Charles Hoskinson: Crypto Could Become the Economic Layer for AI

Charles Hoskinson said the blockchain industry is entering a new phase in which traditional finance is moving closer to open networks, while privacy and cryptographic proofs are becoming necessary for wider business adoption. He estimated that approximately one billion new users and $10 trillion in assets and value could enter the crypto economy by the end of 2030.

His argument places Cardano, Midnight and the ADA ecosystem within an architecture designed for regulated finance, global payments and autonomous AI agents.

Privacy and Midnight Shape the Next Blockchain Phase

Hoskinson identified two parallel trends. The first is the convergence of Web2 and Web3 through regulated products, tokenized assets and financial institutions seeking access to blockchain markets operating globally and around the clock.

He described this transition as a move toward “Web 2.5.” Financial institutions want access to global liquidity and programmable products, but they also require privacy, customer identification, regulatory compliance and controlled disclosure of sensitive information.

The second trend involves privacy, cryptographic proofs, blockchain abstraction and programmable compliance. According to Hoskinson, consumers will not accept financial systems in which every payment exposes their complete transaction history. Institutions also cannot operate effectively on infrastructure that publicly reveals customer relationships, trading positions and commercial strategies.

He presented Midnight as a hybrid response. The network focuses on programmable privacy and selective disclosure, allowing transactions and rules to remain verifiable without making every underlying detail public. Developers can use Midnight to build applications that process sensitive personal or commercial data, while users can prove required information without revealing the underlying records.

Hoskinson argued that private consortium blockchains cannot become neutral global infrastructure because access remains controlled by a limited group of institutions or governments. His preferred model uses an open, permissionless foundation within which regulated organizations can create domains with their own identity, KYC, AML, audit and market-access requirements.

The underlying protocol remains open, while individual applications determine who may enter a regulated environment and which facts participants must prove. Hoskinson connected this model with Midnight and with sovereign networks that rely on an open blockchain for final settlement.

He also expects more traditional assets to move on-chain as institutions pursue continuous trading, programmable compliance and composable financial products. Blockchain infrastructure could allow assets, loans, collateral and yield-generating instruments to interact within the same transaction flow rather than remain separated across institutions and national markets.

Technology Platforms Move Into Global Payments

Hoskinson expects the future of digital payments to be shaped not only by banks and traditional processors, but also by companies such as Google, Apple and Meta. These companies already control devices, operating systems, digital identities and consumer platforms serving billions of people.

Stablecoins could provide the value-transfer layer, while smart contracts enforce settlement and compliance requirements. Mobile devices already support biometric authentication, secure hardware and digital wallets, giving technology platforms a distribution advantage over conventional banks.

Hoskinson argued that Google could build a simpler global payment experience than a large bank because it already controls the platform on which financial applications operate. The same structural advantage applies to Apple, Samsung, Microsoft and Meta, which can connect wallets, stablecoins and digital identity with existing devices and communication services.

The transition would extend beyond payments between people. AI agents change how online content, data and digital services are purchased.

An agent searching for one piece of information may not click an advertisement or purchase a monthly subscription. It could instead pay a small amount for access to a single article, dataset or specialized service. Hoskinson connected this model with stablecoin micropayments and protocols such as x402, which allow software to pay for digital resources programmatically.

Publishers, data providers and software platforms could charge for individual requests, while an agent operates within a predefined research or purchasing budget. Payments would occur directly between software systems without requiring a conventional user account, manual approval for every purchase or card infrastructure designed primarily for humans.

The thesis also has a direct Cardano connection. Cardano components have already been merged into the official x402 codebase for mainnet and testnet use, creating a technical route for ADA applications to participate in programmable agent payments.

The same infrastructure could support automated royalties. If an AI system creates content derived from another person’s intellectual property, blockchain records could document the source and distribute the required payment when the new content is created or sold.

Midnight City Tests an Economy for AI Agents

During the interview, Hoskinson demonstrated Midnight City, an experimental digital environment built to study autonomous agents. According to his description, each agent has an identity, a human owner, its own wallet, an inventory and a persistent presence within the system.

The project explores how users can delegate funds to an agent, define what it is permitted to purchase and allow it to operate without requesting confirmation for every transaction. An agent could locate another specialized agent, pay for data or expertise and use the acquired service to complete a broader task.

Hoskinson gave the example of an agent tasked with building an application but lacking the required programming knowledge. Instead of returning to the user for every decision, the agent could find another agent with the necessary expertise, negotiate a fee and pay to acquire the required capability from its delegated budget.

Cryptographic signatures and zero-knowledge proofs would serve more than a privacy function in this model. They could limit what an agent is authorized to do, verify that predefined conditions were satisfied and prevent transactions that fall outside its delegated permissions.

Midnight City does not yet demonstrate a large commercial economy of autonomous agents. Hoskinson presented it as a testing environment for identity, wallets, delegated authority and coordination between agents, without disclosing production usage figures or measurable economic volume.

His broader prediction is that the internet will eventually contain more AI agents than human participants and that software agents could generate a substantial share of future blockchain transactions. Those agents will require payments, data ownership, privacy, verifiable rules and automated compensation systems.

Hoskinson summarized the thesis directly: “I think crypto is going to eat AI over the next five years or ten years.”

In his model, blockchain could provide capabilities that cards, bank transfers and closed technology platforms do not combine within one infrastructure layer: programmable money, portable identity, data provenance, automated royalties and private execution.

Midnight City gives that thesis a concrete experimental setting. Its agents are not limited to generating responses. They are designed to operate with identities, restricted permissions and funds that can be used to transact with other agents. In Hoskinson’s vision, that distinction separates AI that processes information from AI that can participate independently in a verifiable digital economy.