Author: Trustless Labs
Summary
AI applications combined with blockchain technology are attracting more and more attention from the market. ChatGPT has gained hundreds of millions of users in a very short period of time. Nvidia’s stock price has increased eightfold in 2023, with its market value exceeding one trillion U.S. dollars. AI has become the absolute mainstream of mankind, and its narrative and capital have also begun to spill over into related fields. Combined with regional AI applications of blockchain technology are attracting more and more attention from the market.
As an important complement to AI, Crypto has very extensive integration and huge development potential, and is currently in its early stages. At this stage, many institutions and practitioners are still tentative about the combination of AI and web3, mainly starting from the most direct use case of Crypto, which is "assetization": assetization of computing power (computing power cloud, computing power market), assetization of models ( AI Agent), data capitalization (storage).
From the current point of view, the unique decentralization attribute of Crypto technology cannot improve efficiency and reduce costs in the AI training process. It requires a lot of construction. However, assetization reduces the friction of market transactions and introduces more previously unused features. computing power, which is profitable in the context of scarce computing power; the assetization of the model allows the community to own or use AI in a decentralized manner. The technology implementation of this track is relatively simple and has a low threshold, but it is democratized. The narrative of AI corresponds to the anxiety of AI centralization; decentralized data can indeed reduce costs to a certain extent, but even in the crypto context, its assetization is difficult and still requires a lot of business exploration.
Although the market does not fully agree on AI + Crypto, its track distribution is already clear. Trustless Labs has traversed more than 60 projects in the primary and secondary markets. This article will horizontally expand from 7 modules including computing cloud, computing market, model assetization and training, AI Agent, data assetization, ZKML, and AI applications to explore The future development of AI and cryptocurrency fields, and exploring investment opportunities.
Computing Cloud
With the overall growth of the artificial intelligence market represented by companies such as NVIDIA and the trend of GPU supply shortage, computing power cloud and related computing power market projects have become one of the first areas to benefit from the growth of the AI industry. Computing Cloud attempts to solve the shortage of GPU computing power by integrating the idle computing power of existing cloud service providers into a centralized computing network, reducing computing power costs compared to traditional cloud services. Each of these projects has its own focus, and most of them focus on GPU computing power. However, each project focuses on different areas and technology implementation methods. Some are based on container technology such as Akash Network, while others provide implementation at the virtual machine level.
For computing power cloud projects, the growth of users and business is the basis for its value growth. These projects are generally considered high-PE AI computing power companies because they do not rely on a decentralized narrative. Since the scarcity of GPU resources may be a long-term phenomenon, it becomes particularly important for junior market players to examine the commercial capabilities of these projects.
Akash Network
Akash Network is a decentralized cloud computing marketplace based on Cosmos SDK and Tendermint technology, dedicated to leveraging underutilized resources in data centers. Since its founding in 2015, Akash Network has become one of the earliest projects in the field, initially supporting only CPU computing resources. In September 2020, Akash launched the mainnet supporting CPU computing, and successfully launched the mainnet supporting GPU computing in September 2023, further expanding its service scope and performance. Akash is designed to tap into and capture the huge market of underutilized resources in data centers.
Akash Network uses a unique reverse auction mechanism that allows users to choose the supplier with the lowest bid to reduce cloud computing costs. Its average price is only about one-third of that of mainstream cloud service providers such as AWS and Azure. This pricing strategy not only saves users a lot of costs, but also increases the competitiveness and attractiveness of the market.
The underlying architecture of the project uses the Akash container platform built on k8s to orchestrate and manage containers to meet various cloud computing needs of users. This design allows users to flexibly deploy and manage containers, further improving resource usage efficiency and flexibility.
Akash Network successfully raised US$2 million in 2020, led by Cypher Capital, showing the market's recognition of its technology and business model. The founder of the project, Greg Osuri, has extensive experience in continuous entrepreneurship and cloud architecture experience since 2008. At the same time, Adam Bozanich serves as Co-founder & CTO.
Currently, $AKT has a market cap of $130 million and a future diluted value (FDV) of $220 million.
Ritual
Ritual is a cutting-edge project dedicated to integrating artificial intelligence (AI) with smart contract technology. It has established an open AI infrastructure network. The project is driven by an experienced team, including Eva Zhang, who has worked at top companies such as Socket, Apple, and Google and conducted machine learning research at Stanford University, as well as co-founders Niraj Pant and Akilesh Potti, who have Polychain backgrounds. Ritual's vision is to enable developers to seamlessly integrate AI into any on-chain application or protocol, including the ability to fine-tune, monetize and perform inference, aiming to facilitate the development of fully transparent DeFi, self-improving blockchains, autonomous agents and generated content, etc.
Although in its early stages, Project Ritual has already achieved a number of results, including the launch of its first phase product, Infernet. Infernet is a lightweight library that can introduce calculations to the chain, supports any EVM-compatible chain, and allows smart contracts to natively access AI models for various on-chain use cases and tasks. Infernet can not only optimize the decentralized oracle network, but also enable DApps to implement functions such as DAO proposal risk detection and NFT image generation through its SDK. Ritual also plans to launch its second phase product - Ritual Chain - in the coming months, which will be a sovereign chain with a custom virtual machine dedicated to serving more advanced AI-native applications.
Although Ritual has not yet launched a testnet, and there is no economic model, currency issuance and staking mechanism, the project's open source code, team background, technical roadmap, strategic direction and community-shared concepts all show its huge potential and industry attention. . In addition, Ritual has successfully completed US$25 million in financing, led by Archetype, with participation from Accomplice and Robot Ventures, showing the investment community's recognition and support for its technology and vision.
Render Network
Render Network is a decentralized GPU rendering platform designed to utilize globally distributed high-performance GPU nodes to provide artists and developers with on-demand scalable rendering capabilities. This platform not only focuses on the field of rendering, but also supports AI computing and promotes the development of next-generation rendering and AI technology. Jules Urbach, the founder of Render Network, serves as the strategic vision setter and chief architect of the technology roadmap of OTOY. He is a pioneer in the fields of computer graphics, streaming media and 3D rendering, with more than 25 years of industry experience. Since its establishment in 2002, OTOY has completed 4 rounds of financing by 2020, and made a US$50 million Series B investment in Light Field Labs in February 2023.
