Repost the original title: MT Capital Research: The Intersection of AI x Crypto — Opportunities and Challenges

TL; DR

1. We believe that the development in the AI x Crypto track is sustainable, not just a temporary hype. With the advancement of AI technology over time, we anticipate seeing more funds and attention continuously flowing into this area, bringing multiple rounds of development opportunities. Therefore, laying out a strategy for the AI x Crypto track is not only feasible but a necessary strategic choice.
2. In the AI x Crypto domain, we can identify several sub-fields, including AI Agent, decentralized computing, data, oracles, ZKML (Zero-Knowledge Machine Learning), FHEML (Fully Homomorphic Encryption Machine Learning), co-processors, Memes, Universal Basic Income, generative art platforms, and gaming applications. Among these, decentralized computing is particularly notable. Whether it’s GPU computing or algorithmic models, it represents a huge space for innovation, with an extreme demand for computational power. Computational power becomes a form of consensus, whose potential value can compete with the market cap ceiling of public blockchains. We are also optimistic about the early-stage yet potentially huge fields of ZKML, FHEML, and co-processors.
3. Considering the current market liquidity, project fundamentals, and community influence, Worldcoin, Arkham, Render Network, Arweave, Akash Network, Bittensor, and io.net are the leading projects we believe have a position of leadership and potential for growth.

Introduction

In the past few years, the AI x Crypto domain has experienced unprecedented development and transformation. This emerging field combines two of the most transformative technologies: blockchain and artificial intelligence, aiming to explore how decentralized approaches can empower AI applications, thereby enhancing transparency, security, and user control. With the rapid progress of artificial intelligence technology, especially the rise of generative AI, and the growing demand for decentralized solutions, AI x Crypto has become one of the most exciting innovation frontiers in the technology field.

A New Landscape of Assetization in the AI x Crypto Domain: The Innovative Path of Computational Power, Models, and Data

The most direct use case of Crypto is assetization, and in the AI x Crypto domain, “Computational power assetization,” “Model/Agent assetization,” and “Data assetization” represent three major scenarios.

In the assetization of computational power, there are two main directions: decentralized computing and decentralized inference by AI Agents. Decentralized computing focuses on using distributed networks for the training of AI models. AI Agents mainly utilize trained AI models for decentralized inference. These AI Agents can be deployed on decentralized networks to provide various intelligent services for users, such as automated trading, knowledge assistants, or security auditing.

However, from a technical perspective, the training of current large AI models involves massive data processing and demands high-speed communication bandwidth, which imposes significant requirements on hardware infrastructure. Training Transformer large models typically requires high-end CPUs like NVIDIA’s H100 or A100, NVIDIA’s NVLink technology for GPU connection, and professional fiber switches to achieve over 100Gbps network connections to support training across multiple data centers. These models contain tens to hundreds of billions of parameters, requiring powerful computational capabilities and memory to execute deep network algorithms. At the same time, to rapidly supply data for processing, high-speed storage and network bandwidth are necessary to reduce I/O bottlenecks. Parallel computing strategies, such as model parallelism and data parallelism, demand high-speed internal and external network bandwidth for effective synchronization among multiple GPUs. These requirements indeed present a significant challenge for decentralized AI training under current technological and cost conditions.

The AI inference executed by AI Agents, due to its lower demands on computational power and communication bandwidth, makes the adoption of decentralized methods more feasible and practical. This is also the reason why many projects related to computational power in the current market are more focused on inference rather than training. Despite this, considering cost-effectiveness and reliability, centralized solutions often still surpass decentralized ones at this stage.

The assetization of models/Agents is another important direction, especially under the push of large language models like GPT, becoming a significant trend. Users can interact with AI-based virtual characters. Transforming these AI Agents into NFTs, allowing users to buy, sell, collect, or trade them, similar to art transactions. However, projects in this direction often have a lower technical threshold, lack innovation, and have a lower degree of integration between AI and Crypto. Many projects simply convert AI models into NFTs without deeply considering the integration points between AI and Crypto, leading to homogeneous competition in the market. Moreover, Agents are mostly stored on cloud servers, with only the proof of ownership made into NFTs and placed on the blockchain, resulting in a shallow integration with Crypto.

