A $1.8 trillion asset manager has a message for investors who think owning Nvidia shares covers their AI exposure: it doesn't. In a paper published this week, Sandy Kaul, Head of Digital Assets and Innovation at Franklin Templeton, argued that agentic AI — software that acts, pays, and decides on a user's behalf without checking in at every step — will run on crypto rails rather than Wall Street ones, and that blockchain adoption's long-awaited “killer” use case has arrived.

Why Agents Need Blockchains

The argument turns on what AI agents actually do. Unlike a chatbot, an agent can shop, book, and pay autonomously once a user grants it permissions — and that capability creates a settlement problem traditional payment infrastructure wasn't built for.

Franklin Templeton's paper lays out the mechanics. Payment agents would generate single-use tokens to make purchases and settle transactions, with each token embedding its own rule set: which merchants can accept it, the maximum spend per transaction, and how long it stays valid. Once the purchase completes, the token automatically invalidates. Blockchains can hold, send, and receive those tokens while enforcing their embedded rules — exactly as they already do for smart contracts. The firm also points to cryptographically verifiable agent identity, with each token carrying credentials required to sign transactions.

The Investment Thesis: Own the Tokens, Not Just the Equities

This is where Kaul's conclusion gets pointed. Most investors capture AI growth by buying shares of AI-aligned companies — a playbook that has worked well, with the S&P 500 climbing roughly 20% over the past year to a record high in early June, driven largely by tech and AI names. But Kaul argues that same approach may not capture agentic AI.

The reasoning is mechanical rather than speculative. Recording a transaction on a given chain requires paying fees in that chain's native token — SOL on Solana, ETH on Ethereum. If agents buy data, compute, and API access across blockchains at scale, demand for those tokens follows transaction volume rather than corporate earnings. Kaul's stated view is that capturing the value of decentralized networks and businesses will increasingly require owning the cryptocurrencies and altcoins those entities issue, and that such holdings are likely to become key portfolio positions for anyone targeting the agentic AI opportunity.

You might also like: Clarity Act Explained: The Five Provisions That Would Rewrite the Rules for US Crypto

The Infrastructure Is Already Being Built

The thesis isn't purely forward-looking. The x402 Foundation — 40 organizations including Visa, Mastercard, and AWS building open payment rails for AI — launched formally on July 14. The protocol revives a forgotten HTTP status code from 1991, originally reserved for web payments that never materialized, and repurposes it to let software pay software directly over the internet.

The scale being projected is substantial: estimates cited alongside the paper suggest agentic commerce could reach $3 to $5 trillion by 2030, with initial demand likely coming from machine-to-machine micropayments between enterprise software systems. Franklin Templeton notes Ethereum and its layer-2 networks currently stand as the industry standard, with the largest developer base and the deepest institutional support.

The Caveat Worth Holding

The thesis is a forecast, not a fact — and it comes from a firm with a digital assets division and a commercial interest in the conclusion it reaches. Agentic commerce at trillion-dollar scale remains a projection, the x402 standard is days old, and nothing guarantees that agents settle on public chains rather than permissioned corporate rails.

What makes the paper notable isn't certainty but positioning: one of the world's largest traditional asset managers is telling clients that the AI trade and the crypto trade may be converging — and that the conventional way of owning AI might miss the part that matters most.

Source: Based on Franklin Templeton’s published paper by Sandy Kaul and reporting via Decrypt and CryptoPotato. This article is for informational purposes only and is not investment advice; projections described are forecasts and may not materialize.