AI agents need money rails with subsecond finality, near-zero fees, global programmability, and no human-in-the-loop signups, so the argument lands on X42-style stablecoin payments rather than banks. Solana is presented as the current winner for agent payments on speed, finality, liquidity, and fee structure, but even SOL only covers 18.43% of the base-case load by September 2027. The core claim is that agents will choose permissionless public chains, not Bitcoin, Ethereum, or private enterprise chains, because private chains trap fees and public chains monetize scarcity and usage. The biggest unresolved question is whether SOL can scale fast enough to absorb the projected explosion in on-chain agent payments.
AI agents are already usable for grocery runs, lease negotiation, bank-balance yield shopping, and canceling Netflix.
X42 lets an agent sign a stablecoin payment without traditional rails; listed adopters include AXP, AMZN, CRCL, SOL, MONAD, STRIPE, V, NEAR, COIN, XLM, XRP, SHOP, GOOGL, and MA.
BlackRock’s “machine native economy” framing links digital assets, intelligence, commerce, and compute, and is treated as an underpriced crypto-demand driver.
Agents need API and compute calls plus SaaS accounts, so payments must finalize in subsecond time, with fees below a microcent and no human sign-up or approval loop.
September 2027 conservative case: 5M agents × 100 transactions/day × 20% on-chain = 100M extra on-chain transactions/day.
September 2027 base case: 25M agents × 200 payments/day × 30% on-chain = 1.5B extra on-chain transactions/day; the math is also restated as 60 on-chain transactions per agent per day.
September 2027 bull case: 150M agents × 400 payments/day × 40% on-chain = 24B on-chain transactions/day.
Using an SCP profiler “yesterday” table, SOL is claimed at 163.687M real transactions/day, versus 50M combined for the other 10 chains, or 76.5% of the table.
Against the 1.5B base-case load, SOL is said to have 276.5M/day capacity, or 18.43% of required load; BNB is 209.6 TPS and 18M/day, TRX 138 TPS and 11.9M/day, SUI 123.7 TPS, Base 37.2 TPS, XRP 23.9 TPS, BTC 8.6 TPS, ETH 23.4 TPS, and AVAX 4.6 TPS.
Alpenlow, the Solana Foundation test environment, is claimed to already deliver 150ms finality.
Finality comparison: SOL 0.115s, APT 0.65s, SUI 0.7s, BNB 1.1s, AVAX 2s, TRX 57s, ADA 3m, ETH 12m to 17m, BTC 1h.
Speed multiples versus SOL are given as APT 4x slower, SUI 5x, BNB 7x, AVAX 13x, TRX 38x, ADA 1,200x, ETH nearly 5,000x, and BTC 24,000x slower.
Cost multiples versus SOL are given as Base 2.9x, BNB 10x, AVAX 15x, TRX 17x, ETH 58x, NEAR 68x, and BTC 115x more expensive.
The chain scorecard rejects BTC, ETH L1, and SUI for the agent use case: BTC L2s are too convoluted for USDC micropayments, ETH L1 is “totally cooked,” SUI has no adoption yet, and AVAX subnets fragment liquidity and behave like private chains.
SOL is said to have X42 adoption up 3x in the last week and to handle 76% of all AI-agent transactions; XRP ranks second, Polygon third, and Base fourth.
The forecast cited from Lucanet is that SOL is the only battle-tested platform for AI agents doing absurd weekly volume, with a ceiling estimate of 1 quadrillion AI-agent transactions a week.
The conclusion is that AI agents will use X42, with USDC as the unit of account for now, and that agents will not choose BTC or ETH; private chains lose because fees stay inside the private island and do not accrue to public-chain token holders.