Data assetization is also an important direction in the AI x Crypto track, focusing on using decentralized technology and incentive mechanisms to release and utilize a large amount of data resources, typically confined to private domains, including personal data and internal enterprise data, etc. Once these data are transformed into resources available for training or fine-tuning large models, it can significantly improve the professionalism and efficiency of AI models in different vertical fields. However, factors like the diversity, quality, application scenarios, and privacy protection of data add complexity to data assetization, making standardization a challenge. While non-standardizable data can be NFTized, this also highlights the difficulty of establishing a market with strong liquidity and ease of trade.

Decentralized data labeling, as a part of data assetization, through the “Label to Earn” model or crowdsourcing platforms, incentivizes community members to participate in data labeling, improving data usability and quality while reducing costs and time. This decentralized labor approach not only ensures the efficiency and quality of data labeling but also ensures that participants receive fair compensation, providing a new pathway for data assetization.

Source: MT Capital

From the above, it can be seen that the actual established scenarios in the AI x Crypto track are relatively limited at present, with most directions having low barriers to entry, and the recent market enthusiasm largely driven by capital operations and FOMO sentiment. The AI x Crypto track currently faces several core pain points:

1. Immature Business Models: AI x Crypto is at a very early stage, and many projects attempting to combine the two are not mature enough to fully leverage their respective advantages. With teams that have a deep understanding of both fields getting involved, it is expected that more solutions will be developed that showcase the power of AI technology while deeply integrating Crypto characteristics.
2. The Dual Challenge of Interdisciplinary Expertise and Practitioner Preferences: In AI x Crypto projects, teams often have a deep background in either the AI field or a profound understanding of Web3 and cryptocurrencies, but struggle to excel in both. This not only limits the ability for technological innovation and exploration of business models but also reflects the preference tendency of practitioners when choosing their field, i.e., excellent AI talents often hesitate to venture into the crypto industry. This lack of interdisciplinary expertise and the contradiction of practitioner preferences become major obstacles to innovation in the field. In the future, teams capable of working across boundaries and having insights into both AI and cryptographic technologies will become the key force in innovation and progress in this domain.
3. Internal Empowerment Technical Challenges: When Crypto tries to internally empower AI, such as through ZKML and FHEML, the main pain point is the poor scalability of these technologies, which limits their practical application. Similarly, when AI tries to internally empower Crypto, the challenge is not only the complex engineering problem of integrating AI into existing systems but also ensuring that this integration can work effectively without hindering system performance. These challenges reflect that in deeply integrating AI with Crypto, innovative technical solutions are needed, as well as overcoming the complexity and scalability issues when implementing these solutions.

Despite the current difficulties, we still believe that AI x Crypto is one of the most important tracks of this cycle. The combination of AI and Crypto not only shows strong technical potential and application prospects but also occupies a unique and important position in the current technology and investment fields:

1. The Revolutionary Status of AI in Technology: AI is widely regarded as a key force driving the next round of technological revolution. Compared to the previous cycle centered around concepts like the metaverse, which requires more practical application landings and faces challenges in user data verification, the enthusiasm for the metaverse concepts, as represented by companies like Roblox and Meta, has rapidly declined following their stock price crashes. In contrast, high-tech companies like OpenAI that are not yet public do not need to prove their value through revenue at this stage. Compared to the metaverse, AI has a broader impact on practical applications and technological innovation. It penetrates various fields such as healthcare, education, transportation, and security, capable of advancing the entire high-tech industry chain. Decentralized computational power further unleashes the potential of AI by providing the necessary computational resources through distributed networks to support the training and inference of AI models, promoting the advancement and widespread application of AI technology.
2. The Importance of Computational Power: In AI x Crypto projects, the importance of computational power is self-evident. Computational power is not only directly related to the efficiency and effectiveness of AI model training but also an important indicator of a project’s technical strength and market consensus. The higher the computational power, the stronger the consensus, and the higher the market value. As more enterprises and individuals participate in contributing to decentralized computational power, it not only achieves the optimization of resource allocation but also promotes the exploration of new economic models and methods of value distribution, such as through computational power mining and AI computational power hosting.

Representative Projects

Worldcoin

The reason behind WLD’s recent stellar performance is quite simple. On February 15, OpenAI released a large-scale video generation model called Sora. With text instructions, Sora can generate up to 60 seconds of high-definition videos that include highly realistic backgrounds, complex multi-angle shots, and emotionally rich multi-character narratives, demonstrating a profound understanding of the physical common sense of the real world. Despite the anticipation for the release of GPT-5, the impact of Sora is comparable to a GPT-5 launch.

