A Deep Dive Into The Crypto x AI Ecosystem
Exploring the Diverse Subsections in the Crypto x AI Ecosystem: Compute, AI Agents, Co-processors, & more.
Today’s post is sponsored by Nim Network.
Nim Network is building the future of AI gaming and consumer applications in crypto, creating in its chain an ecosystem that connects ownership with the funding of open-source AI and applications.
Introduction
The convergence of the crypto and AI ecosystems has been rapidly growing, with numerous companies developing innovative solutions to address various challenges within the industry. These efforts span across verticals such as data availability, coordination networks, compute providers, and model providers, essentially covering the entire AI stack.
Over the past year, the landscape has received substantial support from leading key opinion leaders, builders, and innovators in the space. This support has significantly contributed to the advancement and visibility of the crypto x AI ecosystem.
In this report, we aim to delve deeper into this ecosystem and provide a comprehensive understanding of its components. We will cover the following sections:
Ecosystem 101
Sub-categories deep dive
What would the future of Crypto x AI look like?
Closing thoughts
This is the second installment in our series exploring decentralized AI. If you’re completely new to the idea of decentralized AI, you can check out the first paper from this series, here.
Ecosystem 101
We’ve created the market map above to provide a quick overview of some of the major categories within the crypto x AI ecosystem.
The major categories we’ll explore today are:
Compute
AI Agents
Data Availability
Gaming
Privacy, ZKML, FHE
Consumer
Coordination Networks
Co-processors
Model Training
Model Creator
In the sections below, we will briefly explore each of these categories and the projects building solutions within them. We will also provide links to these projects for further exploration. A key focus will be on how modularity becomes a crucial component across the stack, particularly in data availability, AI agents, and coordination networks.
Each of these categories is a crucial component for creating a brighter and stronger future for decentralized AI.
Let’s get started.
Compute
Decentralized compute providers offer computational resources through distributed networks rather than centralized data centers. Currently, most compute resources are controlled by hyperscalers—centralized entities licensed by chip providers to rent out compute power. This centralized model often results in unused compute, causing users to pay more than necessary.
In contrast, decentralized compute platforms allow users to rent out their idle computational power, creating a marketplace for these resources. This approach can significantly reduce costs and increase efficiency by leveraging underutilized power from personal computers, servers, and other devices worldwide. Decentralized networks also enhance security and resilience against attacks or failures that might impact centralized services.
For AI applications, decentralized compute providers are particularly beneficial. AI model training and deployment require substantial computational power, which can be prohibitively expensive when sourced from traditional, centralized cloud providers. Decentralized networks like Akash Network and Render Network offer scalable, affordable solutions for these needs, supporting a wide range of computational tasks beyond AI, including scientific simulations and digital content rendering.
Decentralized compute networks are also more flexible and adaptable than traditional cloud services. They can dynamically allocate resources based on real-time demand, ensuring that users get the power they need when they need it. This flexibility, combined with cost savings and enhanced security, makes decentralized compute an attractive option for businesses and developers in the AI ecosystem and beyond.
Key Players in the Space:
Hyperbolic unites global compute to provide accessible, affordable, and scalable GPU resources and AI services. They offer GPU access, including A100s and H100s, at the lowest market prices and allow users to monetize idle machines. Hyperbolic serves companies, researchers, data centers, and individuals, providing high-throughput, low-latency AI inference services and scalable GPU access with pay-as-you-go plans.
Akash allows users to buy and sell computing resources securely and efficiently. Its permissionless, peer-to-peer communication model focuses on data privacy and payment transparency, making it a flexible, secure, and cost-effective alternative to traditional cloud services. They claim to be almost 5x cheaper than web2 counterparts. Users can explore a wide range of cloud resources and live network pricing, become providers by offering their hardware on the network to earn, and deploy with the user-friendly Akash Console. Akash is general-purpose, aiming to provide cloud compute services to anyone and everyone.
Aethir offers secure, cost-effective access to enterprise-grade GPUs globally. With over $400 million in compute capacity, Aethir focuses on enabling high performance and reliability. They have two major offerings:
Aethir Earth: Provides raw GPU computing power for AI model training and inference.
Aethir Atmosphere: Supports low-latency cloud gaming.
GPU providers can scale easily, earning substantial income and exclusive rewards. Aethir has a massive focus on gaming and AI.
Render Network provides decentralized GPU rendering, aiming to offer near-unlimited GPU computing power for 3D content creation. Launched in 2017, it’s one of the oldest players in the market and focuses on empowering creators and artists to work on content creation without worrying about compute requirements and capabilities. It’s a GPU provider itself, unlike Akash, which has a more community-driven focus.
