in from different angles to solve the core pain points of AI centralization.

• Infrastructure Layer: This is the cornerstone of the entire ecosystem, mainly including projects that provide decentralized compute (such as io.net, Akash), data (such as Grass), and storage (such as Filecoin). They aim to provide the necessary underlying resources for AI operations.

• Middleware/Protocol Layer: This layer is responsible for connecting infrastructure and upper-layer applications, providing core functional modules and protocol standards. For example, projects like Bittensor attempt to create a decentralized collaborative network for AI models; Fetch.ai and Olas focus on providing frameworks for building and interacting with AI Agents; Ritual and Modulus Labs are exploring ZKML, providing verifiability for AI computation.

• Application Layer: This is the layer directly facing users, where AI technology is integrated into various dApps to enhance their functionality and user experience. This includes AI-driven DeFi protocols, smarter wallets, on-chain game NPCs with autonomous behaviors, and various new applications based on “intent.”

At present, most of the ecosystem’s value and attention remain concentrated on the infrastructure and protocol layers, as they are prerequisites for realizing the vision of decentralized intelligence. However, as the underlying technology matures, we expect the application layer to experience explosive growth in the coming years.

Positioning Analysis of Major Participants

There are many major participants within this track, each with clear positioning, forming a complex relationship of both cooperation and competition.

• Competition and Integration in Decentralized Compute: This field has the largest number of participants and is highly competitive. Akash Network (AKT) is known as the “decentralized cloud Taobao,” providing an open cloud computing resource rental marketplace. Render Network (RNDR) cut in from the 3D rendering market, accumulating a large amount of consumer-grade GPU resources, and is actively expanding into AI inference tasks. Rising star io.net (IO) plays the role of a “compute aggregator,” not only integrating GPU resources from networks like Render and Filecoin, but also connecting geographically distributed GPUs through clustering technology to support more complex AI training and inference tasks, directly benchmarking the enterprise market.

• Decentralized AI Models and Networks: Bittensor (TAO) is the flagship project in this field. It does not provide a single model, but instead builds an incentive network where machine learning models across the globe compete and collaborate, together forming an ever-evolving “collective intelligence.” SingularityNET (AGIX) is dedicated to creating an open marketplace for AI services, allowing anyone to create, share, and monetize AI services.

• On-chain AI Agent Platforms: Fetch.ai (FET) is one of the earliest projects in this field, providing a complete set of tools for building, deploying, and connecting Autonomous Economic Agents (AEA). Olas (Autonolas, OLAS) focuses on providing a composable stack for building off-chain autonomous AI services, bringing them on-chain through oracles and other means, demonstrating strong potential in automation and complex task execution.

These projects collectively form the core backbone of the AI + Crypto ecosystem. Their development paths and dynamic relationships with one another will profoundly influence the future trajectory of the entire track.

Investment Framework and Opportunities

Market Size and Growth Forecast

As an emerging cross-domain sector, the market size of AI + Crypto needs to be estimated by comprehensively considering its two underlying macro markets: the AI market and the high-performance computing market (especially GPUs). According to forecasts by Mordor Intelligence, the global AI market size is expected to reach between USD 780 billion and USD 990 billion by 2027. At the same time, as the core hardware foundation of AI development, industry leaders such as AMD predict that the global GPU market size will rise to USD 200 billion to USD 400 billion by 2027. Although the total market capitalization of the AI + Crypto track is still in its early stage, its potential total addressable market (TAM) is enormous. We believe that even if decentralized networks capture only a very small portion (e.g., 1–5%) of these two massive markets, they will create an emerging market worth hundreds of billions of dollars. Its growth potential depends on the speed of underlying technological maturation and the emergence of killer applications.

Investment Logic and Thesis

Our investment logic for the AI + Crypto track is built upon the following core pillars:

• Macro Driving Force: The structural contradiction between the explosive growth in demand for AI compute power and the supply bottleneck of centralized cloud service providers provides a historic opportunity for decentralized compute networks. This is the most solid and clearest fundamental driver of the track.