This event reignited enthusiasm for the AI field. It’s well-known that Sam Altman, the founder of Worldcoin, is also the CEO of OpenAI. With the operation of market makers, WLD quickly became the most eye-catching focus in the market at the beginning of the year.

Worldcoin mainly involves two areas: identity verification and the issuance of digital currency. Rumors have it that OpenAI is developing two types of agent robots capable of deeply understanding human instructions and acting upon them, which is seen as the final step toward General Artificial Intelligence (AGI). Once this step is reached, almost all jobs could be replaced, leaving the majority of people facing unemployment, but they cannot starve. At this time, OpenAI would need to issue a Basic Income (UBI) through Worldcoin, where individuals can collect 6 WLD per month just by iris recognition.

However, a detailed analysis reveals that WLD does not have substantial empowerment, and its existence is more as a hyped-up “air coin”. If WLD is truly used to distribute basic income in the future, this form of non-stablecoin could cause various problems. This is why the white paper and founders of Worldcoin are ambiguous when discussing the role of WLD.

WLD is likely to remain a meme coin forever. Nevertheless, this does not mean WLD lacks investment value. Looking at the market cap, WLD shares similarities with DOGE. If Altman’s fame can surpass Musk’s, WLD might have a chance to reach DOGE’s market cap. However, its high unit price to some extent limits its potential as a top meme coin. If Worldcoin’s price were more accessible, it would undoubtedly greatly increase its appeal as a top meme coin. As a top figure in the AI field, Sam Altman’s every public statement related to AI or major events in the AI field can significantly impact Worldcoin’s market, adding to Worldcoin’s attractiveness and uncertainty as an investment target.

If there are future actions to split the coin, i.e., redefining Worldcoin’s market positioning with a lower unit price and higher circulation, such a strategy could trigger a rapid price increase.

Although the current market positioning and practical application of Worldcoin are somewhat ambiguous, making it viewed by some as a meme coin, Altman’s influence and the rapid development of the AI field provide Worldcoin with unique market dynamics. If reasonable market strategies, such as coin splits, are adopted in the future, Worldcoin has the potential to become a force that cannot be ignored in the market.

Source: https://foresightnews.pro/article/detail/53744

Arkham

Arkham, founded in 2020 and headquartered in the United States, is led by founder and CEO Miguel Morel, with a team that includes Operations Director Zachary Lerangis, Business Development Director Alexander Lerangis, and Institutional Relations Expert John Kottlowski. Arkham has secured over $12 million in funding, including a $2.5 million public round from Binance Labs. The founders are veterans in the crypto industry, having previously established Reserve, a stablecoin project designed for high-inflation economies, with investors including Peter Thiel, Sam Altman, Coinbase, and Digital Currency Group among others.

Binance announced on July 10, 2023, that Arkham’s token $ARKM would be listed on its Launchpad, marking the first time Binance introduced a tool-type product, sparking great interest.

Arkham is a platform that uses artificial intelligence algorithms to analyze blockchain data, capable of associating blockchain addresses with real-world entities, providing users with a complete behind-the-scenes behavior perspective. Arkham recently launched a blockchain intelligence trading platform, Arkham Intel Exchange, which allows users to request needed information through rewards, while information providers can earn rewards by supplying information. Arkham also offers powerful tools that allow users to search, filter, and sort any crypto transaction, revealing the entities and personal information behind market activities.

In addition to its listing on Binance, exchanges like Kraken, OKX, and Hotbit also support $ARKM trading.

Arkham introduced an “Intel-to-Earn” model, matching buyers and sellers on the blockchain to realize an intelligence economy. Its platform token $ARKM is used to pay for analysis platform fees, governance voting, and user incentives. The total supply of $ARKM is 1 billion, with a circulating supply of 150 million (15% of the total supply), and the test website has registered 200,000 users. After listing on exchanges, the trading volume is expected to reach the scale of $100 million.