Since its first ICO in October 2017, Render Network has developed into an early but promising project. The project had a private sales period from January to May 2018. Currently, the market value of Render Network's native token $RNDR has reached US$4.3 billion, while its FDV is US$6.1 billion. It has been listed on the world's leading cryptocurrency exchanges Binance and OKX.
On November 2, 2023, the Render Foundation announced that the Render Network has successfully completed the core infrastructure upgrade from Ethereum to Solana. This transfer is designed to take advantage of Solana's high performance and low cost advantages to further promote the development of the platform. In the process, Render Network launched a new token, RENDER, while minting it on Solana, and the old token, RNDR, will still exist. During the upgrade process, 1 RNDR can be exchanged for 1 RENDER. In the future, the Render Foundation will focus on and fully support the RENDER token, marking an important milestone for the Render Network to enter a new stage.
NetMind.AI
NetMind.AI is a huge artificial intelligence ecosystem, consisting of four core modules: computing power market, ChatBot, Agent, and life assistant. Its core, NetMind. Power, is a computing power market module based on Ethereum, providing a GPU inference and training platform. NetMind.AI's training/inference load management platform has entered the testnet stage since September 2023. NetMind. Power supports a wide range of GPU models, integrates Google Colab's one-click import function, and has a good user experience. It is currently in the free beta testing phase.
The NetMind chain is fully compatible with EVM, and its token $NMT has a market capitalization of 444M and FDV of approximately 13B.
CUDOS
CUDOS is similar to Akash and is a delegated proof-of-stake (DPoS) blockchain using Tendermint Core on the Cosmos SDK, with cross-chain capabilities for Ethereum and L2. Cudos Blockchain Compute is designed to provide multi-chain EVM support and provide GPU computing capabilities.
$CUDOS has a market capitalization of 161M and an FDV of 275M.
Nuco.cloud
Nuco.cloud is a decentralized computing cloud service based on blockchain technology, aiming to provide diversified cloud computing solutions, including use cases such as artificial intelligence training, data analysis, scientific research, and efficient rendering. The project runs on the two major blockchain platforms of Ethereum and Telos, taking advantage of decentralization to provide users with flexible and efficient computing resources. By participating in Nuco.cloud, users can contribute their own computing resources in exchange for computing power, and miners who provide computing power can receive NUCO tokens as rewards. The token is generated based on a proof-of-stake (POS) and proof-of-research mechanism and is supported by Eagle Capital GmbH. It is worth mentioning that Nuco.cloud has established a cooperative relationship with CUDOS to jointly promote the development and application of blockchain technology.
In order to meet the needs of different users, Nuco. cloud has launched a variety of product lines, including nuco. cloud SKYNET, nuco. cloud GO, nuco. cloud PRO and CUSTOM, etc. Each product is targeted at specific user needs and usage scenarios. These products utilize advanced blockchain technology and intelligent pricing mechanisms to provide users with a unique, efficient and decentralized cloud computing service platform.
Since the establishment of the project in 2017, Nuco. cloud launched a test network in August 2023 and is still in the testing phase. In the 2024 roadmap, Nuco. cloud plans to rebrand and release a PRO version of the product for the B-side, and at the same time seeks to be listed on the CEX exchange to further expand its market influence.
In terms of compliance, Nuco.cloud complies with the EU's GDPR standards and German related standards, which makes it stand out among many competitors, including centralized computing cloud services such as AWS, Azure, Google Cloud, as well as GLM, Akash, RNDR , IExec RLC and other decentralized computing projects. According to the project information (Deck), the cost of Nuco.cloud is lower than that of several other decentralized computing clouds. Compared with AWS of the same specifications, its price is only about 30%, showing that it is extremely competitive. Strong price advantage.
Currently, the market value of Nuco.cloud’s token $NCDT has reached 44M, and the FDV is 68M.
Dynex
Dynex is a neuromorphic supercomputing blockchain based on the Dynex Solve chip algorithm, and proposes a Proof of Useful Work (PoUW) method to improve the speed and efficiency of decentralized networks. Dynex hopes to provide computing power for artificial intelligence, machine learning, financial technology, biomedicine, etc. on this basis. Mainly used is the computing power of GPU.
Dynex was launched in 2020. It was originally a project to design neuromorphic chips, and then evolved into a blockchain. The mainnet was launched in Q4 of 2022, and Gate.io was launched in Q1 of 24.
$DNX was launched fairly in August 2022, using a deflation model, with a market cap of approximately 86M and FDV of 110M.
OctaSpace
OctaSpace is an open source scalable distributed computing cloud node infrastructure that allows access to distributed computing, data storage, services, VPNs, and more. OctaSpace includes CPU and GPU computing power, serving disk space for ML tasks, AI tools, image processing, and rendering scenes using Blender.
OctaSpace launches in 2022 and runs on its own Layer 1 EVM-compatible blockchain. The blockchain uses a dual-chain system that combines Proof of Work (PoW) and Proof of Authority (PoA) consensus mechanisms. $OCTA FDV about 73M, fair start
AIOZ Network
AI OZ Network is a Layer 1 decentralized computing platform for AI, storage and streaming media, which can achieve interoperability with Ethereum and Cosmos. AI OZ is powered by a decentralized content delivery network (dCDN), AI computing and thousands of individual nodes operating globally, allowing users to share computing resources for storing, transcoding and streaming digital media content, and supports decentralized Centralized AI computing.
The project was launched more than 6 years ago and the mainnet was launched in December 2021. The infrastructure built includes AI OZ dCDN, AI OZ IPF S, AI OZ W3 S, AI OZ W3 AI, AI OZ Web3 streaming, which provides web3 storage, decentralized artificial intelligence computing, live broadcast and video on demand (VOD). support.
