Coinbase’s chief executive framed the rise of machine-to-machine payments as evidence that artificial intelligence and crypto are complementary rather than competing technologies. Activity on Base and the x402 protocol has crossed major usage thresholds, giving agentic finance a measurable on-chain footprint.
The development matters because it points to utility-driven demand for programmable stablecoin payments. Instead of relying only on speculative trading, Base is now being used for software-initiated payments, making AI agents a new source of transaction activity for crypto infrastructure.
"If you're in crypto, pivot to AI."
I used to hear versions of this, and it's the wrong way to think about the world. It's zero sum, scarcity thinking.
Crypto is a general purpose technology. It's infrastructure, the same way electricity or the internet is infrastructure. It…
— Brian Armstrong (@brian_armstrong) July 26, 2026
Base and x402 Anchor the Agent Payment Stack
The activity surge is concentrated on Base, with USDC serving as the primary settlement currency. Base has processed more than 100 million AI-linked x402 payments, positioning Coinbase’s Layer-2 network as the main venue for early agentic payment experimentation.
Protocol metrics cited by market participants show x402 handling roughly 197 million cumulative payments totaling $52.48 million. The system also recorded about 845,800 AI-agent buyers and 250,000 sellers, suggesting a broadening network of software-driven payment participants.
Recent 30-day throughput figures reached $24.24 million in May 2026. That value remains small compared with major crypto trading markets, but it shows machine-originated payments are moving beyond demonstration traffic into economically meaningful settlement.
x402 revives the semantics of HTTP’s “402 Payment Required” response for stablecoin payments between applications, services and agents. The protocol allows software to request a resource, receive payment terms and settle with USDC, creating a native payment rail for automated internet services.
Coinbase’s broader stack includes Coinbase for Agents, launched in June 2026, which lets AI systems connect to user accounts under constrained permissions. The product is designed to support spot trades, derivatives and payments inside isolated portfolios, making programmable authorization central to agent-based financial activity.
The company has said additional safeguards will include spending caps, trade limits and service access controls. Agent transactions are also routed through the same transaction-monitoring and Know Your Transaction systems used for other activity, placing compliance telemetry inside the autonomous-payment workflow.
Machine Payments Create New Control Challenges
The x402 ecosystem now has governance support through a foundation established with the Linux Foundation and backing from firms including Cloudflare and Stripe. That support signals an ambition to move agent payments beyond Base into broader internet and fintech infrastructure.
Cloudflare’s involvement points toward distribution across web services, while Stripe’s participation shows fintech interest in stablecoin rails for autonomous agents. Together, those partnerships suggest machine payments are becoming a payments-infrastructure race rather than a Coinbase-only experiment.
Still, the adoption data requires careful interpretation. Chainalysis has noted that early x402 activity included meme-coin farming and other incentive-driven behavior, meaning transaction counts alone do not prove durable demand from independent AI agents.
The security risks are also clear. Giving code the ability to move money introduces attack vectors such as compromised agent credentials, weak spending policies, insufficient identity layers and faulty payment logic, making authorization design the core risk-management problem.
Developers must also solve reputation and memory challenges for persistent agents. If autonomous software is expected to transact repeatedly across services, counterparties will need reliable ways to evaluate agent history, permissions and trustworthiness.
APIs, data providers, compute platforms and content services could charge per request or per action, but that model increases dependence on protocol-level safeguards and payment-verification infrastructure.
The next challenge is adapting monitoring systems to non-human actors. Identity, source-of-funds checks, transaction limits, counterparty screening and forensic tracing will need to account for agents acting under delegated authority, creating a new category of machine-originated financial risk.
The broader signal is that stablecoins may become the default settlement layer for AI-driven commerce. Whether that becomes a durable market will depend on real usage, secure permissions, reliable compliance controls and the ability to distinguish genuine agent demand from manufactured transaction volume.