Arkham mainly includes blockchain analysis tools and an intelligence trading marketplace. Analysis tools provide comprehensive data insights through entity pages, token pages, network mapping, etc. Arkham uses its proprietary AI engine Ultra to de-anonymize blockchain data and match addresses with real-world entities through algorithms. The intelligence trading marketplace allows users to buy and sell information through rewards, auctions, and data sharing. Arkham maintains long-term platform operation by charging fees — for listing and auction payments, a 2.5% minting fee is taken, and for reward payments and successful auctions, a 5% acceptance fee is charged.

Compared to other data analysis platforms on the market, Arkham has several unique advantages, such as creating token use cases, realizing on-chain data value transactions through the intelligence exchange, and providing data analysts with a channel for monetizing their knowledge. The platform self-motivates through mechanisms like taking cuts, which is beneficial for sustainable development; it offers users the ability to track the archives of their historical investment portfolios and lowers research costs with visualized data graphs. However, Arkham faces challenges such as limited support for public chains, functional gaps with platforms like Nansen, limited replicability of token scenarios, a user base primarily of professionals limiting appeal to ordinary investors, weaker data processing capabilities reliant on external data teams, and more.

The Arkham project has a first-mover advantage and broad market space in blockchain information analysis but is still in the early stages with a business model that needs validation. Ecosystem construction and scaling require time to cultivate. Risks include the time needed for the popularization of on-chain information analysis, high user education costs, limited replicability of the business model, reliance on personnel for information processing, high operational costs and risks, variable information quality, reputation risks, and regulatory policy uncertainty.

https://foresightnews.pro/article/detail/48222

Render Network

Since its launch in April 2020, Render Network has become a leading decentralized rendering platform, bridging users in need of GPU computational power with providers who have surplus computing resources. This platform primarily serves high-demand computing fields such as artificial intelligence, virtual reality, and multimedia content creation, offering a fair and competitive market environment for all parties involved through its unique dynamic pricing strategy, which takes into account the complexity of tasks, urgency, and available resources. In this way, GPU owners can connect their devices to the Render Network and accept and complete rendering tasks using the OctaneRender software developed by OTOY. In exchange, users pay individuals who complete rendering tasks with RNDR tokens, while OTOY takes a small percentage of RNDR as a fee to facilitate transactions and network operations.

Render Network is headquartered in the United States and was founded by Jules Urbach. Not only is Urbach the founder of Render Network, but he is also the founder and CEO of OTOY, contributing deep insights and advancements in 3D rendering technology and decentralized computing platforms.

Render Network has completed several rounds of funding, including strategic financing. On December 21, 2021, Render Network successfully raised $30 million in a strategic financing round, with investors including well-known investment institutions and individuals such as Multicoin Capital, Alameda Research, Sfermion, Solana Ventures, Vinny Lingham, and Bill Lee. Additionally, Render Network raised $1.16 million through an ICO in January 2018. The successful fundraising not only supported the technological development and market expansion of Render Network but also reflected the market’s recognition of the potential for decentralized rendering services.

Utilizing the peer-to-peer network capabilities of RNDR tokens, Render Network can effectively distribute workloads among providers of idle GPU resources, while incentive mechanisms encourage nodes to share their unused computational capacity. This approach not only maximizes the efficiency of resource utilization but also creates value for participants, driving the prosperity of the decentralized rendering ecosystem.

In December 2023, Render achieved a significant technological leap by migrating its infrastructure from Ethereum to Solana. This transition brought new capabilities to Render, including real-time streaming, dynamic NFTs, and state compression, significantly improving the network’s performance and scalability, and opening up a wider and more diverse range of application scenarios for users.

DePIN (Decentralized Physical Infrastructure Network) emerges as a new concept, comprising two main domains: digital resource networks and physical resource networks, aiming to incentivize individual participation in the construction and efficient use of real-world infrastructure through a Physical Proof of Work (PoPW) mechanism. The advent of DePIN not only brings innovative solutions to the traditional Information and Communication Technology (ICT) industry but also heralds the arrival of a more decentralized and efficient infrastructure network model.

Despite the challenges of high barriers to entry and low resource utilization efficiency faced by the current ICT industry, DePIN introduces a peer-to-peer network model that enables the reuse of idle resources and, through decentralization, lowers the barriers to entry, enhancing market competitiveness and efficiency.

The successful upgrade of Render Network and its close integration with Solana demonstrate the advantages of decentralized rendering platforms in addressing real-time responsiveness and reducing transaction costs. This not only strengthens Render’s leadership position in the DePIN domain but also opens new paths for its future development.

