BlackRock’s Digital Assets Research team has published a whitepaper arguing that agentic AI may become a structural source of demand for blockchains and other programmable payment infrastructure.
Image source: ChatGPT
The paper, titled ‘The Machine-Native Economy’, says AI and digital assets are starting to converge. It describes AI as machine-native intelligence and digital assets as machine-native money.
The authors define agentic AI as systems that plan and carry out multistep tasks with limited human intervention. These systems can buy goods, pay for data and initiate financial transactions. The paper says this will likely increase demand for blockchains and other programmable payment rails. It also says the ecosystem is still nascent and that agentic payment activity remains limited.
ACH and card networks vs blockchain
BlackRock says ACH and card networks already support significant automation. Nevertheless, it adds that they are less suited to always-on, very low-value transactions that need programmable execution. The paper lists four main ‘standard’ network constraints:
- Account setup, credentialing and authorization steps that may require a person.
- Merchant acceptance fees that can make very small payments uneconomic.
- Settlement and finality limits. Most ACH volume settles within one business day or less. Card authorization is near-instant, but merchant settlement and dispute finality can take longer.
- Possible scalability limits as machine-generated transaction volumes grow.
Due to these limitations, crypto-native rails fit high-frequency, sub-cent machine-to-machine payments, such as API calls, on-demand data and consumption-based compute. At the same time, BlackRock expects modified traditional payment systems to remain important for business-to-machine and consumer-to-machine payments.
Agentic payment protocols
The whitepaper also describes a stack of modern protocols that connect AI agents to tools, to each other and to payment systems.
- MCP (Model Context Protocol): introduced by Anthropic in November 2024. It standardizes how AI applications access external data and workflows.
- A2A (Agent2Agent): launched by Google in April 2025. It lets agents communicate and coordinate across platforms.
- x402: an open payment protocol from Coinbase. It uses the HTTP 402 “Payment Required” status code. It is blockchain-agnostic, with stablecoins such as USDC as an early primary use case.
- Machine Payments Protocol (MPP): developed by Stripe and Tempo. It supports payments for APIs and other HTTP resources, settled in stablecoins or through traditional methods.
- Agentic Commerce Protocol (ACP): developed by Stripe and OpenAI. It enables programmatic checkout between agents and businesses, and sellers keep their existing payment infrastructure.
- Agents Payment Protocol (AP2): from Google. It uses cryptographic mandates and audit trails to document user authorization.
- Trusted Agents Protocol (TAP) developed by Visa. It helps merchants verify trusted agents and receive payment credentials securely.
Role of stablecoins
BlackRock expects stablecoins to handle most transactional use in agentic commerce. The paper cites research from the Bitcoin Policy Institute. In controlled simulations, AI model outputs generally favored stablecoins for everyday payments. They favored bitcoin for long-term value preservation. BlackRock stresses that these results come from simulated model responses. They do not reflect observed agent behavior. At the same time, there are three data points that suggest stablecoins could be the most suitable option.
Supply: Stablecoins are the largest category of tokenized real-world assets. Their circulating market capitalization passed $300 billion as of September 2026.
Volume: Adjusted stablecoin transaction volume exceeded $11 trillion in 2025. BlackRock says this is in the same broad range as the annual payment volumes of Visa and Mastercard. It adds, however, that the measures are not directly comparable.
Scale & growth rate versus ACH: ACH moved $93 trillion in 2025. That is still well above stablecoin volume. Yet, from 2020 to 2025, adjusted stablecoin volume grew at an 80% compound annual growth rate (CAGR). ACH grew at about 8.5% over the same period.
The authors expect regulation to support further adoption. They name the GENIUS Act in the United States, MiCA in the European Union, Hong Kong’s stablecoin licensing regime and Singapore’s stablecoin framework.
Many major stablecoins are issued on more than one blockchain. This lets market participants choose a settlement network based on cost and technical fit.
Computing and cloud tech in agentic finance
As prerequisites for agentic finance continue to develop, the whitepaper argues that computing capacity is becoming an increasingly important economic resource. The scale of this shift is reflected in several estimates. Goldman Sachs expects cumulative AI capital spending to exceed $5 trillion between 2025 and 2030. Meanwhile, Bloomberg-compiled analyst estimates put combined 2030 revenue for AWS, Microsoft’s Intelligent Cloud and Google Cloud at around $1.1 trillion, representing a 29% annual growth rate from 2025 levels. At the same time, inference is expected to become the largest AI workload by 2030 and account for a growing share of data-centre power demand, according to McKinsey.
BlackRock says standardized claims on compute capacity could be tokenized. These claims could be transferred, pledged as collateral and settled on programmable infrastructure. The paper points to GPU-backed financings and financing platforms built on usage-linked compute revenue as early examples.
The paper then describes how this could work for agents. An agent queries real-time marketplace APIs and compares capacity by price, performance, latency, location and hardware. It then picks the best option and pays per use or per job through protocols such as x402.
