The investment question is not whether AI and crypto can be mentioned in the same sentence. It is whether autonomous software will generate transactions that existing systems cannot serve efficiently—and, if so, where the economics of those transactions will accrue.

BlackRock’s *The Machine-Native Economy* offers a framework for that question. It considers a future in which AI agents do more than respond to prompts: they select data, procure computing capacity, use external services, and pay for those inputs. If that kind of activity becomes common, payment and compute markets may need to accommodate software as a participant in its own right.

What makes a payment system machine-native?

Most financial infrastructure assumes that a person or institution opens an account, completes identity checks, and initiates or approves a payment. An autonomous agent may instead need to make frequent, low-value purchases through software interfaces, potentially across borders and outside business hours.

Digital assets and stablecoins could be useful in such workflows because they can be transferred programmatically and are not tied to a single domestic banking interface. That is a plausible advantage, not a guaranteed outcome. Card networks, banks, and payment companies may adapt their products for agents, and some transactions may not need a blockchain at all. The report is a thesis about a possible economic structure, not evidence that crypto is already the default payment method for AI agents.

More transactions do not identify the winners

Suppose agent-driven commerce expands. The next question is not answered by the growth in payment volume: which firms or assets capture the value?

A stablecoin payment could benefit its issuer, the application coordinating the transaction, a payment provider, or the settlement network. A network token may capture value only if its design links activity to durable demand—for example, through fees or security mechanisms. Fee abstraction and sponsorship can weaken that link. So can competition among networks and payment rails.

This is why “AI needs payments” is not yet an investment thesis for any particular token. The mechanism connecting usage to asset value has to be demonstrated network by network.

Compute may be bought as a service—but the market is not frictionless

BlackRock also points to computing capacity as something agents may procure. Higher AI workloads could increase demand for GPUs and data-center services, while automated buyers might compare providers and purchase capacity on demand.

But compute is differentiated by performance, latency, geography, reliability, software compatibility, and service terms. A mature market would need comparable listings, dependable supply, proof of delivery, and contracts that agents can interpret and enforce. Demand growth alone does not produce a standardized, liquid exchange.

The useful distinction is between a forecast about what agents could do and evidence about what they currently do. Adoption, payment choice, market infrastructure, and token value capture are separate stages. BlackRock’s report makes those stages worth examining; it does not collapse them into one inevitable outcome.

Sources and related coverage

The primary source is BlackRock Digital Assets Research, *The Machine-Native Economy: How digital assets connect intelligence, commerce, and compute.* Related coverage has appeared in CoinDesk, Cointelegraph, Decrypt and FinTech Weekly. These reports are supplementary; the argument here is anchored to BlackRock’s original paper.

Disclosure

This article is independent commentary based on public materials and does not constitute a recommendation to buy or sell any digital asset.