As Render Network continues to advance technological innovation and ecosystem construction, its potential in leading-edge domains such as decentralized rendering, artificial intelligence, and digital rights management is gradually emerging. Render is more than just a rendering service platform; it is a powerful engine driving innovation, connecting resources with demand, and promoting decentralization and digital transformation. With ongoing technological advancements and growing market demand, Render Network is poised to become one of the key forces driving new developments in the digital economy.

Source: https://dune.com/lviswang/render-network-dollarrndr-mterics

Arweave

Arweave is an innovative decentralized data storage protocol designed for the permanent storage of data. Through its unique permaweb, Arweave enables stored data to be accessed in human-readable forms (e.g., via web browsers), thereby creating a persistent, immutable internet. This capability for permanent storage is revolutionary for ensuring the immutability and perpetual accessibility of information, particularly in applications that require high data integrity and permanence, such as legal document storage, academic research archives, and copyright protection.

Arweave incentivizes data storage providers within its network through its native token, AR, with an economic incentive mechanism that ensures the sustainability and scalability of the network’s storage capacity. As an infrastructure and storage network project, Arweave aims to reshape the way data is stored and accessed. Originally known as Archain, it was founded in 2017 and is headquartered in Germany. The founding team includes co-founder and CEO Sam Williams, COO Sebastian Campos Groth, and Legal Director Giti Said, who bring extensive experience in technology, operations, and legal fields, and are key to driving the development of the Arweave project.

Since the launch of its mainnet in June 2018, Arweave has attracted broad attention and support from several key investors, including notable entities like a16z Crypto, Coinbase Ventures, and Union Square Ventures. In May 2018, a public fundraising round raised $1.57 million. Subsequently, the project conducted two rounds of financing in November 2019 and March 2020, raising $5 million and $8.3 million, respectively, with investors including a16z Crypto, Multicoin Capital, Union Square Ventures, and Coinbase Ventures.

The AO (Actor Oriented) solution launched by Arweave represents a significant innovation in blockchain technology, primarily reflected in its provision of a hyper-parallel computing architecture. This architecture allows an arbitrary number of processes to run in parallel within a decentralized computing environment, greatly enhancing computational efficiency and scalability. The core features of AO include massive computational capacity, the realization of verifiable computing, and high parallel processing and scalability through the construction of three different subnetworks (messenger units, scheduling units, computing units) and basing on Arweave as the foundational layer.

Named AO (Actor Oriented), inspired by the Actor model in computer science, this model is particularly suited for designing and implementing systems that are highly concurrent, distributed, and fault-tolerant. Through AO, the Arweave team demonstrates their profound understanding of and innovative solutions for the future development of decentralized computing environments.

Source: https://foresightnews.pro/article/detail/54511

AO is built upon the foundational layer of Arweave, utilizing Arweave’s on-chain storage as its permanent host for operational data, enhancing its decentralized computing capabilities and allowing an arbitrary number of parallel processes to run simultaneously, akin to the collaborative functioning of data centers and internet computers. Additionally, a key part of AO is AOS, a specific operating system based on the AO architecture, which allows developers to create applications using the Lua language, further enhancing its usability and flexibility.

The launch of AO aligns with Arweave’s long-term goal of supporting a highly scalable blockchain network through its data storage platform. Although the Arweave team faced challenges in achieving this goal, their persistence and innovation ultimately made AO possible. This not only enhanced the functionality of the Arweave chain, enabling it to support more smart contracts and blockchain protocols, but also provided a new and powerful solution for decentralized computing.

The working principle of Arweave AO breaks through the limitations of traditional blockchain technology by decomposing the three main components of the blockchain into independent units that can communicate with each other and execute a large volume of transactions simultaneously, achieving unprecedented scalability. This innovation opens up new possibilities for the development of Arweave itself, as well as providing new perspectives and inspiration for the entire blockchain and decentralized technology field.

Ultimately, Arweave’s goal is to make AO a stable system that requires only infrequent updates, similar to Bitcoin, ensuring the continuity of core functionalities and user rights. This stability and transparency are crucial for users as it allows them to have deeper trust and understanding of the protocols they use. As Arweave AO continues to evolve and improve, it has the potential to become a significant player in the decentralized smart contract platform space, posing a strong competition to existing blockchain technologies like Ethereum.

