Franklin Templeton just told investors to stop staring at NVIDIA and start paying attention to blockchain. Sandy Kaul, the firm’s head of digital assets and innovation, published a thesis arguing that agentic AI — not generative AI — will be the driving force behind the next wave of crypto adoption. Here’s what that argument actually looks like, and where we think it holds up.
What Is Agentic AI, and Why Does It Need Blockchain?
Most people know AI as the thing that writes emails and generates images. That’s generative AI — it responds to prompts. Agentic AI is different. It acts on its own. An AI agent can compare flight prices, buy cloud computing power, negotiate a data subscription, or hire another AI agent to complete a subtask — all without waiting for a human to click confirm.
That distinction matters because it creates a new kind of economic activity: machines paying machines, thousands of times per hour, for fractions of a cent per transaction.
Here’s where the problem starts. Visa’s network settles in one to three business days. Credit card fees can run 2–3% per swipe. When an AI agent needs to pay $0.001 for a single API call, traditional payment rails make zero sense. The fee alone would cost more than the transaction.
Kaul’s argument is that blockchain networks solve this by design. Public chains like Solana, Aptos, and BNB Chain settle transactions in seconds with fees measured in fractions of a cent. They’re programmable, permissionless, and always on — exactly what autonomous software needs to move money at machine speed.
Visa Already Agrees — And Has the Data
This isn’t just Franklin Templeton talking. A week before Kaul’s post, Visa and blockchain analytics firm Artemis published a joint report breaking agentic commerce into two categories: macro-commerce (human-directed purchases like booking travel) and micro-commerce (sub-dollar, machine-to-machine payments).
Their finding: credit cards work fine for the first category, but stablecoins are the natural fit for the second. And micro-commerce is where the volume explosion will happen.
The x402 protocol — an open payment standard developed by Coinbase and Cloudflare — had already processed 165 million transactions and $50 million in volume across 69,000 active AI agents by April 2026. The x402 Foundation now counts Google, Visa, AWS, Circle, and Anthropic among its members.
What we find most telling isn’t the transaction numbers — it’s who is building this. When Visa says stablecoins will power a new category of commerce, and the protocol is backed by the Linux Foundation with 22 institutional members, we’re past the “interesting experiment phase.
Circle Sees One System, Not Two
Circle CEO Jeremy Allaire pushed the thesis even further. In a recent paper, he argued that AI and blockchain aren’t two separate trends converging — they’re one technological shift.
His reasoning: AI is driving the cost of knowledge work toward zero, while blockchain does the same for payments and coordination. When businesses rely on specialized AI agents, those agents become economic actors — buying services, hiring other agents, and exchanging value autonomously. Blockchain provides the identity, settlement, and governance layer.
Allaire also suggested software pricing could shift from subscriptions to pay-per-task models, with AI agents as both buyers and sellers. That increases demand for native tokens, since agents need them to pay network fees.
We think this vision is ambitious, but the plumbing is already being laid. OKX recently launched a beta marketplace where AI agents can find jobs and pay each other, and Coinbase’s Base network already lets AI tools like ChatGPT execute live crypto trades. Robinhood launched its own AI-powered investing tools in May. The infrastructure race is on.
What This Means for the “AI Trade”
Kaul’s core message for investors: the first AI wave was obvious — buy chipmakers and cloud companies. But if the next phase involves autonomous agents transacting across digital economies, the infrastructure powering those transactions becomes investable too. That means blockchain networks, native tokens, and stablecoin settlement rails.
According to a Bain & Company forecast cited by Franklin Templeton, AI agents could account for 15% to 25% of all U.S. e-commerce sales by 2030. If even a fraction of those transactions settle onchain, the demand implications for networks like Solana and Ethereum’s layer-2s could be substantial.
Of course, the gap between vision and reality is still wide. As CoinDesk reported in March, x402’s daily volume was just $28,000 at the time — most of it from testing. The narrative around agentic commerce is currently running ahead of actual adoption.
But that’s also how every major crypto cycle has started. DeFi had minimal usage before it didn’t. NFTs were a joke until they weren’t. The difference this time is that the institutional backing arrived before the mainstream adoption wave — not after.
Where We Stand
We’re not calling this a guaranteed outcome. The technical infrastructure for AI agent security is still maturing, regulatory questions around autonomous payments remain unanswered, and most AI agents today are still glorified chatbots.
But when a $1.8 trillion asset manager, the world’s largest payment network, and the biggest crypto exchange all point in the same direction within the same month, it’s worth asking whether the market is pricing in what’s coming — or whether it’s still stuck watching NVIDIA’s earnings calls.
The smartest investors notice infrastructure before the application layer takes off. Franklin Templeton is betting blockchain is that layer for AI. We think they’re looking in the right direction.
FAQsWhat are AI agents in crypto, and how do they use blockchain?
AI agents are autonomous software programs that complete tasks — like buying data, paying for computing power, or hiring other agents — without human approval at every step. They use blockchain because it offers instant, low-cost settlement and programmable digital wallets that let machines hold and transfer value on their own.
How does the x402 payment protocol work?
x402 uses the HTTP 402 status code to embed stablecoin payments directly into web requests. When an AI agent hits a paid endpoint, the server returns a price, the agent pays in USDC, and retries the request with proof of payment. The whole cycle takes seconds with no human involvement. Coinbase and Cloudflare built and open-sourced it.
Which blockchains are best suited for AI agent payments?
Solana, Base, Aptos, and BNB Chain are the most cited because they offer sub-second finality and near-zero fees. Ethereum mainnet is too expensive for micropayments, but its layer-2 networks like Arbitrum and Base handle high-volume DeFi transactions at much lower cost.
What are the biggest risks of AI agents making autonomous payments?
Liability is the biggest unknown. Current payment dispute rules assume a human made the purchase. If an AI agent makes a bad transaction, it’s unclear who’s responsible — the user, the developer, or the platform. Regulation hasn’t caught up yet, and crypto security challenges add another layer of risk.
Which AI crypto tokens could benefit from the agentic AI trend?
Tokens tied to decentralized computing (Render, Akash), AI coordination networks (Bittensor, Fetch.ai), and high-speed settlement chains (SOL, APT) are positioned around the agentic AI narrative. Our guide to the best AI crypto coins in 2026 covers the strongest candidates and what separates real utility from hype.