USDC captures about 99% of x402 AI-agent payments, shifting the fight to Base vs Solana
Coinbase counts 69,000 active agents and 165 million x402 transactions, while Ripple’s XRPL streaming pitch waits on an unscheduled upgrade.
USDC has become the default settlement asset for AI-agent “machine payments” using Coinbase’s x402 standard, with Circle putting its share at 98.8% of agent-driven volume and CEO Jeremy Allaire citing 99.3% in Q2 2026. With the “which coin pays” question largely settled inside this workflow, the competitive edge shifts to which chain becomes the default routing layer for micro-invoices: Base, Solana, and potentially XRPL later.
Key Takeaways
- USDC represented 98.8% of agent-driven transaction volume in Circle’s materials, and Circle CEO Jeremy Allaire separately cited 99.3% share for x402 AI-agent payments in Q2 2026.
- Coinbase said that by late April 2026 the x402 ecosystem reached 69,000 active agents, 165 million transactions, and about $50 million in cumulative volume, with average transfers described as fractions of a cent.
- Chainalysis tracked over 100 million cumulative x402 transactions on Base through Q1 2026, but cautioned that late-2025 spikes were largely meme-coin driven rather than agents buying services.
- Ripple adopted x402 in June 2026 and joined the Machine Payments Protocol on Sept. 17, 2026, but XRPL “streaming payment sessions” depend on a future upgrade that is not yet scheduled.
USDC Becomes the Default Settlement Asset for x402 Agent Payments
The settlement-asset race inside x402 agent payments is already lopsided. Circle’s press material pegged USDC at 98.8% of agent-driven transaction volume, and Circle CEO Jeremy Allaire put the figure at 99.3% for payments made by AI agents using x402 in Q2 2026.
Those two numbers are close enough to tell the same story, even if the packet does not reconcile the measurement window or definition difference. In this specific workflow, competing volatile assets are not fighting for “payment coin” mindshare so much as trying to attach themselves to the rails that move the payments.
Circle’s product posture lines up with that outcome. The company introduced Agent Stack in May 2026, offering policy-controlled wallets, nanopayments, and a marketplace that supports over 900 paid services. The packet also states USDC circulation reached $73.3 billion and USDC transaction volume rose 151% year over year, which is the kind of scale narrative that makes “default stablecoin” feel less like a bet and more like plumbing.
How Big Is x402 Really? Coinbase’s Agent Counts vs Chainalysis’ Meme-Coin Caveat
The headline x402 numbers are big, but they are not clean. Coinbase said that by late April 2026 the x402 ecosystem had 69,000 active agents, 165 million transactions, and around $50 million in cumulative volume. The average transfer size was described as “just a fraction of one cent,” and both sides of the trade were programs.
That mix matters for interpretation. A high transaction count can mean a lot of tiny service purchases, or it can mean a lot of automated activity that is not actually “agents buying services” in the way traders imagine when they hear “machine payments.”
Chainalysis’ caveat is the friction point. It tracked over 100 million cumulative x402 transactions on Base through Q1 2026, but warned that much of the late-2025 surge in x402 transactions was driven by meme-coin activity rather than agents purchasing services. The packet does not quantify how much of the total should be discounted, which leaves the “actual working market” as an open variable even while the settlement asset looks decided.
The practical read is that x402 has proven it can generate machine-to-machine throughput, but the composition of that throughput is still the question that decides whether this is a durable services market or a standard that also got used for other onchain behaviors.
Why Stable Value Wins Micro-Invoices: How x402 Uses HTTP 402 to Trigger Payments
x402 is a machine-payments standard built on HTTP 402 Payment Required, a web status code that signals a payment is needed before a server will fulfill a request. The flow described in the packet is mechanical: a server returns a 402 with payment instructions, the agent reads those instructions, signs a stablecoin transaction, includes proof of payment, and retries the request.
That design pushes the system toward stablecoins because the unit prices are microscopic and the cadence is high. The packet’s example is an API call priced at $0.0004 that needs to hold constant even a minute later. If the payment asset moves, every invoice becomes a volatility trade, and the agent cannot reliably budget or price services.
Coinbase effectively baked that conclusion into the spec early. The packet says Coinbase integrated stablecoin functionality into the x402 reference implementation “before any developer began working with the specification,” which is another way of saying the default developer path assumed stable value from day one.
The timeline also explains why the standard is now bigger than a single vendor’s product. Coinbase created and open-sourced x402 in May 2025, announced a foundation with Cloudflare in September 2025, and transferred governance to the Linux Foundation in April 2026. By the time governance moved, five deployments were operational, including Coinbase’s Agent.market, Stripe Machine Payments, CoinGecko’s paid endpoints, and the Cloudflare Agents SDK.