According to its Roadmap, the project Q1 will implement AI OZ Transfer dApp, AI OZ multi-chain wallet browser extension, AlOZ Node V3, AI OZ W3 IPFS infrastructure; Q2 implement AI OZ W3 AI; Q3 implement AI OZ W3 Stream, and AI OZ various token standards; Q4 realizes AI OZ DEX.
Phoenix
Phoenix is a blockchain infrastructure platform for decentralized AI, computing expansion and data-driven networks. Its products include AlphaNet, an AI platform for the cryptocurrency trading market, and NYBL, an AI GC metaverse project. Phoenix's AI computing layer is a Web3-based infrastructure designed for scaling AI computing tasks and easily deploying AI-enabled applications, including deep learning, predictive analytics, LLM, federated learning, and AI edge computing. Phoenix’s artificial intelligence vertical solutions cover multiple fields.
Recently, Phoenix has cooperated with Helium to launch the AI computing mining machine Phoenix Node, and its Phoenix LLM large language model service has been launched on Telegram. In addition, Phoenix will be integrated with the entire Telegram ecosystem.
Aethir
Aethir is focused on building scalable decentralized cloud infrastructure for games and artificial intelligence, based on Arbitrum. The project was launched in 2023 and is currently in the testnet stage. The user experience is close to that of web2's cloud game platform, and it meets the testing, cross-platform, and distribution needs of AI and games. The main use cases in artificial intelligence include delay-free reasoning, language interaction, and visual interaction. Checker node licenses will be available on the whitelist at 10:00 am (UTC) on March 18th and 19th. The public sale will start at 10:00 am (UTC) on March 20th. The starting price of each node is 0. 1259ETH. , increasing step by step, as ERC721 can also be purchased at the second level. The Checker node can be installed and run in 2024 Q2, and it is expected that the currency will be issued and the main network will be launched in Q2. 15% of $ATH will be used to reward Checker nodes.
Iagon
Iagon is a decentralized storage and computing market based on Cardano. The testnet will be released in 2023 and the project will be open source.
OpFlow
OpFlow is a decentralized computing power cloud platform focusing on AI and rendering. OpFlow Hosting is a decentralized deployment cloud service, compatible with TG Bot. Its rendering service exclusively uses NVIDIA L40 GPUs. According to its Roadmap, the first phase of the plan is to implement: release tokens on Uniswap; launch OpFlow hosting (Telegram Bot); launch network nodes such as ETH, Linea, and Eigen Layer.
$OpFlow is an ERC20 Token, more details have not been announced yet.
OpSec
OpSec is a decentralized cloud computing solution. Aiming to build the next generation of supercomputers, solutions include Cloudverse, Cloudsec, OpSec networks, one-click node setup, DePIN hardware, managed services, etc.
The project is relatively new and has not released a testnet yet. Currently, $OPSEC is online on DEX.
Hash rate market
The computing power market relies on a decentralized mechanism, allowing users to provide their own GPU and CPU resources to participate in computing power leasing or network training projects. Although the emergence of this kind of market has not theoretically significantly reduced the cost or improved efficiency of AI training, its openness and threshold-free characteristics enable a large number of GPU resources to be mobilized in a short period of time to provide a powerful network computing power support. This approach not only leverages huge asset endorsements, but also realizes the leveraged tokenization of GPU assets, bringing new value and significance to the computing power market.
However, such projects are often regarded as "more virtual than real". Their core does not lie in the actual utility of the technology or service, but relies more on its decentralized attributes and the creation of related narratives (for example, from traditional AI to The process from decentralized reasoning to decentralized training). In fact, the business models of these projects largely revolve around this narrative.
In terms of technical support, computing power market projects that enable decentralized operation of GPUs are usually combined with the DePin concept. The combination of the computing power market represented by io.net and nosana plus DePin, combined with Solana, a high-performance chain, shows significant growth potential. Therefore, we recommend that investors pay long-term attention to early-stage projects that combine Solana and AI. These projects are not only expected to bring technological innovation, but may also become investment hot spots.
From an investment perspective, early participation in the GPU computing power market, especially when the market shows a FOMO effect, may bring higher returns. This method of participation can not only obtain incentives, but is also expected to achieve a high input-output ratio and bring considerable economic benefits to investors.
Clore.ai
Clore.ai is a platform based on PoW that provides GPU computing power rental services. Users can rent out their own GPUs for tasks such as AI training, rendering, and mining. With an optional proof-of-hold (PoH) mechanism, the more tokens a provider holds, the more CLORE they receive.
$CLORE market capitalization is 90M, FDV is 110M. 50% of each block belongs to the miners, 40% to the custodian, and 10% to the team, with a cap of 1.3 billion tokens.
Nosana
Nosana is a cloud computing service provider focusing on GPU on Solana, and the project is open source . Nosana was established in 2021 and currently has no financing information. Co-founder Sjoerd Dijkstra has rich DevOps experience, Jesse Eisses has a good technical/ML background, and Laurens Verspeek also has a technical background and has rich development experience. In mid-October 2023, Nosana announced a shift from CI/CD use cases to AI inference. Its platform is currently in the early stages of development. There is very little data on nosana explorer . Current statistics show that there are 94 GPU nodes and 160,000 completed inferences. It needs Further research to validate the project.
Llama 2 and Stable Diffusion's AI inference workloads will be integrated into the network. The test network is divided into three phases. It is currently in the first phase. According to official Twitter data, more than 1,000 devices have signed up for the second phase. The data and revenue model have not been made public. It plans to launch the mainnet in the first half of the year and improve consumer node support; in the second half of the year, it will implement community connector libraries and official connectors for PyTorch, HuggingFace, TensorFlow, etc.
The currency will be issued in January 2022. The current market value of $NOS is 510M and FDV is 608M.
io.net
io.net is an innovative AI decentralized computing power market based on Solana blockchain technology and is currently in the testnet stage. This platform aims to provide users with access to cloud computing resources at a lower cost than traditional centralized services, and its valuation has reached US$500 million. The core attraction of io.net is that it can support various AI computing needs, such as batch inference, parallel training, hyperparameter tuning, reinforcement learning, etc., and its back-end infrastructure is composed of a series of modular layers. Able to achieve effective management and automated pricing of resources. Currently, it offers NVIDIA RTX 3090 computing resources priced at $0.20 per hour.
