BlackRock published a research paper titled "The MachineNative Economy" arguing that AI adoption represents an underappreciated source of demand for digital assets, tying the growth of autonomous AI agents directly to stablecoins, blockchain infrastructure, and a new investable market for computing power.
"Our latest research paper explores the growing connection between AI and digital assets and explains why broad AI adoption may drive new demand, utility and applications across the digital asset economy," BlackRock posted on X.
The paper's authors, Will Su, Robert Mitchnick, Jay Jacobs, and William Helm, lead BlackRock's digital assets research, digital assets, US equity ETFs, and US iShares product innovation teams respectively.
The tokenization parallel driving BlackRock's core thesis
BlackRock's central argument rests on a structural comparison between how large language models process language and how blockchains process value.
"AI represents machine-native intelligence, while digital assets represent machine-native money," the paper states.
In an LLM, a tokenizer divides text into words or sub-words and maps them to numerical identifiers, which an embedding layer then converts into vector representations for computation. Blockchains apply what the paper calls "a functionally comparable, though technically distinct, process" to financial claims, representing cash, fund interests, or ownership stakes as standardized digital tokens on a distributed ledger.
This tokenization parallel is not merely rhetorical framing. Both systems solve the same underlying problem: converting unstructured real-world inputs, whether human language or economic ownership, into discrete, machine-processable units that computational systems can act on without human interpretation at each step. The paper's argument is that this shared architecture gives LLM-based AI agents a more direct interface with blockchain data than with the fragmented legacy databases and APIs that dominate traditional finance, since both AI tokens and asset tokens are already structured for machine consumption rather than requiring translation layers.
Why existing payment rails struggle with agent transactions
The paper identifies specific structural weaknesses in traditional payment infrastructure that make it poorly suited for AI agent commerce. These include account setup and credentialing processes that require human involvement, merchant acceptance fees that make very low-value transactions uneconomic, and settlement delays, noting that "most ACH volume settles within one business day or less, while card authorization is near-instant but merchant settlement and dispute finality can take longer."
The paper details how emerging protocols address these gaps. MCP, introduced by Anthropic in November 2024, standardizes how AI applications access external data. A2A, launched by Google in April 2025, enables agents to communicate across platforms. Coinbase's x402 protocol uses the HTTP 402 "Payment Required" status code to facilitate machine-initiated payments, described in the paper as blockchain-agnostic with stablecoins like USDC as an early primary use case. The paper also references Stripe and Tempo's Machine Payments Protocol, Stripe and OpenAI's Agentic Commerce Protocol, Google's AP2, and Visa's Trusted Agent Protocol as parallel standards connecting agentic transactions to both blockchain rails and traditional payment infrastructure.
The paper walked through an illustrative example: a user asks an AI agent to book a trip under $2,500. The primary agent accesses calendar and payment details through MCP, delegates flight and hotel research to a sub-agent through A2A, and that sub-agent pays for airfare and rate data through x402 settled on-chain, before the primary agent completes bookings and returns receipts to the user.
The scale of stablecoin volume already recorded
BlackRock's paper cites adjusted stablecoin transaction volume exceeding $11 trillion in 2025, placing it in the same broad range as Visa and Mastercard's annual payment volumes. That figure remained below the $93 trillion transferred over ACH the same year, but stablecoin volume grew at an 80% compound annual growth rate from 2020 to 2025, compared with approximately 8.5% for ACH over the same period.
The paper's own citations point to Bitcoin Policy Institute research examining what stablecoins AI models themselves favor in simulated payment scenarios, reporting that model outputs generally favored stablecoins for everyday payments and bitcoin for long-term value preservation. BlackRock was careful to characterize this as reflecting "simulated model responses rather than observed agent behavior," a distinction that matters because it separates a theoretical preference AI models express in controlled testing from confirmed real-world transaction patterns that have not yet materialized at meaningful scale.
Compute as the next digital asset frontier
Beyond payments, the paper argues that computing capacity itself is emerging as a distinct asset class. It cites consensus estimates that combined revenue from Amazon Web Services, Microsoft's Intelligent Cloud, and Google Cloud could reach approximately $1.1 trillion by 2030, a 29% compound annual growth rate from 2025 levels.
The paper points to Stripe's August 2026 agreement to acquire OpenRouter, a platform that distributes AI workloads across more than 400 models from over 80 providers, as an early signal that model routing and compute optimization are becoming part of financial infrastructure. Given Stripe's existing footprint across payments, stablecoins, and agentic commerce protocols, BlackRock frames the acquisition as pointing toward a future where agents autonomously source and pay for compute through programmable payment rails, extending the same tokenization logic the paper applies to money into the computing resources that power AI itself.
McKinsey data cited in the paper projects that AI inference will become the largest AI workload by 2030, driven by an addressable user base of enterprises and individual consumers that the paper describes as considerably larger and more fragmented than the concentrated set of AI training market participants.
BlackRock characterized the overall ecosystem as still nascent, noting that agentic payment activity and compute-market liquidity remain limited today, while framing digital assets as an increasingly integral part of AI's economic infrastructure as agentic systems mature.
Gunel Ismayilova
Disclaimer: All materials on this site are for informational purposes only. None of the material should be interpreted as investment advice. Please note that, despite the nature of much of the material created and hosted on this website, HODL FM operates as a media and informational platform, not a provider of financial advisory services. The opinions of authors and other contributors are their own and should not be taken as financial advice. If you require advice, HODL FM strongly recommends contacting a qualified industry professional.