• Cross-domain Innovation Dividend: AI brings unprecedented intelligence and automation capabilities to the crypto world, capable of fostering more complex DeFi strategies, more adaptive DAO governance, and more immersive blockchain gaming experiences. Conversely, Crypto provides AI with an open, permissionless economic model and global collaborative networks, solving the problem of AI’s “production relations.”

• Attention from Institutional Capital: Top VCs such as a16z, Paradigm, and Multicoin Capital have already made deep investments in this field. Their research and capital injections provide strong endorsement and catalysts for the development of the track.

Valuation System and Key Metrics

Valuing AI + Crypto projects is a complex challenge, as traditional valuation models (such as P/E, DCF) are difficult to apply because they rely on stable cash flows and profit forecasts. Therefore, we need to build a multidimensional valuation framework by combining crypto-native metrics with new metrics specific to this track.Crypto-native Valuation Metrics:

• Protocol Revenue and Price-to-Sales (P/S) Ratio: Measures the network’s ability to capture value, serving as a core indicator for assessing the financial health of a project.

• Token Staking Ratio: Reflects token holders’ confidence in the long-term value of the network and their willingness to participate in network security.

• Network Value to Transactions (NVT) Ratio: Similar to the P/E ratio, used to evaluate the valuation level of network value relative to its on-chain transaction activity.

• Number of Active Addresses and Developer Activity: Measures the actual usage of the network and the health of its ecosystem.

AI + Crypto Specific Metrics:

• Compute Utilization Rate: For decentralized compute projects, this is key to measuring the efficiency of their supply-demand matching and the real business demand.

• Model API Calls: For decentralized AI model platforms, the volume of API calls directly reflects the frequency and value of their models’ actual use.

• Data Contribution Volume and Value: For decentralized data projects, measures the scale and quality of data collected and processed by their networks.

• Agent Interaction Frequency and Value: For AI Agent platforms, measures the activity of agents within the network, the complexity of the tasks they perform, and the amount of funds mobilized.

Risks, Challenges & Strategies

Technical Risks and Bottlenecks Although the prospects are enticing, AI + Crypto still faces numerous severe challenges and bottlenecks on the technical level, which are the main obstacles preventing its large-scale application in the short term.

• Scalability and Performance: The low TPS (transactions per second) of most public blockchains cannot meet the demands of high-frequency AI applications. At the same time, communication latency between nodes in decentralized networks is a huge barrier to achieving efficient distributed AI training (which requires large-scale data synchronization).

• Model Complexity and Cost: The cost of directly deploying and running large language models (LLMs) on-chain is astronomical. Whether in terms of storage costs or compute costs, it is unbearable. Current on-chain AI mostly executes inference results rather than running the models themselves.

• Contextual Understanding Ability of Agents: Current AI agents have limited capability in understanding complex, dynamic, and adversarial on-chain environments. They can easily be exploited by malicious actors, or make wrong decisions when facing unforeseen market conditions.

• Maturity of Core Technologies: Privacy-preserving computation technologies such as ZKML (Zero-Knowledge Machine Learning) and FHE (Fully Homomorphic Encryption), which are the cornerstones of trusted computation, are still in very early research stages and remain far from large-scale commercialization.

Economic and Market Risks Apart from technical challenges, this track also faces significant risks in terms of economic models and market dynamics.

• Overhype and Bubbles: AI is currently the hottest narrative in the tech sector, which has caused the valuations of many projects in the AI + Crypto track to be seriously overestimated, leading to huge bubble risks. Market sentiment, rather than fundamentals, largely dominates short-term prices.

• Fragility of Token Economic Models: Designing a token economy model that can sustainably maintain supply-demand balance and incentivize all participants (compute providers, model developers, users, etc.) is extremely complex. A slight misstep may plunge the system into a “death spiral.”

• Competition with Traditional Cloud Services: In terms of reliability, stability, ease of use, and comprehensive service costs, decentralized compute networks are difficult to compete head-on with centralized giants such as AWS and Azure in the short term. They must find differentiated market positioning to survive.

Security and Ethical Risks The combination of AI and Crypto also brings new security and ethical dilemmas.

