Discover - Article 1
Long developed in parallel, AI and Web3 now seem on the verge of converging
Discover - Article 2
DePINs for decentralised AI management
Discover - Article 3
Blockchain and AI: Synergies across the entire AI value chain
Discover - Article 4
Jonathan Schemoul ( "True AI with free will can only be achieved with decentralised technologies".
Go deeper - Article 5
How blockchain can help you prove your humanity in cyberspace
Go deeper - Article 6
Worldcoin, the flagship AI project of the crypto
Go deeper - Article 7
When crypto miners diversify into AI
Go deeper - Article 8
Map of Web3 players working in AI
Go deeper - Article 9
Market analysis: how are AI tokens performing?
Go deeper - Article 10
How AI can help make blockchain more secure
No items found.

Blockchain and AI: Synergies across the entire AI value chain

Subscribe to our smart subscription to read this report in full

Blockchain and AI: Synergies across the entire AI value chain

Tout ce qui compte en Web3. Chaque semaine.
Try insider for free for 30 days.

In 2024, worldwide, nearly 85% of company directors plan to increase their investment in AI. The current focus is mainly on identifying and implementing use cases, motivated in particular by the search for cost savings and productivity gains. However, structural issues are emerging that will eventually need to be addressed: what about the confidentiality and protection of data used for training in particular? How can we prevent the centralisation of power in the hands of the tech giants, who are often American? How can model training be made more affordable? These are all issues to which decentralised technologies can provide answers.

Thus, following a presentation of the main elements of the AI value chain, we will explore the synergies between blockchain and AI.

What is the AI value chain?

This value chain starts with the hardware, which is needed to provide the computing power, followed by the data that needs to be collected and stored, then the algorithmic brick that trains the artificial intelligences and makes them available to users. Finally, we find the applications that use these models.


An image containing text, screenshot, Font, automatically generated designDescription
Figure 1 - AI value chain

Classified according to this value chain, here are some use cases, starting with the application level: 

APPLICATIVE - Copyright protection for content created by generative artificial intelligence

The advent of generative AI raises many questions about intellectual property: how can the authors of works used to train models enforce their rights? What about copyright on a work generated by generative AI? If an AI creation infringes the copyright of a work, who is responsible?

While the legal answers to these questions are as yet unclear, blockchain technologies could nevertheless provide powerful technical means of tracing the development and training data of generative AIs, enabling authors to assert their rights. It is also easy to imagine being able to trace the creations generated by AI on the blockchain in order to assert copyright to all the players in the chain.

Story Protocol is an example of an initiative trying to respond to these issues. It is presented as a "programmable IP layer" that creators could use to trace the use of their work and ensure that copyright is collected. It is currently in the beta testing phase on Ethereum's Sepolia testnet.

APPLICATIVE - Decentralised infrastructure of autonomous agents

Autonomous agents in artificial intelligence (AI) are software or hardware entities endowed with a certain degree of autonomy and capable of acting in a given environment to achieve specific goals. They are designed to perform tasks autonomously, adapting to changes in their environment and learning from their experiences.

Some examples of autonomous agents are: autonomous vehicles (cars, drones.), robots (manufacturing or warehouse), customer service agents such as certain chatbots used on websites.

The market for autonomous agents is booming, recording a high growth rate. Companies such as Meta, Open AI with their Meta AI agents and GPT are major players. It is therefore a highly centralised market with a strong American dominance. 

This is why it is relevant to look at the possibilities of decentralised infrastructures for building these agents. This is, for example, what is proposed by the Fetch.AI protocol, the main features of which are as follows:

  • Integrated digital economy: integrates a token economy (FET) that incentivises agents created on its platform to provide useful services and contribute to the network, creating an ecosystem of economically viable autonomous agents.
  • Decentralisation and security: uses blockchain technology, enabling a decentralised and secure infrastructure for autonomous agents. This makes them more resistant to breakdowns, attacks and centralised control.
  • Data protection: Through the use of blockchain and advanced cryptography technologies, Fetch.AI ensures that interactions and transactions between agents are secure and that sensitive data remains confidential.
  • Cost reduction: By allowing agents to interact directly without intermediaries, Fetch.AI reduces the costs associated with transactions and services. The platform creates an open market environment where services can be negotiated at the best price, stimulating efficiency and innovation

It should be noted, however, that to date, Fetch.AI-type solutions still have some way to go, particularly in order to be truly scalable and interoperable. 

Following these two use case examples at the application level, let's move on to the algorithmic level.

ALGORITHM - Decentralising machine learning

Learning artificial intelligence (AI) models is a process by which a computer model learns from data to perform specific tasks such as image recognition, natural language understanding or trend prediction.

These learning processes are extremely expensive, with Google's Gemini Ultra reportedly costing nearly $191 million to train and OpenAI's GPT-4, reportedly costing $78 million. These exorbitant costs are a major barrier to entry for smaller players. 

The decentralisation of learning enabled by blockchain technologies therefore makes it possible to reduce training costs and thus open up the possibility of developing their models to a wider range of players.

This is notably what the Bittensor protocol offers. Bittensor through its decentralised platform and its digital token the TAO, allows developers and data scientists to collaborate securely and transparently on learning and training AI models while using the principles of the decentralised economy. The network players are rewarded in CAT for their involvement/contribution. In addition, all the AIs in the network can use data made available to them.

For the purposes of optimising resources, obtaining optimum performance, security as well as remaining flexible and scalable, the Bittensor network is divided into "subnets" specialising in a particular category of AI tasks, such as text generation, image recognition. Ownership and governance of these subnets are decentralised, enabling open collaboration. After this major use case at the algorithmic level, let's move on to the data level.