In terms of project background, io.net was founded by Ahmad Shadid, who built a GPU computing network for Dark Tick, a machine learning quantitative trading company, in 2020. As for the financing situation, io.net has been led by Hack VC, and has also received support from a number of well-known investment institutions and individuals, including Multicoin Capital, Solana Labs, Aptos Labs, etc., as well as some well-known figures in the industry such as the founder of Solana Anatoly Yakovenko.
In terms of participation opportunities, io.net provides diverse ways for users to participate in its network, including mining income and staking income. From March 1 to April 28, 2023, the platform launched the first round of points incentive program and expects to issue its token $IO in April. In terms of mining income, the platform calculates points by detecting the node’s Internet bandwidth, memory, CPU clock speed, and floating point computing power to provide rewards to users. In addition, users can also earn income by staking IO Coin, and stakers will receive 1-3% of all rewards received by participants.
It is worth mentioning that the io.net team drew on RNDR’s airdrop rules to design its own points calculation method, emphasizing the importance of bandwidth in calculating rewards. The analysis shows that high-bandwidth devices have a clear advantage in the final reward distribution, which prompts users to pay special attention to improving the bandwidth performance of their devices when participating in the io.net network.
Gensyn
Gensyn is an L1 protocol focused on training deep learning models. Its main goal is to improve the efficiency and accessibility of deep learning model training through innovative designs and technical solutions. Since the launch of the project, Gensyn has completed two rounds of financing, including a $6.5 million seed round led by EdenBlock in March 2022, and a $43 million Series A round led by a16z in June 2023.
The core of Gensyn's technology lies in its verification system and computing power supply solution, as well as a set of efficient solutions. Its verification system consists of four main roles: submitter, problem solver, verifier and whistler, which together ensure the trustless nature of the system, that is, a trustless distributed verification mechanism. In addition, Gensyn is committed to making it easy for various devices to access its computing network, including daily game-level GPUs and GPUs used for ETH PoW mining. Its efficient machine learning training solution is based on probabilistic learning proofs, graph-based precise positioning protocols, and Truebit-style incentive games, aiming to significantly reduce training costs and improve training efficiency.
In terms of cost, Gensyn has demonstrated significant competitiveness. The average hourly cost of its machine learning training is about US$0.4, which is not only lower than the US$1.5 one-hour GPU call cost provided by Akash, but also within the cost range of US$0.1-1 provided by io.net, which reflects Gensyn’s competitiveness. Advantages in cost control.
Gensyn's code has not yet been open sourced, and there are currently no plans to issue coins. The development of Gensyn will be divided into three stages: testnet, canarynet on the Kusama relay chain, and mainnet on the Polkadot relay chain. Currently, the project is still in the testnet stage. Over the next 3-6 months, Gensyn plans to continue optimizing its technology and prepare to enter the next stage of development. Its vision is to become the basic layer for machine learning computing, similar to Ethereum for smart contract execution, which demonstrates Gensyn's long-term plan for the future development of deep learning and blockchain integrated applications.
Nimble
The core vision of the Nimble project is to create a decentralized AI ecosystem that covers the three major markets of data, computing power and builders. Through this design, Nimble aims to lower the threshold for training AI models and enable more participants to contribute to the innovation and development of AI. This project is positioned as a Composable AI Project, which means that it emphasizes the composability of various AI resources and components under a decentralized framework.
The Nimble mainnet was launched on March 21, 2024, and was selected into the seventh season of BinanceLabs Most Valuable Builder (MVB) Incubator Program on March 1. The full security audit report has not yet been seen. Nimble is valued at $30 million and has raised $6 million.
The mainnet was launched last weekend, and users can mine by donating GPUs. Currently, 1,500 GPUs have been connected to the network.
Morpheus AI
Morpheus AI is a decentralized computing power marketplace. This project is based on Arbitrum and aims to help users build agent AI based on large language models, and then call smart contracts through the agent AI. This design not only broadens the potential functions of decentralized applications (DApps), but also creates an incentive mechanism for multiple participants such as computing power providers, stETH stakers, protocol development contributors, and community API operators. In the process of participating in the Morpheus AI ecosystem, users can obtain MOR tokens by staking or submitting code.
Kuzco
Kuzco is a distributed GPU cluster on Solana for LLM inference. Although this project is new, it is progressing rapidly. It has launched a public beta version for Mac and Linux users, and the number of online working nodes has reached 1,400. Kuzco provides an efficient environment and code support, allowing users to pre-install the required models and environments locally, and complete tasks only through network transmission. Compared with ionet, it performs better in resource usage. The usage rate of a single card can reach 90%.
Currently, participants who contribute computing power can receive $KZO points in return. Although its expected returns, development roadmap, and Token economic model have not yet been clearly announced, they have attracted the official attention of Solana. No financing has been carried out yet, but the issuance of Token may occur in the near future.
Golem
Golem is a decentralized computing power market based on Ethereum. Since its launch in 2015, it has been focusing on providing CPU computing power sharing services based on Ethereum. Golem is one of the earliest peer-to-peer computing power network protocols and has established a network node system including providers and requesters. Starting in 2022, Golem will lay out the GPU market, launch the first phase of the GPU Beta test plan, and launch the second phase in March 2024.
The total supply of $GLM is 1 billion, all of which are currently in circulation, with a market value of $561 million and the code is open source .
NodeAI
Node AI is a decentralized GPU computing power market whose parent company is EyePerformance. Rental fee reference: A10 0 is about $0.85/h, A10 is about $0.22/h. $GPU’s current market value is 121M.
GPU.Net
GPU.Net is a decentralized GPU resource network that aims to meet the computing power needs of new technologies such as generative AI, Web3 metaverse, cryptocurrency mining, and high-end graphics rendering by providing decentralized GPU infrastructure. In August 2023, GPU. Net completed US$500,000 in financing, with participation from Momentum 6 and Alphablockz, and support from NVIDIA and Taanga Studios. Early online pre-registration is currently online.