• Vulnerabilities in Smart Contracts and AI Agents: The complexity of code increases exponentially, and even a tiny vulnerability can be exploited, causing large-scale user asset losses. The autonomous execution of AI agents may amplify such risks.

• Liability Issues of AI: When an autonomous AI agent makes a wrong decision and causes massive losses, who should bear the legal responsibility? The agent’s owner, the developer, or the protocol itself? This is an unresolved legal and ethical question.

• Data Privacy and Misuse: Although decentralization aims to protect data, if improperly designed, publicly available on-chain data may also be used to train biased AI models, or exploited for malicious purposes, such as precision user behavior analysis for exploitation.

Regulatory Uncertainty Regulation is the greatest external uncertainty shared by both the AI and Crypto fields.

• Complexity of Cross-Border Regulation: The ethical and data usage norms of AI, as well as the financial attributes of Crypto, all face inconsistent and rapidly changing regulatory policies across countries. How a cross-border decentralized network adapts to such a complex regulatory environment is a major challenge.

• Contradiction Between Compliance and Decentralization: Regulators usually require the implementation of measures such as Anti-Money Laundering (AML) and Know Your Customer (KYC), but these are in fundamental conflict with the core ideals of permissionlessness and censorship resistance pursued by Crypto. Project teams need to find a difficult balance between compliance and decentralization.

Long-term Value and Historic Opportunity

The integration of AI and Crypto is far from a temporary technological hotspot, but a profound paradigm revolution aimed at reconstructing the creation, distribution, and ownership of value in the digital world. It combines the unparalleled intelligence and analytical capability of AI with the decentralized trust, coordination mechanisms, and programmable value networks provided by Crypto. The long-term value of this integration lies in the fact that it is not merely a stacking of technologies, but a co-evolution of productivity (AI) and production relations (Crypto), laying the foundation for building a more equitable, open, and autonomous digital economy. We are standing at a historic turning point in the transition from an “internet of information” to an “internet of value and intelligence.”

The Next 3–5 Years: Technological Breakthroughs and Market Inflection Points In the mid-cycle of the next 3–5 years, we predict that the following areas will become key technological breakthroughs and trigger market growth:

• Practicalization of Verifiable Compute: Zero-Knowledge Machine Learning (ZKML) will move from theory to practice, taking the lead in achieving commercial adoption in specific scenarios with high requirements for privacy and verifiability, such as financial risk control, decentralized identity (DID), and on-chain voting. This will be the cornerstone for the establishment of the “trustworthy AI” concept and a key factor for attracting institutional-grade applications.

• Combination and Emergence of AI Agents: As standards for agents mature (such as the protocols proposed by Olas), we will see large-scale collaboration among AI agents across protocols and across chains, forming a complex “agent society.” The market inflection point will be the emergence of the first “intent-driven” killer application, which can simplify complex on-chain operations into natural language commands, thereby fundamentally transforming the user experience of Web3.

• Rise of Specialized DePIN Networks: Decentralized Physical Infrastructure Networks (DePIN) targeting AI training and inference will become more specialized and efficient. By optimizing network protocols and hardware clustering solutions, they will demonstrate cost-effectiveness in specific niche markets (such as model fine-tuning and domain-specific inference) comparable to centralized cloud service providers, thereby capturing a considerable market share.

Conclusion

The Beginning of a Structural Transformation In summary, the coupling of AI and Crypto is a long-cycle structural transformation. It addresses the increasingly prominent problems in AI development such as centralization, data privacy, and value distribution, while injecting unprecedented intelligence and autonomy into the world of Crypto. Although it still faces multiple challenges such as technical bottlenecks, market bubbles, and regulatory uncertainty, the complementarity of its underlying logic and its historic inevitability are already clear.

We firmly believe that this is not merely a new investment track, but also the blueprint for the next generation of the internet —— a value network that is autonomous, intelligent, and fair. On this path, AI agents will gradually become the main economic participants, data and models will return to the hands of creators, and a truly “decentralized intelligent society” is slowly rising on the horizon. For today’s builders and investors, now is the best time to devote themselves to this grand experiment and to jointly shape the future digital landscape.