XRP Ledger's 13.5 million AI-agent payments have settled 12,010 RLUSD and 6,982 XRP

Spread across every transaction on the XRPL AI Hub, the cumulative settlements of 12,010 RLUSD and 6,982 XRP ^[3] come to about 0.00089 RLUSD and 0.00052 XRP per payment, or less than one RLUSD for every thousand agent payments ^[11]. Each average is a floor for its token, because each total is divided by all 13,489,468 transactions ^[2], including those settled in the other token. On an even split across the 168 registered merchants ^[2], each would have handled about 80,000 payments ^[15] and collected roughly 71 RLUSD and 42 XRP ^[12].

Akinyele's October 10 update gave counts of transactions and merchants ^[1]. The settlement totals come from the dashboard ^[3]. The crypto.news report that carried both noted that counts of payments and tokens transferred do not establish revenue ^[9].

The index is operated by t54 and tracks payments associated with t54's own x402 infrastructure ^[5]. So the count covers the XRP Ledger's agent payments, or rather the ones that run through t54. Ripple-backed t54.ai launched the hub before the ledger's first million x402 payments on July 8 ^[6]. Its co-founder, former Ripple product lead Chandler Fang, expects more real merchants to put services on the ledger and settle through t54's facilitator ^[18].

On the count alone, growth is fast. About 1 million payments on July 8 ^[6] became more than 13 million on October 10 ^[1], an increase of 12 million in 94 days, or roughly 128,000 a day ^[13]. "100M might actually come sooner than we expect!" Akinyele wrote ^[17]. His "sooner" depends on the daily rate rising, because at the July-to-October average the remaining 86.5 million or so would take about 678 days ^[14].

The gap between count and cash has more than one explanation. One is developer traffic: Ripple's AI Starter Kit, released in June, supports automated x402 payments in XRP and RLUSD ^[7], and a developer testing an integration generates payments without generating a business. Fang's version is another, in which merchants move paid APIs, data subscriptions and model inference onto the rail, the uses Ripple's documentation describes ^[8], and settled value begins to rise faster than the count. A third is that very small per-call prices are intended in a protocol built for software agents paying without a person approving each charge ^[10]. In that case revenue would show up only at counts many times today's.

We think the 13 million is real use of the payment rail and, at roughly 71 RLUSD per merchant on an even split ^[12], not yet a revenue line for the businesses accepting it. The view is wrong if settled RLUSD on the hub starts growing faster than the transaction count. Value growing faster than volume would mean agents are paying more per call, the first evidence of the merchants Fang describes ^[18].