GamerHash
GamerHash utilizes the remaining computing power of players’ computers when playing high-end games for cryptocurrency mining. For example, a player may only be using 15% of their computing power while gaming, and GamerHash allows the remaining unused power to be automatically used to mine cryptocurrency without requiring additional action or monitoring by the user. In order to serve computer users with lower configurations, GamerHash launched the Play & Earn function. Users can choose to complete specific tasks, such as playing specific games or downloading apps, to earn GUSD, an internal currency pegged to the U.S. dollar, bringing a new revenue channel to users with limited hardware capabilities.
NodeSynapse
NodeSynapse is a GPU computing power marketplace that provides token holders with server hosting, Web3 infrastructure and GPU computing as well as a unique revenue sharing model.
Model assetization and model training
In the trend of combining blockchain technology with AI, an important development direction is to capitalize the AI model, or run the business model itself through tokenization. This approach has gradually shown its value in many specific application scenarios, such as in security audits, chat robots, AI consultants and other fields. Since the technical threshold for this type of project is relatively low and the implementation difficulty is not high, there are both top-level projects with abundant resources and many small-scale projects that are just starting out in the market. Therefore, in projects that combine blockchain and AI, it becomes particularly important to identify the quality and investment value of the project.
The key is to identify whether there are real innovation and technical barriers to the project. Simply trading the ownership or use rights of an AI model does not constitute true technological innovation. Real technological breakthroughs should focus on how to effectively verify the output of the model and how to ensure that the model is run in a decentralized environment.
saharalabs
Sahara Labs aims to solve data sharing and privacy issues in AI model training through its two core products: Knowledge Agent and Data. The project promotes data sharing and decision-making by providing individuals and enterprises with semi-automated autonomous agents and data operation tools while ensuring privacy security. Currently, Sahara Labs has successfully attracted 30 enterprise customers and achieved positive cash flow, demonstrating the market feasibility and practical value of its solutions. Sahara Labs is led by professors Sean and Tyler and has a team of more than 30 people. Professor Sean is a professor at the University of Southern California and has received many honors, including Samsung AI Researcher of the Year. Tyler has extensive experience in the blockchain industry and served as the investment director of Binance Labs.
Its product Knowledge Agent provides users with customized semi-automatic agents that support sharing and processing external or internal data to make decisions. These agents can be fine-tuned for private data, while providing privacy protection and data provenance verification tools, as well as a development toolkit to support multi-agent collaboration. Sahara Data Marketplace provides a comprehensive set of data operation tools including data collection, labeling, quality control (QA) and project management, solving data privacy issues. It has been adopted by well-known institutions such as MIT and Microsoft.
Sahara Labs has raised US$600M in funding and received US$6 million in support through a seed round, led by Polychain Capital, with participation from Sequoia Capital, Samsung Next, Matrix Partners, etc. The development roadmap of the project includes: Q2 of 2024, the Sahara Data data market will be launched; Q3, the test network and Sahara ID will be launched; Q4, the main network and Sahara Agent will be officially launched.
Bittensor
The Bittensor project is a decentralized open source project aimed at creating a neural network protocol on the blockchain. This protocol allows the creation of AI-driven decentralized applications (dApps) and enables peer-to-peer value exchange between AI models. Bittensor was co-founded by former Google software engineer Jacob Robert Steeves and former University of Toronto assistant professor and University of Waterloo postdoc Ala Shaabana, with James Woodman serving as chief operating officer. The project conducted a "fair launch" in 2021 and did not conduct any pre-mining tokens, VC rounds, private placement rounds, etc. The token TAO was generated through mining by miners. Known investors and market makers include DCG, GSR, Polychain Capital and Firstmask.
The core of the Bittensor network consists of miners and validators. Miners are responsible for submitting pre-trained models to receive rewards, while validators ensure that the outputs of these models are valid and accurate, and select the best outputs to return to users. For example, when a user requests an AI chatbot to answer a question, the question can be answered no matter how many nodes the Bittensor network is running. Queries are sent by users to validators, who then pass the queries to miners and rank the miners’ outputs, eventually returning the highest-ranked output to the user.
The development highlights of the Bittensor project include:
- Subnet subnet: All of Bittensor's AI applications actually run on its subnet. There are currently 32 subnets, each with specialized use cases. The threshold to participate in the subnet is $6,000 TAO, over $4 million in fees. Miners and validators can receive rewards of 7200 $TAO per day, and if the subnet is running well, the daily income can reach hundreds of thousands of dollars; otherwise, the subnet may be eliminated.
- Miners/Verifiers: The threshold for becoming a miner or verifier is relatively high. Miners need to conduct machine performance testing, and becoming a verification node requires staking a large amount of $TAO and ensuring that the equipment remains online. For miners, if they cannot provide optimized models or unique large-scale LLM models, they will soon be kicked out of the network.
The total supply of token TAO is 21,000,000, all of which have entered circulation, with a market value of $4,474,463,212. Already listed on exchanges MEXC, Gate.io, KuCoin, Bitget. Retail investors can choose to stake $TAO to existing verification nodes and enjoy the current annualized rate of return of approximately 15.85%.
iExec RLC
iExec RLC is a decentralized cloud resource platform based on Ethereum, aiming to change the current cloud computing model through integrated DApp, computing power market, and data market. This platform utilizes the original iExec Proof-of-Contribution (PoCo) consensus mechanism to ensure the effectiveness and security of off-chain calculations while protecting the platform from interference by dishonest participants. iExec was co-founded by Dr. Gilles Fedak and Dr. Haiwu He. Dr. Fedak serves as CEO. He has made pioneering achievements in the development of software and algorithms for cloud computing and large-scale parallel systems before entering the field of blockchain entrepreneurship, and has won several awards. Best Paper Award. Dr. Haiwu He, as the head of the APAC region, was a "100-person professor" at the Computer Network Information Center of the Chinese Academy of Sciences, leading research work in high-performance computing (HPC) and cloud computing. In 2017, iExec completed US$12 million in financing through ICO.
