Investors buying Nvidia or Microsoft to ride the AI wave may be positioning for the wrong phase. That is the argument from Sandy Kaul, Franklin Templeton's Head of Digital Assets and Innovation, who published a piece this week titled "Agentic AI The Killer Use Case for Blockchain and Crypto" on the firm's verified X account.
Her case hinges on a simple infrastructure mismatch. A typical card transaction costs 2-3% plus a flat $0.30 fee. AI agents, by contrast, make payments averaging $0.001 to buy a single second of compute or run a data query. Credit card rails cannot process that economically. Blockchains can, and every transaction on those networks requires the underlying token as a fee. That mechanic, Kaul argues, is why altcoins will end up as portfolio holdings for anyone serious about the agentic AI trade, not just as speculative side bets.
The numbers behind the thesis
Kaul cited external forecasts placing agentic commerce at $3-5 trillion by 2030. She also pointed to survey data showing 38% of organizations expect AI agents to work alongside human teams by 2028. As proof that the infrastructure is already forming, she referenced Coinbase's x402 payment standard, now under the Linux Foundation, and a Machine Payments Protocol jointly developed by Stripe and Visa. The phrase she uses: "software can pay software." Solana's SOL token appeared in the piece as an example of a network fee asset, though no specific token was recommended.
What this actually changes for crypto investors
Franklin Templeton manages roughly $1.5 trillion in assets. When its digital assets head publishes a first-party market view arguing that altcoins are structurally necessary for capturing AI value, that is a different signal than a crypto-native fund making the same point. The audience here is traditional allocators who currently have no crypto exposure at all.
The piece carries heavy risk disclosures. Franklin Templeton explicitly notes that "concepts discussed may not come to pass" and that crypto investments carry the risk of total loss. So this is a directional thesis, not a research finding, and Kaul is careful to frame it that way. Still, the underlying mechanical argument, that legacy payment infrastructure is simply incompatible with machine-speed micropayments at fractions of a cent, is hard to dismiss on technical grounds alone.
This article is for informational purposes only and does not constitute financial advice or a recommendation to buy any asset.