DATA - Decentralised Data Markets

Data is the keystone of all artificial intelligence systems. The accuracy and performance of algorithms derive from the quantity and quality of the data on which they have been trained. For example, OpenAI's GPT-3 model has been trained on a dataset containing hundreds of billions of words. In 2022, the global market for training data for AI was valued at $1.62 billion.

We are, however, faced with a paradox, because while these algorithms need vast amounts of data, a widespread awareness of confidentiality is on the march, hence the advent of web3. This follows numerous scandals in the web2 - Cambridge Analytica, or more recently Linkedin Data Breach in 2021 impacting over 700 million users.

The subject of training data sourcing must therefore be treated with great caution. And it is in this way that blockchain technology is a vector of value, notably by offering decentralised data markets that enable companies, governments and individuals alike to share and value their data while protecting their ownership.

This is notably what Ocean Protocol offers, through three mechanisms: 

  • Data Tokens: Ocean Protocol allows data owners to tokenise their data assets in the form of Data Tokens. These tokens represent ownership over datasets and can be traded or sold on Ocean Protocol's marketplace.

  • Ocean Market: a decentralised marketplace developed by Ocean Protocol, which allows data provider users to publish datasets as Data Tokens representing either ownership or access to specific data. Data-consuming users can then search for and purchase the data that meets their needs.

  • Compute-to-data: Ocean Protocol's Compute-to-Data technology enables users to analyse and use data without having to take it out of their secure environment. This minimises the risk of data confidentiality breaches.

The benefits of Ocean Protocol's marketplace, Data Tokens and Compute-to-data system are as follows:

  • Privacy protection and data security: via the Compute-to-data system on the platform

  • Dynamic pricing: The platform uses a dynamic pricing mechanism based on supply and demand. Data Token prices can vary depending on the quantity purchased and the frequency of use of the data.

  • Transparent and secure access: All transactions on Ocean Market are recorded on the blockchain, guaranteeing transparency and security. Users can check the full transaction history and authenticity of Data Tokens.

  • Interoperability: Ocean Market is designed to be interoperable with other blockchain and AI services, facilitating the integration and use of purchased data in various applications and platforms. 

However, decentralised data market players face considerable challenges as they seek to compete with giants such as AWS Data Exchange, Databricks or Google Cloud Platform. Their strong market footprint and users' habits of using centralised platforms maintained by established companies make the move to a smaller decentralised marketplace complex, especially given the perceived complexity of blockchain.

DATA - Decentralised Data Storage

The development of AI and machine learning is strongly driving demand for massive data storage solutions, as these technologies require huge amounts of data for model training. As a result, the next-generation storage market, which includes storage solutions for AI, is expected to reach USD 95.13 billion by 2029, with a CAGR (Compound Annual Growth Rate) of 7.37% over the period 2024-2029.

The market is currently held by players such as AWS, Microsoft Azure, or Google Cloud Platform. This centralisation of massive players can pose several issues: 

  • Data control and surveillance leading to privacy and surveillance concerns, especially if companies or governments exploit this data for unethical purposes or without the explicit consent of users.

  • Cost: Managing and maintaining centralised data centres can be very expensive. These costs include not only the physical infrastructure and hardware, but also energy consumption, cooling and technical maintenance. These high costs can be passed on to users in the form of higher charges for storage and data access.

  • Performance and latency issues: Centralised data centres can also suffer from latency issues, particularly when users accessing data are geographically remote from the data centre. This can slow down access to data and reduce the responsiveness of applications.

In response to these centralised players, decentralised storage initiatives such as Filecoin are tackling the market and offering innovative solutions for data storage that are based on decentralised storage infrastructures. The value propositions of these decentralised storages are: 

  • security and reliability: These solutions incorporate advanced cryptography protocols to ensure that data remains secure and private, offering peace of mind for users concerned about the confidentiality of their information. In addition, data is distributed across a wide network rather than concentrated in a few data centres. This distribution helps guard against data loss and service interruptions

  • competitive costs: by allowing individuals to rent their unused storage space and offering a hypercompetitive market, costs are driven down

  • improved access speeds thanks to the geographical proximity of servers to end users. 

As with the decentralised data market, decentralised storage players are up against stiff competition and usage habits that are now well established.

Let's move on to the hardware level with the hardware brick.

HARDWARE - Decentralised computing power

Computing power is the ability of a computer system to perform a given number of calculations per second. It is essential for data-intensive applications such as artificial intelligence (AI). 

The market for it was valued at USD 45.7 billion in 2023 and is expected to reach USD 81.3 billion by 2032. By way of example, in April 2023, the execution cost for a day of ChatGPT was estimated at USD700,000.

The computing power market, like the data storage market, is highly centralised. This leads to control and monitoring issues, as well as monopolistic situations, particularly on the pricing offered to users. 

This is why players such as iExec, Golem or Render (specialising in 3D content) are offering decentralised computing power marketplaces. These blockchain-based marketplaces allow anyone to buy and sell computing power.

As we have seen from this non-exhaustive list, blockchain has relevant application cases across the AI value chain, providing avenues for decentralising the existing ecosystem in particular. However, we need to bear in mind that we are still in the early stages of this work, and that there is unfortunately still a long way to go before we can compete with the centralised market leaders. We should also remain cautious before investing in Blockchain x AI projects, as many players are currently taking advantage of the AI craze to launch dubious projects.

An image containing text, screenshot, Font, automatically generated designDescription
Figure 2 Examples of Blockchain use cases on the AI value chain