The core functionality of the iExec platform allows users to run containerized applications, primarily for short-running jobs and one-time tasks. While it does not currently support a wide range of use cases such as web or API-based applications, iExec has successfully powered DApps on its sidechain and has built worker pools to sort by task size. The platform's computing resource bidding system enables users and applications to issue work orders based on preset criteria, and cloud providers (workers) provide computing power based on the price they are willing to accept. In addition, iExec also introduces the concept of pool to organize workers (worker nodes). These pools can be either public or private, similar to traditional cloud service providers.
The PoCo consensus algorithm serves as the defense mechanism of the iExec market, ensuring trustful execution on the decentralized network and preventing possible adverse results and unreasonable objections. As of now, $RLC has a market cap of $322M, while its fully diluted value (FDV) is $387M. Since its launch in 2016, iExec has made remarkable achievements in the field of decentralized cloud computing and became an important member of the Intel AI ecosystem in 2020.
i Exec is continuously developing its technology and service scope, and future plans include: Research and Development (R&S) of AI and Large Language Models (LLMs); Research and Development of Privacy-Enhancing Technologies (PET); i Exec Oracle IDE development; Privacy Pass activity; Worker Pass second round issuance; Dapp store web interface release; Questbook bounty task season 3; Exec Web3 incubator season 3 and 4; PoCo consensus algorithm upgrade.
Allora
Allora is a self-improving, decentralized artificial intelligence network that can achieve more accurate inferences by building a probabilistic market. AI/ML agents in the Allora network use their data and algorithms to broadcast their predictions across the peer-to-peer network, and each reference these predictions to evaluate the predictions of all other agents. The network consensus mechanism combines these predictions and evaluations and allocates rewards to agents based on the quality of their predictions and evaluations. This carefully designed incentive mechanism allows Allora to continuously learn and improve, adjusting as the market evolves.
Allora builds its independent chain based on cosmos and implements pos, rewarding verifiers and ai working nodes. Its nodes verify the accuracy of each other's reasoning while performing reasoning, and predict the reasoning quality of each agent by directly evaluating each other. , distribute rewards based on evaluation results.
Allora's internal test network Edgenet was released on March 5. It is currently running test nodes in cooperation with multiple project parties and institutions, and plans to launch the main network in Q2. Its investment institutions are also relatively luxurious, with Polychain , Framework , Blockchain Capital , CoinFund , Delphi Ventures , dao5 , and Distributed Global all participating.
lPAAL
l PAAL AI is an AI ecosystem built using custom data sources and LLM. Users can create personalized AI, conduct professional strategic transactions, obtain market intelligence, etc., and deploy it to multiple platforms such as Telegram and Discord. PAAL AI tools include MyPaal, AutoPaal, and AutoPaal X. Invested by Google Cloud, Coingecko, OK X, etc. The total supply of PAAL is 1 billion, the circulating supply is approximately 650 million, and the market value is approximately US$100 million.
MyShell
My Shell is an AI platform based on opBNB that allows users to create chatbots, allows creators to call different models and external APIs in the same way, and allows third-party model providers and API providers to integrate their services into the platform for Used by developers. Incentivize creators and users to be active through Shell points and Shell Coin (Shell Coin is an airdrop voucher, and points can be exchanged for Coin). In October 2023, it raised US$5.6 million in seed round financing, led by INCE Capital, with participation from Folius Ventures, Hashkey Capital, SevenX Ventures, TSVC, OP Crypto, etc. Binance Labs also officially announced that My Shell has entered the sixth phase of incubation camp. The valuation is approximately US$57 million. Currently in the testnet development stage.
Qubic
Qubic is an L1 PoW platform that uses PoW computing power for AI training. The mainnet will be launched in April 2022. Founded by Sergey Ivano (also known as Come From Beyond or CFB). CFB is the third person in history to mine Bitcoin, and the NXT he founded was once one of the most successful ICOs. At the same time, he is also the founder of Iota.
Quorum Protocol: This protocol uses 676 computers (validators) to process transactions and execute smart contracts, ensuring the finality and reliability of the results.
AI Aigarth: This is an AI software running on top of Qubic, which obtains information from tens of thousands of Qubic - AI miners who use their computing power to create billions of artificial neural networks.
AI Agent
An AI agent can be understood as an intelligent agent with the ability to autonomously understand, remember, make decisions, use tools, and perform complex tasks. This kind of agent can not only teach users "how to do it", but also actually help users complete tasks. The AI agents specifically referred to here are those projects that use AI models to interact with blockchain technology, such as conducting transactions, providing investment advice, operating bots, enhancing decentralized finance (DeFi) functions, and conducting on-chain data analysis, etc. .
This type of AI agents has specific and direct application scenarios, that is, conducting transactions. They are particularly closely integrated with blockchain technology and can directly create revenue, introduce new transaction scenarios, and optimize the blockchain usage experience. This combination is actually an advanced narrative of decentralized finance (DeFi), which creates profits through trading activities, attracts capital investment, and creates speculation opportunities, thus driving the operation of the Ponzi flywheel effect.
AI trading agents are typically launched from the community with a low market value, low participation costs, and therefore good return on investment potential.
Morpheus
Morpheus is an AI decentralized computing power market built on Arbitrum to help users build agent AI based on large language models, and the agent AI calls smart contracts. Incentives include miners who provide computing power, stETH pledges, protocol development contributors and community API operators. The project team uses character names from the "Matrix" movies (Morpheus, Neo, Trinity) as pseudonyms on the website.
The core developer is David Johnston, who has many years of experience in investing and serving as CEO. He has been involved in the Bitcoin Foundation since 2013, then joined Multicoin Capital and Space Fund, and has been working in DLTx since 2021. In 2023, he created the "intelligent agent" framework and began to invest in the development of Morpheus.