IO.net is an aggregator provider focusing on global GPU resources, aiming to provide accessible, affordable, and scalable compute solutions. Users can monetize idle GPUs, earning revenues with high utilization rates. IO.net emphasizes robust security with SOC2/HIPAA compliance and end-to-end encryption. They have partnered internally with other compute providers like Aethir and Render, aggregating the compute services provided by these partners.
AI Agents
Decentralized AI agents are autonomous programs that operate within a distributed network, performing tasks and making decisions without centralized control. These agents interact with other agents and systems, creating complex, multi-agent environments for collaborative task execution.
The primary advantage of decentralized AI agents is their independence and ability to collaborate, enhancing robustness and scalability with no single point of failure. They can operate across different blockchain networks, interacting with smart contracts and other decentralized applications for seamless, integrated services.
Decentralized AI agents are valuable in scenarios requiring trust, security, and transparency. In financial services, they can autonomously manage and execute trades while ensuring regulatory compliance. In supply chain management, they can track and verify the movement of goods, providing real-time insights and enhancing transparency. Organizations leveraging decentralized AI agents can build more resilient, efficient, and secure systems at scale.
Key Platforms:
Talus Network is a Layer 1 blockchain that combines the security and performance of Move's smart contracts to create a robust ecosystem for AI smart agents. These agents can be used in various applications such as DeFi for monetizable smart agents, intent networks for optimal user outcomes, automated gaming resource collection, and DAO governance. The core principles of Talus are security, speed, and an enhanced developer experience, enabling the creation of secure, high-performance AI applications. This ensures that smart agents within Talus can be owned, managed, and monetized securely and transparently.
Guru Network is a Layer 3 blockchain that is building a multi-chain AI compute layer, allowing dApps and retail users to embed orchestrated AI agents into their routines and earn rewards. Guru Network's Flow Orchestrator serves as an infrastructure-as-a-service (IaaS), enabling AI models and processors to be published and integrated into applications. The network supports autonomous agents and compute nodes, creating marketplaces for these services. With a focus on interoperability and scalability, Guru Network aims to integrate AI-driven orchestration across both on-chain and off-chain activities.
Myshell is developing an AI consumer layer that connects users, creators, and open-source AI researchers. The platform allows users to build, share, and own AI agents, enabling voice and video interactions through AI companions like Shizuku. Leveraging state-of-the-art generative AI models, Myshell transforms ideas into AI-native apps quickly, empowering anyone to become a creator, own their work, and be rewarded for their contributions.
Data availability
Data availability in AI and blockchain refers to accessing and utilizing data stored across a distributed network, which is crucial for decentralized applications (dApps) and AI models. Platforms focus on securely storing and ensuring data is readily available when needed, employing techniques like sharding and cryptographic proofs.
Modularity is crucial in data availability (DA) as it allows components to scale independently to meet increasing demands. It separates data availability from consensus and other blockchain functions, enabling specialized optimization and integration with various applications. Modular systems can interact with multiple blockchain ecosystems, providing a versatile foundation for decentralized AI and dApps.
Reliability is critical for AI applications requiring large datasets for training and inference, even during network disruptions or attacks. These platforms distribute data across multiple nodes to promote transparency and trust, reducing the risk of manipulation or censorship. This reliability is especially vital in sectors like finance, healthcare, and governance, where data integrity and transparency is paramount.
Key Platforms:
Celestia is the first modular blockchain network designed for scalable and efficient data availability solutions for dApps. By separating consensus and data availability layers, Celestia enables developers to deploy customizable blockchains as easily as smart contracts. Its modular architecture supports abundant throughput through data availability sampling (DAS), which scales while maintaining verifiability for any user.
Eigen DA, built on EigenLayer, stores rollup transactions until their state is finalized on the rollup bridge. Its scalability, security, and decentralization make it ideal for developers needing reliable on-demand data. Eigen DA's core components include operators, dispersers, and retrievers, which work together to store and verify data efficiently.
0g Labs provides infinitely scalable data availability and storage systems to scale Web3 and enable novel on-chain use cases. Their programmable data availability infrastructure facilitates scalable and secure applications with low-latency data feeds. The 0G Storage Network offers a flexible data storage system for structured or unstructured data, supporting applications, network state-offloading, and more. This flexibility enables developers to customize data pipes, build on-chain AI applications, and perform decentralized inference or fine tuning using OPML or ZKML.
Nuffle Labs has two major offerings:
Near DA leverages NEAR Protocol’s sharded architecture to offer a modular data availability layer for rollups, ensuring high throughput and low costs.
