By BitBrainers Editorial
The thesis is clean. AI agents will transact with each other autonomously, at high frequency, for amounts too small for traditional payment rails to handle. An API call, a second of compute, access to a dataset. Fractions of a cent per transaction. Traditional networks charge $0.30 plus 2-3% and settle in one to three business days. That structure breaks for machine-to-machine commerce at scale.
AI agents cannot open bank accounts. KYC requirements built for humans do not map to software. Blockchains do not ask for a passport. That is structural, not incidental.
The numbers she cited are real. Aptos is recording up to 12,933 transactions per second. Solana 6,284. BNB Chain 3,252. Visa's network runs at roughly 1,700 TPS under normal load. The throughput gap is not marginal. If you want exposure to the chains pulling ahead in this race, Kraken lists SOL, ETH, and the major L2 tokens.
The Protocol Nobody Is Stress-Testing
x402 is the HTTP-native payment protocol developed by Coinbase that lets AI agents pay for APIs and services over standard HTTP, using stablecoins, without accounts or subscriptions. Visa, Mastercard and Stripe have all backed it. It has processed $15 million in adjusted volume across 109 million transactions since launch.
Franklin Templeton presents x402 as chain-agnostic. That is technically accurate. The spec is open. The Linux Foundation now maintains it.
The operational reality is different. Solana accounts for an estimated 50 to 80% of all x402 transactions. Base is the dominant EVM settlement layer. Coinbase developed x402. Coinbase runs Base. The protocol is chain-agnostic in its architecture and Coinbase-adjacent in its actual flow capture.
That is not a reason to dismiss the thesis. It is a reason to watch who controls the settlement layer when volumes scale.
Decentralization in Practice
The TPS numbers are peak figures recorded under specific conditions. They are real, but they describe ceiling performance, not operating reality under adversarial load or validator concentration pressure.
High-throughput chains tend toward validator concentration over time. The economics push that direction. Larger validator sets cost more to coordinate and often trade throughput for decentralization. The chains Kaul highlights are fast precisely because they made architectural trade-offs that centralize decision-making in fewer nodes.
Telegram's TON network is the clearest recent example. Non-custodial wallet, strong user numbers, genuine adoption. Telegram is also the network's dominant validator. Users hold keys. Telegram influences whether transactions clear. That is decentralization in architecture and a chokepoint in practice. We covered that dynamic in detail here.
Regulators do not need to attack a blockchain to disrupt it. They reach the domains, the app stores, the fiat on-ramps, and the banking relationships of the entities running the largest validator nodes. The attack surface for agentic payment rails is not the chain itself. It is everything around it. If self-custody matters to you while navigating this environment, Trezor is the standard starting point.
Institutions publish the bull case. We read the footnotes.
Bitcoin, macro, and AI infrastructure. Before the crowd finds it.
SubscribeWhat Franklin Templeton Gets Right
Settlement finality is genuinely structural, not marketing. Machine-speed commerce cannot wait for T+1 or T+2. The moment agents are buying compute in real time, settlement delay is a functional failure, not an inconvenience.
The UX flywheel argument holds. If agent payments become invisible, one of crypto's largest adoption barriers disappears. People adopt applications, not infrastructure. Invisible rails are the only rails that scale.
If agentic commerce grows on-chain, demand for native gas tokens grows with it. That is a direct mechanism, not a speculative narrative.
The Question the Report Does Not Ask
Which chain actually wins? Franklin Templeton lists Aptos, Solana and BNB Chain as high-throughput candidates. It does not argue for one. That is appropriate restraint for an institutional report. It is also the question that determines where the value goes.
x402 adoption patterns suggest Solana and Base are pulling ahead in real transaction flow. Early integrations compound. A chain embedded into the first generation of agentic stacks is difficult to displace even if a superior alternative appears later.
The GENIUS Act and CLARITY Act are moving. Neither directly addresses autonomous software agents transacting on-chain. When regulators write those rules, the chains named in them gain structural advantages over those that are not.
We covered the first major attempt to put AI agents inside a brokerage account and what it got wrong: Robinhood Just Gave AI Agents a Brokerage Account.
On The Radar
Watch x402 facilitator concentration. If one entity controls the dominant settlement path through a chain-agnostic protocol, the protocol's neutrality is theoretical. Track which chains are getting embedded into enterprise AI agent stacks by default, not which chains have the highest peak TPS.
Watch for regulatory language that names AI agents specifically in the context of on-chain financial transactions. The first jurisdiction to write those rules sets the template. The chains and protocols they reference gain legitimacy by inclusion.
Sources
Franklin Templeton — Agentic AI: The Killer Use Case for Blockchain and Crypto
CoinDesk — Forget Nvidia: The Next Big AI Trade Could Be Crypto and Blockchain
Cointelegraph — Agentic AI is Next Killer Use Case for Blockchain: Franklin Templeton
Chainstack — x402 Protocol: Architecture and Payment Flow for AI Agents
This is market commentary, not financial advice. BitBrainers holds Bitcoin and other digital assets. Nothing here is a recommendation to buy or sell any asset. Do your own research.