Morpheus has a high level of community attention and is distributed in the form of fair launch. The pledge code in the first phase is well completed and has undergone security audits. There are also bug bounty in the community and the code security is also strong. However, the code update for the AI Agent is slow and the progress of the core module is unclear. The main methods of participation include: pledging stETH to participate in the distribution of the capital part; providing code support for Morpheus; participating in mining after the launch of the computing network on May 8; making some network billboards and tools and becoming a community contributor.
QnA3.AI
QnA3.AI provides full-life-cycle, full-scenario services such as information management, asset management, and rights management. Technically, QnA3 uses RAG (Retrieval Augmented Generation) technology, combined with LLM understanding, pre-training, scalability and reasoning capabilities, to optimize information retrieval and generation, and improve timeliness and accuracy. At the same time, the team emphasizes the concepts of AI + Trading and AI + DePIN in asset management and rights management, and promotes the practice of decentralized machine learning. QnA3.AI was incubated by Binance Labs and received investment from the Solana Foundation on February 25, 2024.
The QnA3.AI project started in January 2023 and has experienced rapid development in just one year. The project launched a question and answer function in June, and the number of users increased rapidly; in September, it launched an intent-centered Telegram bot, and the number of users surged to 300,000; by December, the number of daily active users reached the first place on BNB Chain, and the total number of users exceeded 2 million. So far, the number of users has exceeded 10 million.
Autonolas
Autonolas is an open marketplace for creating and using decentralized artificial intelligence agents. Autonolas also provides developers with a set of tools for building off-chain hosted AI agents with the ability to connect to multiple chains including Polygon, Ethereum, Gnosis Chain and Solana. They currently have a number of active proxy proof-of-concept products, including products for prediction markets and DAO governance.
David Minarsch is the CEO and co-founder of Valory, the parent company of Autonolas. He holds a PhD in economics from Cambridge University and specializes in multi-agent services. Previously, he served as director of multi-agent services at Fetch.ai. Investors include Signature Ventures, Semantic Ventures, True Ventures, Proof Group and others.
SingularityNet
SingularityNET is an open, decentralized artificial intelligence service network whose mission is to create a decentralized, democratized, inclusive, and beneficial general artificial intelligence. Developers can publish their services to the SingularityNET network, where they can be used by anyone with Internet access. Developers can charge for the use of their services using native AGIX tokens.
The creators are Dr. Ben Goertzel and Dr. David Hanson, one of the core developers of the earlier famous intelligent humanoid robot Sophia. SingularityNET's investors include Fundamental Labs. In May 2022, SingularityNET and Singularity DAO received a $25 million investment commitment from the investment group LDA Capital.
SingularityNET's services can provide inference or model training across multiple domains, such as image, video, speech, text, time series, bio-inspired AI and network analysis. These services can be as simple as packaging well-known algorithms, to complete end-to-end solutions to industry problems, or to standalone AI applications. Developers can also deploy autonomous AI agents that interoperate with other services on the network. For example, promote trust and automate transactions through multi-party escrow, launch new artificial intelligence services and organizations on the blockchain, track successful API calls, and define pricing strategies.
Fetch.AI
Fetch.AI is one of the first deployed artificial intelligence agent protocols and has developed an ecosystem for building, deploying and using agents on-chain using its FET token and Fetch.AI wallet. The protocol provides a comprehensive set of tools and applications for working with proxies, including in-wallet functionality to interact with proxies and issue commands.
Most of the Fetch development team graduated from prestigious schools or came from Fortune 500 companies, and they are all experts in AI or algorithm-related fields. It is worth mentioning that the CEO and CTO are both related to Deepmind, an AI company owned by Google. Its signature product Alphago once defeated Korean Go champion Lee Sedol and also defeated professional players in StarCraft 2. Most of the consulting team are professors from prestigious universities with strong academic backgrounds. In terms of financing, in March 2021, Toronto digital asset company GDA Group invested US$5 million; in March 2022, Fetch launched a US$150 million development fund to encourage developers to develop projects in its ecosystem. The fund is led by MEXC Global Huobi and Bybit participated in the investment.
Humans.ai
is an AI blockchain platform that brings together an ecosystem of stakeholders around the use of AI for large-scale creation, integrating a library of AI tools into a creative studio suite where users can pick and realize their own ideas . Individuals have the right to create and own digital likenesses of themselves, which can be used by themselves and others to create any number of digital assets. Synthetic media, AI applications, and other digital assets can leverage blockchain technology to generate NFTs. The total supply of Humans.ai token HEART is 7.8 billion, the circulating supply is approximately 5.12 billion, and the market value is approximately US$85.96 million.
Metatrust
Metatrust is a Crypto-powered AI Agent network. MetaTrust builds the world's first comprehensive set of Web3 security solutions covering the entire software development life cycle (SDLC), founded by the world-renowned research team of Nanyang Technological University. The seed round raised $10 million, led by M23 Fund, with participation from Redpoint, ABCDE, Long Hash Ventures, and Hash Capital.
AgentLayer
AgentLayer is based on the core chain structure of OP Stack + EigenDA and is a decentralized agent network built by the Metatrust team. EigenDA can take into account the efficient availability of data and greatly improve the overall performance and security of AgentNetwork. As a complete toolbox for layer2, OP Stack supports Optimism and other Rollup implementations, and provides key components such as sequencers, nodes, contracts, etc., which can be independently applied to different expansion implementation plans.
DAIN
DAIN is building an agent 2 agent economy on Solana. The goal of DAIN is to allow agents from different enterprises to seamlessly interact with each other through a common API, thereby greatly opening up the design space of AI agents, with a focus on implementing agents that can interact with web2 and web3 products. In January 2024, DAIN announced its first cooperation with Asset Shield, enabling users to add "proxy signers" to their multi-signatures, who can interpret transactions and approve/reject according to rules set by the user. DAIN is developing an Agent that can obtain information on the chain and conduct transactions, analyze and process the data on the chain, and give purchase and other suggestions in the form of chat.