The Nuffle Fast Finality Layer (NFFL) provides a fast settlement layer leveraging EigenLayer, enabling quick information access across participating networks.
Celestia, Eigen DA, 0g Labs, and Nuffle Labs support AI in the crypto space by providing infrastructure for storing and retrieving large datasets crucial for AI models. These data availability layers ensure secure and accessible data for AI model training and inference, fostering innovation in AI dApps.
Gaming
By leveraging decentralized networks and AI-driven processes, Web3 games and platforms create dynamic gaming environments that adapt and evolve based on player interactions. This approach enhances player engagement by providing unique, personalized experiences that are not possible with traditional, centralized game servers.
AI algorithms analyze player behavior and preferences to tailor gaming experiences based on each individual user: adjusting difficulty levels, suggesting in-game purchases, and generating custom content. This personalization enhances engagement by offering unique challenges and rewards based on individual preferences. Additionally, AI enables the creation of sophisticated non-player characters (NPCs) and opponents that can learn and adapt to player strategies, resulting in more challenging and unpredictable gameplay.
AI optimizes in-game economies by adjusting the supply and demand for virtual goods based on player activity. This maintains balance and fairness, ensuring a sustainable economic environment within the game.
Key Players:
Nim Network is a Dymension RollApp that focuses on the intersection of web3 gaming & AI. It leverages the Dymension modular framework, providing compatibility with both the Cosmos ecosystem and EVM chains, ensuring flexibility and scalability. AI agents on Nim Network act as intermediaries between users and blockchain applications, simplifying interactions and enhancing the user experience. Collaborations with platforms like Jokerace and Ocean Protocol, along with participation in the AI Gaming Coalition, underscore Nim Network's commitment to innovation and scalability in AI gaming.
Today the Game
Today the Game will allow players to create their dream island and nurture relationships with AI-powered residents. It’ll be interesting to see what they will build
AI Arena is an action game where AI characters learn behavior patterns and engage in battles. Players train their AI characters, influencing their strategies and observing their performance in combat, creating an immersive blend of AI and gaming.
Colony is an AI-powered web3 survival simulation game that features highly autonomous AI agents called “avatars”, who continuously learn from the world around them. Players guide and collaborate with these AI avatars, which possess a wide range of skills and capabilities, to navigate a future Earth populated by distinct colonies competing for survival. Colony's AI avatars have unique personalities and worldviews, drawing individual lessons and insights from their experiences. Additionally, these avatars can autonomously transact onchain via dedicated wallets they control, enabling them to trade with other in-game avatars.
PlayAI is a modular chain designed specifically for AI in gaming, it enables creators to deploy sophisticated gaming AI, allowing players to monetize their gameplay, and helping games enhance the overall user experience. PlayAI aggregates gameplay data from the gaming community, processes it through data nodes to create model datasets, and ensures the highest quality data for training AI models.
Privacy, ZKML, FHE
Privacy-preserving technologies like Zero-Knowledge Machine Learning (ZKML) and Fully Homomorphic Encryption (FHE) are crucial for ensuring data privacy and security in decentralized AI applications. These technologies enable computations on encrypted data without revealing the data itself, which is particularly important for sensitive industries such as finance and healthcare.
ZKML allows AI models to be trained and deployed without exposing the underlying data. By using zero-knowledge proofs, one party can prove to another that a statement is true without revealing any additional information. This ensures that AI models respect user privacy and comply with data protection regulations. ZKML also facilitates secure multi-party computations, where multiple parties can jointly compute a function over their inputs while keeping those inputs private. This capability enables AI models to be more widely used in sensitive areas without compromising data privacy.
FHE allows arbitrary computations to be performed on encrypted data, meaning sensitive data can remain encrypted at all times, even during processing. This is particularly valuable for cloud computing, where data security is a major concern. By using FHE, AI applications can process sensitive data without ever exposing it, protecting against data breaches and leaks. This enhances the trustworthiness of AI systems and enables their use in highly regulated industries, providing robust data security and privacy.
Key projects:
Fhenix facilitates the deployment of encrypted smart contracts, ensuring that sensitive data remains secure and private. The project’s roadmap includes various phases, such as the launch of the Helium Testnet, the Nitrogen Testnet v2, and the Gold mainnet.
Inco Network competes with Fhenix by focusing on building a modular, privacy-preserving machine learning ecosystem. Integrating privacy-preserving methods, Inco Network ensures secure handling of sensitive data in machine learning applications, reducing risks related to data breaches and unauthorized access.