The New Battleground: Base vs Solana as the Routing Layer While USDC Settles
Once USDC is the settlement unit, Base and Solana compete as routing and execution venues, not as “payment coins.” The packet’s description of agent stacks is blunt: flows typically default to Base for Ethereum-compatible tasks and Solana for high-throughput cases, while the payment still settles in USDC.
The performance claims in the packet frame the trade. Solana confirmations are described as about 0.5 seconds versus roughly 2 seconds on Base, with both chains running at very low gas costs. Ethereum mainnet is supported, but described as too costly for sub-dollar transactions, which effectively disqualifies it from the sub-cent micro-invoice use case that x402 is targeting.
What stands out here is how much power sits in defaults. If the reference client and the most common agent frameworks route to Base unless they need Solana’s throughput, then Base becomes the “normal” path for machine invoices even if Solana is technically faster. If the default flips, the habit flips, and the economics follow the habit.
That is also why the competitive surface shifts away from token narratives. Solana can win routing share without SOL being the invoice unit, and Base can win routing share without ETH being the invoice unit. The value accrues through fees, developer mindshare, and integration gravity, not through being the denomination.
XRP Ledger’s Pitch Is Streaming Sessions—But the Upgrade Isn’t on the Calendar Yet
Ripple is positioning XRP Ledger as the session layer rather than the default invoice currency. The packet says Ripple adopted x402 in June 2026 and joined the Machine Payments Protocol on Sept. 17, 2026, framing XRPL’s differentiator as “streaming payment sessions,” meaning value is sent continuously over time during an ongoing session rather than as a one-off transfer.
The catch is that the differentiator is not fully live. The packet says RLUSD can handle one-off payments on XRP Ledger, but session streaming requires a future Ripple upgrade that has not been scheduled. It also notes XRP’s share of agent volume is “not measurable yet,” which makes the XRPL angle more about optionality than current flow.
That puts XRPL in a different competitive lane than Base and Solana. Base and Solana are described as already carrying routing activity for x402-style payments, while XRPL is trying to define a feature category that could matter if agents start paying for time-based services where continuous settlement is cleaner than per-call billing.
Trader’s Read: If the Coin Is Decided, Watch the Default Chain and the Standards Body
The part that decides this trade is not whether USDC is “winning” agent payments. It already is, at 98.8% by Circle’s materials and 99.3% by Allaire’s Q2 2026 statement. The part that decides the next leg is whether x402’s reference implementations and the Linux Foundation-governed standard harden around a default routing chain, because defaults become the path of least resistance for thousands of integrations.
The threshold that matters is whether the ecosystem starts publishing clearer composition data. Coinbase’s late-April 2026 snapshot (69,000 active agents, 165 million transactions, ~$50 million cumulative volume) reads like scale, but Chainalysis’ warning that late-2025 spikes were meme-coin driven means raw counts are a noisy proxy for “agents buying services.” If future disclosures separate service payments from other activity, that would turn x402 from a throughput story into a demand story.
Circle’s Agent Stack is the other lever that can turn dominance into lock-in. It launched in May 2026 with policy-controlled wallets, nanopayments, and a marketplace supporting 900+ paid services. If that marketplace grows and Circle updates the agent-payment share beyond Q2 2026, the settlement layer becomes less contestable and the routing layer becomes the only place left to compete.
Ripple’s XRPL pitch is the cleanest example of timeline risk. Joining x402 and the Machine Payments Protocol gets XRPL into the standards conversation, but “streaming payment sessions” remain gated by an upgrade that is not scheduled. If Ripple puts that upgrade on a dated roadmap, XRPL becomes a real routing candidate for time-based services. If it stays unscheduled, the market will keep routing around it, and USDC will keep settling on whichever chain the reference client makes easiest.
What USDC dominates AI agent crypto payments Tells Me
I read the 98.8% versus 99.3% split less as a discrepancy and more as a sign that the measurement window is now the only thing left to argue about. In a market where the average transfer is “a fraction of one cent,” the asset that removes denominator risk wins by default, and x402’s reference path was built around stablecoin settlement early.
The real test is whether the routing layer consolidates the same way the settlement layer did. If the Linux Foundation-governed x402 maintainers and the major deployments converge on Base-first defaults for most EVM-compatible work, then Base becomes the habitual micro-invoice lane and Solana’s role stays specialized around throughput-heavy bursts. If Solana becomes the default for more stacks because the 0.5-second confirmation profile is operationally cleaner, then Base’s advantage narrows to “Ethereum compatibility” rather than being the assumed path.
XRPL is the conditional branch. If Ripple schedules and ships the upgrade required for session streaming, XRPL can compete on a payment primitive that is not just “faster” but structurally different for time-metered services. If that upgrade remains off-calendar, then the agent-payments story stays a two-chain routing contest with USDC as the settled unit, and the only number that will matter is which reference client default becomes the habit for millions of daily micro-invoices.