ChainGPT
ChainGPT is an AI model designed for blockchain and encryption. Its products include AI NFT generator, AI generated news, AI transaction assistant, smart contract generator, smart contract auditor, etc. ChainGPT won the BNB Ecosystem Catalyst Award in September 2023. The market value of the token CGPT is approximately US$40.63 million.
Data capitalization
The data track combines AI and encryption technology and is regarded as an area with huge potential. Its core value lies in data and computing power as the basic production materials of AI. Unlike the decentralization of computing power, which may reduce efficiency, it is reasonable to decentralize data processing to a certain extent because the production of data itself is decentralized. Therefore, in theory, the data track has great development potential in the field of combining AI and encryption technology.
However, the core challenge facing this field is the lack of mature data trading markets in traditional markets, which makes effective valuation and standardization of data extremely difficult. It is difficult for project parties to attract large amounts of capital through token incentives because in the absence of an effective evaluation mechanism, the value of data is difficult to be reflected in token prices. This situation has led to a potential rupture of the "flywheel effect". Even projects with great potential within the data track have difficulty leveraging large-scale capital through low-cost token incentives.
Despite the challenges, there are also projects that are trying to approach the data track from different angles and explore new models and solutions. For example, projects like Sapien focus on providing a market for labeled data required for AI, trying to solve the problem of obtaining AI training data. In addition, considering the high demand for bandwidth for AI decentralized training, bandwidth providers have become a new entry point in the data track. Traffic is an easily quantified asset, and projects such as grass.io are trying to build a decentralized bandwidth ecosystem to indirectly provide data for AI training. This demonstrates attempts to find new opportunities and solutions within the data track.
Synesis One
Synesis One is a data crowdsourcing platform on Solana that allows anyone to earn $SNS by completing microtasks for training AI. The off-chain annotation is passed to the on-chain verifier for verification and rewards will be issued.
The Korean development team was reorganized on October 11, 2023. Isaac Bang, former head of growth at Synesis One, serves as CEO. Synesis One is currently cooperating with Mind AI. Mind AI will use the data collected by Synesis One to train AI, and is expected to support audio, video and image data as well as RPA (robotic process automation) in the future. Mind AI has signed cooperation agreements with General Motors and the Indian government. Financing amount is US$9.5 million.
Grass.io
Grass.io is a decentralized bandwidth marketplace. Users can sell their excess bandwidth to artificial intelligence companies that use it to scrape data from the internet. There are currently more than 2 million personal IP network addresses, and only users who are online in real time can obtain Grass points.
It has received $3.5M in seed round financing, led by Polychain Capital and Tribe Capital, with participation from Bitscale Capital, Typhonv and others. Tokenomics has not been announced, but according to team members, tokens are expected to be issued by the end of the year.
GagaNode
GagaNode is the next generation decentralized residential bandwidth market, aiming to alleviate the global IPv4 address shortage problem through Web3.0 technology. Currently, it has good support for clients of all major systems (including mobile terminals). The code has been open sourced .
Ocean
Ocean Protocol allows anyone with a sought-after data set to tokenize their data and make it available on the Ocean Market, where users can create data NFTs representing their data and purchase them using data tokens. You can participate by building an AI trading robot, staking tokens, and participating in data challenges.
Founder Bruce Pon has worked at Mercedes-Benz for over 5 years, guest lectured at MIT, Oxford, NYU, and the European School of Management (ESMT), and conducted research on blockchain, cryptocurrency, and decentralized technologies. He has given more than 200 public lectures. Has over 35 consultants worldwide.
ZKML
By introducing ZK into the field of machine learning, it is possible to quickly verify the computational correctness of the model without revealing information. The trustless verification of the model allows zkml to be used as a trustless proxy, or be used in some privacy scenarios, such as the application layer represented by Worldcoin.
At the same time, machine learning may be run in smart contracts through ZK, allowing smart contracts to achieve greater automation and flexibility; at the same time, as a "black box", the ML model's trust problem can also be solved through ZK verification.
As mentioned above, the infrastructure part of Zk ML is divided into on-chain ML represented by Modulus Labs, trusted inference proof services represented by Giza and EZKL, and also includes co-processors represented by Risc Zero and Axiom. This track is still in its early stages, and its technical capabilities and team background need to be considered.
WorldCoin
Launched by Open AI, WorldCoin requires users to scan their irises to generate a unique ID. In the future, biometric IDs could be self-stored on personal devices by using encryption, with the required model used to locally verify these biometric information. Users can provide proof of their biometric information without revealing their identity, thus resisting Sybil attacks while ensuring privacy. It is currently the leader in the field of AI. Worldcoin has raised over $250 million in multiple rounds of funding from investors including a16z, Khosla Ventures, Bain Capital Crypto, Blockchain Capital, and Tiger Global.
There are currently more than 1 million active users. Authenticate on DRiP using a device-verified World ID or an Orb-verified World ID and instantly unlock bonus Droplets.
EZKL
EZKL is open source and proves that a model produced certain results by generating zero-knowledge proofs without having to reveal the model itself. Over the past few months, EZKL has made significant progress in enhancing its zkML solution, primarily focused on reducing costs, improving security, and speeding up proof generation. For example, in November 2023, EZKL integrated a new open source GPU library that can reduce aggregate proof time by 35%; in January this year, EZKL released Lilith, a software solution for when using EZKL proofs Integrate high-performance computing clusters and orchestrate concurrent job systems. At present, it has been officially announced to cooperate with OP.
Jason Morton is the CEO of EZKL. He also has a PhD in mathematics from the University of California, Berkeley, and a BA in economics from the University of Michigan.
Giza
Giza is a machine learning platform built on StarkNet. In addition to providing tools for creating verifiable machine learning models, Giza also plans to implement a web3 equivalent of Hugging Face, open a user marketplace for zk ML collaboration and model sharing, and eventually integrate decentralized computing products. Recently Giza Datasets is a Web3 dataset framework optimized for ML workflows and structured for the purpose of model training.
Giza was founded in October 2022 by Cem Dagdelen, Fran Algaba and Renç Korzay. Completed a $3 million Pre-Seed round of financ