Akash Network

The core value of Akash Network lies in its role as a decentralized computing platform that taps into the globally underutilized GPU resources, connecting these resources with users in need of GPU computational power. This platform not only offers a profitable opportunity for GPU resource owners but also provides a more cost-effective option for users requiring these resources. According to data from September 2023, Akash Network has successfully deployed between 150 to 200 GPUs on its network, achieving a utilization rate of 50% to 70%. This achievement translates to an annual transaction value of $500,000 to $1,000,000, demonstrating the market potential of the decentralized computing resource sharing model.

Further analysis of Akash Network’s business model draws a fitting parallel to Airbnb in the real estate market. Akash has created a marketplace that allows GPU resource owners to rent out their unused computing power, similar to how Airbnb hosts rent out their properties, while users needing these resources can access the required computing power at lower costs. This model not only increases the utilization rate of GPU resources but also lowers the barrier to entry into fields like artificial intelligence and machine learning.

With the rapid development of artificial intelligence, the demand for high-performance computing resources like GPUs has surged dramatically. Nvidia, a leading manufacturer of GPUs, is expected to see its revenue grow significantly in the coming years, from $27 billion in 2022 to $60 billion in 2023, with an anticipated reach of around $100 billion by 2025. This growth forecast reflects the robust demand for GPU computational power globally, providing Akash Network with a vast market space.

Akash Network’s decentralized model is particularly suited to the current market environment, where demand for cloud computing services is increasing, and a significant amount of global GPU computational capacity remains underutilized. Through Akash, providers can offer idle GPU resources, while consumers can access the necessary computational power at a lower cost. This model not only optimizes the distribution of resources but also democratizes computational power, enabling more businesses and individuals to participate in the research and development of artificial intelligence and high-performance computing.

The native token of Akash Network is AKT, which plays several important roles within the network. First, AKT is used to pay for computational resources on the network, including but not limited to GPU computation, storage, and bandwidth. Second, AKT is also part of the network governance, where holders can participate in the decision-making process through token voting, such as protocol updates and improvement proposals. Additionally, AKT serves as an incentive mechanism, encouraging users to participate in network maintenance, including providing computational resources and validating transactions.

To encourage more users to offer unused computational resources, Akash has designed an incentive mechanism primarily through two methods: token rewards and transaction fees.

• Token Rewards: Akash Network provides rewards to users who offer computational resources through the issuance of new tokens. These newly issued tokens are distributed to resource providers as incentives, encouraging them to connect more resources to the Akash Network. Furthermore, network validators and users participating in network governance can also receive AKT token rewards, motivating them to contribute to network security and governance.
• Transaction Fees: Akash Network charges fees for transactions using its services, paid in AKT tokens. According to Akash’s policy, a portion of the transaction fees is allocated to nodes providing computational resources, serving as a direct economic incentive for their service provision.

Akash charges a 4% fee for transactions paid with AKT, while transactions paid with USDC (a stablecoin) incur a higher fee of 20%. This differentiated fee structure aims to promote the circulation and use of AKT tokens, while also providing financial support for the network’s maintenance and development.

Akash Network has also established a community pool, collecting a portion of the network’s revenue, including tokens generated from inflation and transaction fees. The funds in the community pool are used to finance projects and proposals for network development, such as technical improvements and marketing campaigns, with the allocation of funds decided by community voting.

Through this complex but effective token model and incentive mechanism, Akash Network not only ensures the network’s active and healthy development but also offers users the opportunity to participate in the network and benefit from it. These incentive measures help attract more resource providers and users to join the Akash ecosystem, driving the long-term success and continued growth of the decentralized computing platform.

However, despite the broad market prospects for Akash Network, the challenges it faces cannot be overlooked. In addition to competing with traditional cloud service providers, Akash must continually optimize its technical platform to ensure efficient and secure services. Moreover, building and maintaining a decentralized market requires continuously attracting new resource providers and users and maintaining high market activity.

