Franklin Templeton: Agentic AI Is Crypto's Killer Use Case for Blockchain Rails
Franklin Templeton's Sandy Kaul publicly argued that agentic AI is the killer use case for blockchain and crypto, that legacy payment rails are structurally unfit for machine-to-machine micropayments, and that most investors are missing the payment infrastructure layer entirely.
A $2 trillion asset manager just told the market it's positioned for the wrong layer of the AI trade.
Key takeaways
- Franklin Templeton's Sandy Kaul published an essay on July 21 arguing that agentic AI is the "killer use case" for blockchain and crypto, citing legacy payment networks as structurally incompatible with machine-to-machine micropayments.
- Kaul contends that investors buying Nvidia and cloud stocks are missing the payment and settlement infrastructure layer where agentic commerce value will actually accrue, with some estimates placing that market at trillions of dollars by 2030.
- The chains Kaul identifies as front-runners all have identifiable, legally addressable validator sets, which means the permissionless argument that makes blockchain suitable for autonomous agents applies most cleanly to a rail Kaul does not mention.
Sandy Kaul, Head of Digital Assets and Innovation at Franklin Templeton, published a lengthy essay on July 21 arguing that autonomous AI software agents will require blockchain payment infrastructure to function at scale, and that the current investor playbook of buying semiconductor and cloud stocks will fail to capture the next layer of value creation. Franklin Templeton manages approximately $2 trillion in assets, which makes this less a blog post and more a shot across the bow at every TradFi allocator still treating crypto as a speculative side pocket.
The essay, originally published in Kaul's LinkedIn newsletter "Revolution, Not Evolution" and cross-posted to Franklin Templeton's website, frames the problem in structural terms: AI agents capable of spending money autonomously cannot use legacy payment networks because Visa settlement takes one to three business days and transaction fees make micropayments economically prohibitive. Blockchain networks settle in seconds. That gap, Kaul argues, is the architectural prerequisite for a functioning agentic economy.
The KYC Wall Nobody Is Talking About
Kaul names the speed problem explicitly. The identity problem is just as important and gets less attention.
AI agents cannot open bank accounts. They have no legal identity. They cannot pass KYC or AML checks. The entire correspondent banking infrastructure, and every payment network layered on top of it, is built on the assumption that a human being with a legal identity sits at one end of every transaction.
Autonomous agents running software do not meet that assumption.
"Agentic AI will likely need to rely on crypto technologies and blockchains to enable their activities as these rails are ideally suited for these use cases," Kaul writes. The framing validates what Bitcoiners have argued for years: permissionless money is not a niche ideology, it is an architectural requirement for any payment system that doesn't require human sign-off at every step.
Kaul also challenges the prevailing investment thesis directly: "To capture the AI growth opportunity today, most investors buy shares of AI-aligned companies and related verticals. But will the same playbook work for agentic AI?"
The emerging x402 protocol and agentic payment infrastructure more broadly are already surfacing this question in live markets. Cloudflare's x402 payment gateway opened its waitlist on USDC, not Bitcoin, which illustrates exactly the fork in the road Kaul is gesturing at without fully mapping.
The Rails Problem the Essay Doesn't Solve
Here is where the analysis earns out. Kaul's argument is directionally correct and institutionally significant. The specific chains identified as front-runners for the agentic economy are not.
Aptos, Solana, and BNB Chain all settle in seconds. They are all also operated by identifiable validator sets that are legally addressable by regulators and can be pressured by the foundations that control protocol upgrades. A regulator who wants to freeze an AI agent's payments on Solana has a list of people to call.
The OFAC sanctions on Tornado Cash show exactly how that plays out. Speed without permissionlessness is just a faster version of the same counterparty risk.
Bitcoin's base layer and Lightning Network have no such list. A Lightning-enabled agent spending sats is structurally different from a Solana-enabled one. The payment cannot be frozen by contacting a foundation. There is no issuer to blacklist a channel.
Kaul's architectural argument, applied rigorously, points at Lightning. The essay does not go there.
The falsifiable version of this thesis: if Solana or BNB Chain scale materially as agentic payment rails over the next 24 months without a significant censorship event, validator collusion, or regulatory intervention, the permissionlessness premium weakens. If Lightning-attributable agent payment volume remains negligible at the same horizon, the Bitcoin-as-the-correct-rail argument is directionally right but operationally premature.
What to Watch
Franklin Templeton's compliance and communications teams approved this essay, which shifts what other $1 trillion-plus asset managers are allowed to say publicly about crypto as AI infrastructure. Watch for allocation disclosures and follow-on research from other major asset managers in the next 60 to 90 days. The more interesting signal will be whether any of them name Lightning.
Sources
Frequently Asked Questions
AI agents have no legal identity and cannot pass KYC or AML checks required to open accounts with legacy payment providers. Beyond the identity barrier, Visa and Mastercard settlement takes one to three business days and carries transaction fees that make machine-speed micropayments economically unworkable. Blockchain rails solve both the speed problem and, on permissionless networks, the identity problem.
Kaul's essay does not mention Bitcoin or Lightning. The chains named as front-runners have identifiable validator sets that are legally addressable, which means they carry censorship risk that permissionless rails do not. The architectural argument Kaul makes, that agentic AI needs rails that don't require human identity attestation, applies most completely to Bitcoin's Lightning Network, the only globally reachable payment rail with no issuer, no foundation, and no validator set a regulator can pressure.
A chatbot responds to prompts. An agentic AI takes actions autonomously on behalf of a user or another system, including spending money, booking services, or executing multi-step tasks without human approval at each step. The payment problem emerges precisely because these agents act without a human in the loop, which breaks every assumption the existing financial system makes about transaction authorization.
