Coinbase CEO Bets on “Agentic Finance” as AI Drives 100M Payments on Base
———————————————————————
Coinbase is doubling down on the idea that artificial intelligence will *supercharge* crypto adoption rather than replace it.
Chief executive Brian Armstrong argues that AI agents – autonomous software programs that can make decisions and transact on their own – will need native digital money and open financial rails. In his view, that puts crypto at the center of the next wave of financial innovation.
Armstrong points to three pillars of this vision: the Base network, the USDC stablecoin, and Coinbase’s x402 payment protocol. Together, he says, they form an early version of “agentic finance,” or AiFi – an economy where machines routinely pay each other without human involvement.
“AI makes crypto more important, not less”
Responding to the growing narrative that crypto companies should pivot into AI, Armstrong pushed back, insisting that AI and blockchain are complementary technologies.
He argues that as AI becomes a “megatrend,” intelligent agents will need programmable, always-on money, not legacy banking infrastructure with business hours, intermediaries and manual approvals. That, he says, is where crypto excels.
In public comments, Armstrong stressed that AI agents will require:
– Global, internet-native payment rails,
– Assets that can be transferred via code,
– And settlement systems that do not depend on traditional banking “pipes.”
From his perspective, these requirements naturally align with crypto networks and stablecoins rather than card networks, bank wires, or closed fintech APIs.
Base crosses 100 million AI-driven payments
Armstrong’s thesis is not just theoretical. Agentic activity on Coinbase’s Base network has already crossed a significant milestone.
According to blockchain analytics from Chainalysis, more than 100 million agentic payments have been executed on Base via the x402 protocol within roughly nine months of activity. The firm identified x402-related flows directly onchain and noted that transactions with a value of at least 1 dollar represented about 95% of the total value transferred.
These figures suggest that AI and other autonomous agents are not just experimenting with onchain payments, but beginning to use them at real scale for digital services and machine-to-machine commerce.
Chainalysis found agentic payment wallets on Base tend to:
– Be relatively new addresses,
– Hold a wider variety of assets,
– Maintain smaller balances than the average Base user.
That profile is consistent with wallets controlled by software rather than long-term human investors – frequently funded just enough to complete specific tasks or workloads.
How Coinbase’s AiFi stack was assembled
Coinbase did not originally build its infrastructure with AI agents as the sole target, but the pieces have come together as a coherent stack for agentic finance:
1. Base
Launched in 2023, Base is Coinbase’s layer-2 network built on Ethereum. It is designed to make onchain applications:
– Faster,
– Cheaper,
– And more user-friendly.
Base was conceived as a general-purpose scaling solution, enabling everything from DeFi and gaming to onchain identity. The emergence of AI agents as users of the network is a second-order effect: low fees and reliable throughput make it suitable for high-frequency, automated transactions that agents generate.
2. x402 protocol
Roughly two years after the Base launch, Coinbase introduced x402 – a payments protocol inspired by the HTTP “402 Payment Required” status code. That long-unused internet standard was originally envisioned to support pay-per-use web resources. Coinbase’s x402 revives that idea for the AI era.
x402 lets software applications, including AI agents, automatically:
– Request payment for access to APIs, data, and compute,
– Settle those payments with stablecoins,
– And complete the process without human checkout flows, passwords, or accounts at a traditional financial institution.
In practice, x402 turns payments into a background function of software, similar to how HTTP turned data retrieval into a simple, standardized action.
3. USDC stablecoin
USDC, a dollar-pegged stablecoin launched in 2018 by Circle in partnership with a Coinbase-backed consortium, is one of the core assets used for x402 payments.
For AI agents, USDC offers:
– Price stability relative to the US dollar,
– Instant onchain settlement,
– And programmability via smart contracts and APIs.
That combination makes it suitable for automated, high-volume microtransactions where volatile assets like Bitcoin or Ether would introduce unnecessary complexity or risk.
Together, Base, x402, and USDC form Coinbase’s current “AiFi stack”: network, protocol, and asset working in concert to enable machine-to-machine payments at scale.
From human-first to machine-first finance
Traditional financial infrastructure is designed for people: forms, identity checks, manual approvals, customer support. AI agents, by contrast, need something closer to a software API: deterministic, permissionless, and available 24/7.
Agentic finance attempts to reorient financial rails around those needs. Under Armstrong’s vision, AI agents will:
– Buy data feeds or access premium APIs in real time,
– Spin up or shut down cloud compute resources as needed,
– Subscribe to digital services and manage those subscriptions,
– Negotiate prices and settle bills directly in code,
– All while holding and spending digital assets without a human clicking “pay.”
Crypto rails allow these behaviors because ownership, authorization, and settlement can all be expressed as code on a blockchain rather than through an offline contract and a human-operated payment process.