Source: https://www.modularcapital.xyz/writing/akash

Bittensor

Bittensor was founded in 2019 by AI researchers Ala Shaabana and Jacob Steeves, originally conceived as a parachain for Polkadot. In March 2023, the project strategically pivoted to develop its own blockchain, aimed at incentivizing global machine learning nodes with cryptocurrency to promote the decentralization of AI development. Bittensor introduced a new paradigm by enabling these nodes to collaboratively train and learn, enhancing the network’s collective intelligence through the integration of incremental resources and expanding the contributions of individual researchers and models to the whole.

Bittensor introduced several innovative concepts and mechanisms, such as the distributed Expert Model (MoE) and Proof of Intelligence, designed to promote the development of a decentralized AI ecosystem by rewarding useful machine learning models and outcomes. Its token economics and ecosystem structure are aimed at supporting and rewarding participants in the network, encouraging fair distribution practices and network participation through the TAO token.

The architectural design of Bittensor reflects its pursuit of establishing a robust AI ecosystem. Through a layered structure consisting of miners, validators, enterprises, and consumers, Bittensor aims to build a network that comprehensively supports AI innovation. In this structure, the miner layer drives innovation with AI models, the validator layer maintains network security and integrity, and the enterprise and consumer layers ensure that technological achievements are transformed into practical applications to meet market and societal needs.

The core participants of the Bittensor network include miners and validators. Miners submit pretrained models in exchange for rewards, while validators are responsible for confirming the validity of model outputs. Bittensor creates a positive feedback loop through incentive mechanisms, encouraging competition among miners, which promotes the refinement and performance improvement of models.

Although Bittensor itself does not directly participate in model training, its network provides a platform that allows miners to upload and fine-tune their models. This approach enables Bittensor to integrate various models, processing different tasks through specific subnetworks, such as text generation and image generation, among others.

Source: https://futureproofmarketer.com/blog/what-is-bittensor-tao

The subnetwork model adopted by Bittensor is a significant feature of its architecture, with these subnetworks focusing on executing specific tasks. Through this approach, Bittensor aims to achieve model composition and decentralized intelligence, although this goal still faces challenges due to current technological and theoretical limitations.

Bittensor’s token economy model is heavily influenced by Bitcoin, featuring a similar token issuance mechanism and incentive structure. The TAO token is not only part of the network rewards but also key to accessing services on the Bittensor network. The project’s long-term goal is to democratize artificial intelligence technology by promoting the iteration and learning of models within the intelligent network in a decentralized manner.

Compared to traditional centralized AI models, Bittensor’s greatest advantage lies in promoting the openness and sharing of AI technology. This enables AI models and algorithms to be iterated and optimized within a broader community, accelerating technological progress. Furthermore, through its decentralized network structure, Bittensor hopes to reduce the application costs of AI technology, enabling more individuals and small businesses to participate in AI innovation.

io.net

io.net is an innovative decentralized GPU network designed to address the challenge of accessing computational resources in the machine learning (ML) field. The project integrates GPU resources from independent data centers, cryptocurrency miners, and participants in projects like Filecoin and Render to create a vast pool of computational power. The idea was conceived by founder Ahmad Shadid in 2020 while building a GPU computing network for the machine learning quantitative trading company Dark Tick, facing high costs and difficulties in accessing resources. Subsequently, the project gained broader attention and recognition at the Austin Solana Hacker House.

The main challenges io.net aims to solve include limited availability of computational resources, lack of options, and high costs. By aggregating underutilized GPU resources, io.net offers a distributed solution that enables machine learning teams to build and scale their model service workflows on a decentralized network. During this process, it leverages advanced distributed computing libraries, such as RAY, to support data and model parallel processing, optimizing task scheduling and hyperparameter tuning.

In terms of products, io.net provides a range of tools and services, including IO Cloud, IO Worker, and IO Explorer. IO Cloud is designed to deploy and manage decentralized GPU clusters, featuring seamless integration with IO-SDK and offering a comprehensive solution for scaling AI and Python applications. IO Worker offers a comprehensive user interface that allows users to efficiently manage their computational resource supply operations, including account management, real-time data display, and tracking of temperature and power consumption. Meanwhile, IO Explorer provides comprehensive visualization of network activity and key statistics, helping users better monitor and understand the network status.

To incentivize participation and balance supply and demand, io.net introduced the IO token, which functions to reward continued use by AI and ML deployment teams, price computation units for IO Workers, and participate in community governance. Additionally, considering the volatility of cryptocurrency prices, io.net specifically developed a stablecoin pegged to the US dollar, IOSD, to stabilize the payment system and incentive mechanisms.