Giza leverages zero-knowledge proofs to ensure data remains secure and private. They aim to streamline the process of building, managing, and hosting verifiable machine learning models, enabling developers to create trustworthy AI solutions
They provide:
python-supported workflows for easy integrations
an action-based SDK for creating actions in a privacy-first manner
They recently announced the launch of their AI agents support framework too.
Modulus Labs specializes in developing actionable AI solutions. By leveraging zero-knowledge cryptography, Modulus Labs ensures that AI results are verifiable and tamper-proof. This capability, known as "Accountable AI," allows smart contracts to access AI outputs without compromising trust. They have various integrations with ML libraries and platforms, providing a seamless development experience for creating verifiable AI models.
Bagel Network focuses on creating a credibly neutral, peer-to-peer machine-learning ecosystem. Designed for both human and artificial intelligence, Bagel Network enables the seamless, verifiable, and computable evolution from siloed networks into an integrated machine learning ecosystem. The platform supports autonomous AIs.
Consumer AI
The Consumer AI category focuses on delivering decentralized AI solutions directly to end-users. Platforms focus on providing user-friendly interfaces and applications that leverage decentralized AI and blockchain technology. These platforms aim to democratize access to AI, particularly in inferencing applications.
Decentralized consumer AI applications offer significant advantages in terms of security and privacy. Unlike centralized services that store user data on a single server, decentralized platforms distribute data across multiple nodes, reducing the risk of data breaches and unauthorized access. This is particularly important for applications that handle sensitive personal information.
Some players:
Gemz is a platform designed to enhance engagement and loyalty between creators and their communities. It allows creators to craft and deploy custom 3D interactive tokens, essentially NFTs, which can be used to reward community members and drive fan engagement. These tokens are unique, collectible, and provide a direct connection between creators and their fans. Gemz aims to ensure that interactions are secure, transparent, and verifiable, fostering a deeper sense of loyalty and community among fans.
Chain GPT provides multiple features and services including:
A smart contract generator and auditor
AI-driven market analysis
AI-powered trading assistant.
Offers daily crypto market updates
Users can create and deploy smart contracts, perform technical and chart analysis, and receive daily crypto market updates. The platform also offers an AI chatbot for answering blockchain and crypto-related questions.
Coordination Networks
Coordination networks in web3 are essential for enabling seamless interaction and collaboration among data providers, compute providers, model developers, and inference providers. These networks ensure that high-quality data is readily available for training, computational resources are optimally utilized, and AI models are efficiently developed and deployed.
Strong incentive mechanisms within these networks encourage active participation and collaboration, fostering an open and inclusive environment for AI development.
Key Players:
Allora Network is designed to enhance the intelligence and security of applications through a modular network of ML models. By integrating crowdsourced intelligence, federated learning, and zero-knowledge machine learning (zkML), Allora aims to create a more secure, efficient, and collaborative AI ecosystem. Its modular architecture allows continuous evolution and improvement, fostering a collaborative environment where builders can share knowledge to drive innovation. The network supports various applications, including smart contracts and decentralized applications (dApps). Incentives are provided through its native token - $ALLO
Bittensor is a decentralized coordination network that incentivizes the sharing and collaboration of AI models. It’s open-source first and ensures that all transactions and contributions are transparent and verifiable, fostering trust and encouraging innovation within the community. There are ‘subnets’ - you can think of them as targeted models aimed at serving specific use cases such as data training, models for healthcare, or data scrapping services.
Co-processors
Decentralized co-processors provide specialized processing power for web3 apps by offloading specific tasks to specialized hardware. This distributed approach allows for more efficient and cost-effective processing, as tasks can be spread across a network of co-processors rather than relying on a single centralized system.
Co-processors are particularly beneficial for AI applications, enabling high-performance computing for tasks such as model training and inference. By leveraging trusted execution environments, decentralized co-processors ensure the confidentiality and integrity of computations, protecting sensitive data during processing.
Key Players:
Ritual is pioneering a decentralized execution layer for AI, starting with Infernet, a decentralized oracle network (DON) that enables smart contracts across any blockchain to access AI models. The next phase will introduce a sovereign chain with a custom VM optimized for AI-native operations, leveraging Celestia’s modular architecture for enhanced scalability and verifiability. Infernet allows users to access AI models on and off-chain, providing flexibility for various use cases.
Phala Network is building AI-agents / co-processors. This innovative framework facilitates autonomous AI agents to execute tasks, manage assets, and interact with both humans and other agents. Central to Phala's offering are its five critical features that collectively foster a robust AI-Agent ecosystem.
Decentralized network with over 30,000 nodes.
Smart contract integration with large language models.