Why Base is attractive to AI agents
For agents performing frequent, low-value tasks, several characteristics of Base matter:
– Low transaction costs: Microtransactions become viable when fees are a fraction of a cent or a small percentage of payment size.
– Predictable performance: AI workflows rely on consistent confirmation times to avoid cascading delays.
– Open access: Agents do not need to “ask permission” or onboard as a traditional customer; they only need an address and funds.
By hosting x402 on Base, Coinbase effectively gives AI developers an L2 chain tailored to automated operations: fast, cheap, and programmable.
Business impact: letting software become a customer
One immediate implication of agentic finance is how it changes what it means to have a “customer.”
With x402 and USDC on Base, a business can effectively:
– Treat an AI agent as a paying client,
– Expose metered APIs or services that agents can consume,
– Accept stablecoin payments automatically as those services are used.
That model:
– Reduces friction for human users (no more manual billing for every micro call),
– Creates new pricing options (true pay-per-use, dynamic pricing),
– And opens up markets where humans are not involved at all, such as machine-to-machine data marketplaces.
For example, an AI agent tasked with monitoring markets could:
– Purchase streaming data from multiple providers,
– Continuously evaluate which source offers best value,
– And route payments automatically to whichever service it chooses, all settled in USDC on Base.
Competitive and ecosystem context
Coinbase is not alone in targeting AI-related payments, but its approach is distinct in several ways:
– It leans heavily on existing crypto primitives (stablecoins, L2 networks, HTTP standards) rather than building an entirely new stack from scratch.
– It focuses on machine-to-machine payments as a *native* use case for crypto, rather than simply adding AI branding to existing products.
– It positions itself as infrastructure – the “pipes” – rather than trying to become an AI application developer.
Other players in the AI and blockchain space are exploring similar themes: tokenized access to compute, decentralized data markets, and onchain identity for agents. The growth of agentic payments on Base gives Coinbase an early data point that this category is not just hypothetical.
Risks, challenges, and open questions
Despite Armstrong’s optimism, agentic finance faces non-trivial hurdles:
– Regulatory treatment of AI-controlled wallets:
It is still unclear how regulators will treat wallets managed by algorithms. Questions around KYC, AML, and liability in case of misuse have not been fully resolved.
– Security and abuse prevention:
Giving AI agents control over money raises the stakes of security. Poorly designed agents could overspend, be exploited, or trigger unintended financial actions.
– Economic viability of microtransactions:
While Base’s fees are low, extremely granular pay-per-use models need to balance infrastructure costs, fraud prevention, and user experience.
– Standardization of protocols:
For agentic finance to scale beyond a single platform, cross-chain standards and interoperable protocols will likely be necessary so agents are not locked into one network or provider.
These uncertainties mean that AiFi is still experimental, even as transaction numbers on Base suggest genuine traction.
Earnings backdrop and investor lens
Coinbase’s agentic finance narrative emerges ahead of its second-quarter earnings report. Market expectations point to:
– Revenue around 1.29 billion dollars,
– Representing an estimated 13.8% decline compared with the same period a year earlier,
– With earnings per share projected to be roughly flat.
In that context, the AiFi story serves a strategic purpose: it frames Coinbase not just as an exchange dependent on trading volumes, but as a broader infrastructure provider for emerging internet-native financial use cases – including AI.
If agentic payments continue to climb from 100 million transactions toward billions, they could one day represent a meaningful stream of onchain activity, fees, and enterprise business for the company.
What this means for developers and enterprises
For builders and businesses, Coinbase’s push into agentic finance translates into several practical opportunities:
– API monetization:
Developers can price digital services, data, and tools on a fine-grained, per-call basis and settle in USDC automatically.
– AI-native products:
Companies building AI agents can give those agents wallets on Base, enabling them to transact, subscribe, and manage resources autonomously.
– Automation of back-office tasks:
Software agents could handle invoices, recurring payments, and vendor settlements onchain, reducing manual reconciliation and delays.
– Global accessibility:
Because Base and USDC operate over the internet, AI agents built anywhere can pay or get paid without integrating with local banks.
As these patterns solidify, “having a wallet” may become as essential for an AI agent as “having an API” is for traditional applications.
The road ahead for agentic finance
The crossing of 100 million agentic payments on Base is likely an early milestone rather than an endpoint. Over time, several trends could shape the next phase of AiFi:
– Integration of more stablecoins and assets, giving agents multi-currency capabilities.
– Development of richer policy layers – letting humans define clear constraints and permissions for what their agents can spend and where.
– Emergence of reputation and credit systems for agents, allowing them to enter more complex financial relationships than simple pay-per-use.
– Greater regulatory clarity around automated wallets and machine-controlled funds.
For now, Armstrong’s core message is simple: AI does not diminish the role of crypto – it creates a new kind of user for it. As AI agents become more capable and more numerous, the demand for programmable, internet-native money could grow in parallel, with Base, USDC, and x402 positioned as early infrastructure for that future.