Source: https://io.net/

io.net demonstrates strong innovation and market potential in both its technology and business model. Through collaboration with Filecoin, it is expected to further expand its capabilities in model storage and computational resources, providing robust support for the development and expansion of decentralized AI applications. By offering a cost-efficient, accessible, and user-friendly platform, io.net aims to become a strong competitor to traditional cloud service providers like AWS, driving innovation and progress across the entire AI field.

In terms of capital, io.net has successfully completed a Series A funding round, raising $30 million with a valuation of $1 billion. This funding round attracted participation from several renowned investment institutions, including Hack VC, Multicoin Capital, Delphi Digital, Animoca Brands, Solana Ventures, Aptos, OKX Ventures, and Amber Group. This series of investments reflects the market’s high recognition of io.net’s innovation capabilities and market potential in decentralized computing and artificial intelligence.

Conclusion

As AI and blockchain technology continue to evolve, the AI x Crypto domain has shown tremendous potential and opportunities, while also facing a series of challenges. By delving into the three core scenarios of “Computational power assetization,” “Model/Agent assetization,” and “Data assetization,” we can see the innovative paths and existing obstacles in this field. Decentralized computational power opens new possibilities for AI training and inference, despite the dependency on high-performance computing resources and communication bandwidth. The assetization of models and Agents through NFTs provides proof of ownership and enhances interactive experiences, but deeper technological integration is still needed. Data assetization unlocks the potential of private domain data, facing challenges in data standardization and market liquidity, yet also paving new paths for AI efficiency and specialization.

It is noteworthy that with the continuous development and iteration of AI technology, there will periodically be influxes of attention and capital into the AI x Crypto domain, bringing continuous waves of development to AI, rather than just single-stage opportunities. The enduring value and innovation potential of the AI x Crypto domain signify its role as a key track in the technology and investment fields.

Looking forward, the development of AI x Crypto will rely on technological innovation, interdisciplinary cooperation, and exploration of market demands. By breaking through technological limitations, deepening the integration of AI with blockchain, and developing practical application scenarios, this field is moving towards long-term development, offering safer, more transparent, and fairer AI services. In this process, the decentralized philosophy and technological practices will continue to drive the AI x Crypto domain towards more open, efficient, and innovative directions, ultimately achieving dual leaps in technological innovation and value creation. Thus, the AI x Crypto track in the current cycle represents an important opportunity not to be missed, signifying not only the forefront of technological innovation but also indicating significant trends in future technological advancements and investment directions.

References

1. https://foresightnews.pro/article/detail/54962
2. https://foresightnews.pro/article/detail/55156
3. https://foresightnews.pro/article/detail/54807
4. https://foresightnews.pro/article/detail/55053
5. https://foresightnews.pro/article/detail/55054
6. https://twitter.com/mo_baioumy/status/1760296558539501698
7. https://twitter.com/Wuhuoqiu/status/1755922300799693108
8. https://foresightnews.pro/article/detail/53518
9. https://foresightnews.pro/article/detail/53744
10. https://foresightnews.pro/article/detail/38689
11. https://foresightnews.pro/article/detail/37989
12. https://foresightnews.pro/article/detail/37907
13. https://foresightnews.pro/article/detail/37579
14. https://foresightnews.pro/article/detail/48222
15. https://foresightnews.pro/article/detail/38545
16. https://foresightnews.pro/article/detail/37458
17. https://www.modularcapital.xyz/writing/akash
18. https://foresightnews.pro/article/detail/48972
19. https://foresightnews.pro/article/detail/49581
20. https://foresightnews.pro/article/detail/53218
21. https://foresightnews.pro/article/detail/54511
22. https://foresightnews.pro/article/detail/54515
23. https://foresightnews.pro/article/detail/54819
24. https://foresightnews.pro/article/detail/21045
25. https://foresightnews.pro/article/detail/52521
26. https://foresightnews.pro/article/detail/47729
27. https://foresightnews.pro/article/detail/47532

Disclaimer：

1. This article is reprinted from [MT Capital], Forward the Original Title‘MT Capital Research: The Intersection of AI x Crypto — Opportunities and Challenges’, All copyrights belong to the original author [Xinwei, Ian]. 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.