Support for advanced models like GPT-4.
Autonomous collaboration between AI agents.
Data privacy ensured by Trusted Execution Environments.
Model Training
Decentralized model training platforms like Gensyn provide a distributed network for training AI models. Traditional AI model training requires substantial computational resources, which can be costly and time-consuming. Gensyn addresses this challenge by leveraging distributed computing resources to reduce costs and increase the speed of training large AI models. By distributing the training process across multiple nodes, Gensyn enables more efficient use of computational power and reduces the time required to train complex models.
One of the key advantages of decentralized model training is its ability to democratize access to AI capabilities. Smaller organizations and individual developers can leverage the distributed network to train their models without the need for expensive infrastructure. This opens up new opportunities for innovation and development, as more entities can participate in the AI ecosystem. Additionally, decentralized model training enhances the resilience and security of the training process, as tasks are distributed across multiple nodes rather than relying on a single, centralized system.
Furthermore, decentralized model training allows for more flexible and scalable processes. Developers can dynamically allocate resources based on real-time demand, ensuring that their models receive the power they need when they need it. This flexibility, combined with cost savings and enhanced security, makes decentralized model training an attractive option for AI researchers and developers. By leveraging platforms like Gensyn, organizations can accelerate their AI development and bring innovative solutions to market more quickly.
Key Player:
Gensyn specializes in providing decentralized solutions for AI model training by creating a marketplace where computational resources can be efficiently allocated to train AI models. This decentralized approach not only reduces costs but also increases the availability of computational power for AI researchers and developers. Gensyn's platform allows users to contribute their idle computing resources to the network and earn rewards based on their contributions. This creates a more accessible and scalable environment for AI model training, enabling researchers and developers to access the computational power they need without significant upfront investments. The use of blockchain ensures that all transactions are secure and transparent, fostering trust and collaboration within the community.
Model creation
The Model Creator category encompasses platforms that facilitate the creation and deployment of AI models. Platforms like Nous Research provide decentralized tools and frameworks for developing AI models, enabling researchers and developers to collaborate and share their work in a secure and transparent environment. This approach aims to accelerate AI innovation by fostering a community-driven model development process.
Nous Research focuses on developing advanced tools and frameworks for creating AI models - focusing on pushing decentralised open-source. Their AI pipelines can
run offline on edge devices
remain customizable due to open weights
are capable of generating synthetic data for production use
Future of Crypto x AI
The intersection of cryptocurrency and AI is still in its infancy, with many projects in various stages of development. The excitement is evident, signaling strong potential.
Collaboration among stakeholders will be pivotal. No single solution can address all challenges, making cooperation essential for developing orchestration layers, customizable solutions, and innovations tailored to specific use cases.
Components will need to become more modular, enabling a "plug-and-play" environment. This modularity will simplify integration and encourage open-source contributions. By designing components that can be easily swapped or added, developers can build complex systems more efficiently. Modularity will also:
Facilitate Innovation: Developers can experiment with new ideas without overhauling entire systems
Enhance Flexibility: Users can customize their solutions to fit specific needs, improving adaptability across different applications.
Promote Interoperability: Standardized interfaces will allow components from different projects across the stack to work together seamlessly
Reducing dependency between components is crucial. Establishing baseline solutions that can be easily adapted for diverse use cases will foster innovation and accelerate development across the ecosystem.
There are a few other upcoming projects like Mira who are building methods that help combine aggregate compute, data, and models (anything across the AI stack ) to perform specific tasks. Think of these as modular ‘plug-and-play’ blocks which can be used by projects to quickly build on top of.
Overall, the Crypto x AI ecosystem holds great promise. As more people recognize the value of decentralized systems, Crypto x AI will likely become a key area of development over the next 12 months.
Closing thoughts
That concludes our exploration of the broad categories within Crypto x AI. To recap, the top modular AI categories highlight:
Compute: Dominated by market cap and immediate demand.
AI Agents: Driven by innovations and development.
Data Availability: Enhanced by modularity support and versatility.
Privacy: Progressing through dedicated research.
Gaming: Advanced by novel AI-led gamifications.
One of the key opportunities is to be early to projects and monitor the broader Crypto x AI segment's performance, influenced by catalysts like NVIDIA’s revenue numbers and support from industry leaders like Balaji and Erik Voorhees.
I believe success in this space hinges on how effectively projects can:
Coordinate with each other
Integrate seamlessly
Create strong incentivization loops
For this, it’ll be important to focus on a ‘modular-first’ approach!
In future editions, we will delve deeper into these categories and explore the exact workings of various products.
Thank you for reading and following along! 👋