Base Hits 100 Million AI Agent Payments — and Armstrong Says Bots Will Dominate

Coinbase CEO Brian Armstrong isn’t buying the idea that crypto startups should drop blockchain and run toward AI.

On July 27, he pushed back hard on that advice. His argument: the two technologies complement each other rather than compete.

Positioning crypto as payment infrastructure for AI isn’t a pivot, in his view — it’s an acknowledgment of where adoption is already going.

Armstrong expects AI agents to become the dominant users of crypto rails.

He offered a milestone as evidence. Base, Coinbase’s Layer-2 network, has now processed over 100 million payments initiated by AI agents.

Why Armstrong Says The Pivot Crowd Has It Backwards On Crypto As AI

The argument Armstrong made, reported by CoinDesk on July 27, targets a wave of crypto founders who have quietly rebranded their projects as AI companies over the past 18 months.

His case is that these founders are misreading the causal direction. Understanding crypto as AI infrastructure, he argues, is the honest read — not a marketing angle.

Armstrong framed blockchain as fundamental infrastructure for future automation, not a legacy technology to escape.

The reason is practical. AI agents need to transact autonomously, instantly, and across borders, without waiting for banking hours or human approval.

Traditional payment rails cannot support that workflow. Cryptocurrency networks, by design, settle transactions without intermediaries and without sleep.

Seeing crypto as AI payment rails is not a rebranding exercise — it reflects how agents are already operating in production environments.

The 100 million figure is not a prediction. It reflects payments that have already cleared on Base, Coinbase’s own Layer-2 network built on Ethereum (ETH).

A Layer-2 network processes transactions off the main Ethereum (ETH) chain to reduce fees and increase speed, then posts compressed proofs of those transactions back to Ethereum for final settlement. That architecture lets Base handle the kind of high-frequency, low-value micropayments that AI agents generate when they pay for API calls, data, or compute services.

What Agentic Finance Actually Looks Like

The term “agentic finance” describes a model where software agents, not humans, originate financial transactions.

An AI assistant that independently books a hotel, pays a vendor, or buys compute time from a rival service is engaging in agentic finance. Each action requires a payment that is authorized and settled without a person clicking “confirm.” This is precisely where crypto as AI infrastructure proves its value over legacy systems.

Traditional banking cannot serve that use case cleanly.

ACH transfers in the US take one to three business days. Credit card networks require merchant accounts and charge interchange fees that erode the economics of sub-dollar transactions.

Stablecoins on low-fee blockchains, by contrast, settle in seconds for fractions of a cent. Base’s 100 million figure suggests that AI developers are already routing agent payments through exactly that infrastructure.

Armstrong did not frame this as speculative.

Coinbase disclosed the milestone alongside his comments, grounding the thesis in measurable traction rather than roadmap promises. The data reinforces the crypto as AI narrative with concrete transaction volume rather than projection.

From Coinbase Curiosity To A Contested Industry Argument

Coinbase’s interest in AI-agent payments predates this week’s comments.

The exchange began routing stablecoin settlement infrastructure through Base in 2024 and has steadily expanded the network’s developer tooling. The 100 million payment milestone represents compounding adoption, not a single event. Coinbase (COIN) has been positioning this crypto as AI payments stack for well over a year before Armstrong’s public remarks.

The pivot Armstrong is criticizing has been visible for months.

Projects with blockchain-native architectures have added “AI” to their names, white papers, and investor decks, often without changing their core product. Armstrong’s argument is that this rebranding reflects a misunderstanding: the AI economy will need crypto more, not less, as agents multiply.

Genuine crypto as AI integration means building payment settlement into agent workflows, not swapping a ticker suffix.

The counter-argument, held by some in the industry, is that AI companies have no ideological attachment to decentralization. If a centralized API can settle micro-transactions cheaply enough, agents will use it.

Armstrong’s bet is that no centralized solution will match the censorship-resistance, global reach, and programmability of public blockchains for the specific demands of autonomous software.

Also Read: Ripple Backs AI Compliance Push as XRP Ledger Hits 10M Agent Transactions

The Federal Reserve Decision Adds Urgency

Armstrong’s comments landed on a day when cryptocurrency markets were already absorbing macro signals. Bitcoin (BTC) held near $65,000 as traders positioned ahead of a Federal Reserve interest-rate decision, and Ethereum outperformed BTC as DeFi activity ticked higher with the easing of Strait of Hormuz tensions.

That broader market backdrop matters for the agentic finance thesis. A Fed rate cut would lower the opportunity cost of holding stablecoins rather than yield-bearing dollar instruments, potentially accelerating stablecoin adoption by AI agent developers who currently balance treasury yield against payment convenience.

Armstrong’s core point cuts across the rate cycle.

Whether the Fed cuts or holds, the structural case for crypto as AI payment infrastructure rests on settlement speed and programmability, not yield. The 100 million payments on Base suggest developers are already acting on that logic.

Read Next: Can Bitcoin ETFs Reclaim The Flow Crown?

Ethereum Has Won Three Weeks Running