See more like this on Google
Share article
Share article
Crypto payments are back on the tape, not because merchants suddenly love QR codes, but because the next "user" might not be human. Coinbase CEO Brian Armstrong is pushing a clean thesis: AI agents will need native internet money, and banks are not built for autonomous software. If that plays out, stablecoins and on-chain settlement stop being a niche and start looking like default rails for machine-to-machine commerce. [1]
That is the setup Armstrong highlighted last week (March 10), arguing AI agents could be the catalyst for the next crypto payments boom. [2] The timing matters. With total crypto market cap hovering around $2.47 trillion and Bitcoin$62,424.01 dominance near 57%, the market is still trading "store of value first." Armstrong is essentially pitching a new demand driver that is transactional, not speculative.
Enjoy articles without ads?
Register for free and get unlimited access to all articles.
The Armstrong bet: payments demand that does not wait for TradFi
The old utility narrative said TradFi and DeFi would converge, slowly, with banks modernizing and blockchains absorbing only the edge cases. Armstrong's version flips that: the rise of AI agents widens the gap between banks and crypto, because both systems were designed for humans, but only one is composable enough for software.
The key idea is simple:
- AI agents will buy things: data, compute, APIs, ad inventory, digital services, even physical fulfillment via intermediaries.
- They will do it frequently, globally, and often in small ticket sizes.
- They need a settlement layer that can be automated, permissioned by code, and integrated into apps without waiting on bank onboarding.
That is not a philosophical argument, it is a product constraint.
Why banks break when the "customer" is software
Banks can absolutely move money digitally, but they do it inside an identity and account framework that assumes a human or a legally registered entity with signers. That creates friction AI agents do not tolerate.
A few mismatches Armstrong is implicitly pointing at:
- Onboarding is human-paced: KYC, forms, account reviews, jurisdiction questions, business purpose checks. Agents operate at API speed.
- Permissions are coarse: Banking credentials were built for people, not granular spending rules like "spend up to $12/day on dataset X unless model accuracy drops."
- Cross-border is still messy: Even when fast, it is fragmented, and agents will route around fragmentation the same way developers route around legacy infrastructure.
- Programmability is bolted on: Banks can offer APIs, but the money itself is not "programmable" in the same way a smart contract wallet can be.
Crypto rails, especially stablecoins, were built with a different default: a wallet can be created instantly, controlled by code, and used globally.
Stablecoins look like the obvious "agent currency"
If agents trigger a payments boom, the base unit is unlikely to be a volatile asset. It is almost certainly stablecoins, because they match the agent use case: predictable unit of account, 24/7 settlement, and easy integration into software.
Armstrong's Coinbase sits directly in that path through USDC$1.0005 distribution, exchange infrastructure, custody, and its on-chain footprint. The bigger point for the market is structural: an agent economy prefers money that behaves like an API. [3]
Where stablecoins fit cleanly:
- Micropayments and metered billing: Pay per query, per token, per inference, per second of compute, per impression.
- Instant settlement for automated vendors: If an agent buys from another agent (or an API marketplace), neither side wants chargeback risk.
- Global access by default: Agents will not care about local bank hours or local transfer schemes.
This is also why "payments" in this framing is not just retail checkout. It is internet commerce plumbing, similar to how cloud billing became the backbone for software businesses.
The missing piece: wallets that look like infrastructure, not UX
The bullish case depends on one unsexy detail: agents need operational wallets that are safer than hot keys and more flexible than today's manual flows.
Expect the stack to center around:
- Smart contract wallets (programmable control): Spending policies, rate limits, allowlists, time locks, multi-party approvals.
- Account abstraction and session keys: Let an agent act continuously without exposing a master key every time it pays.
- Gas management: If every action requires a user to think about gas, adoption stalls. Gas sponsorship, paymasters, and "fees in stablecoin" experiences matter.
- Compliance layers: Not everything will be permissionless. Agent commerce at scale will pull in monitoring, allowlists, and identity attestations.
Coinbase's positioning here is obvious. Base and Coinbase Wallet style primitives (whether branded or white-labeled) are designed to make on-chain activity feel like normal web infrastructure. If Armstrong is right, the winners are the platforms that make "agent pay" a one-line integration. [4]
What could invalidate the thesis
This narrative is clean, but it is not guaranteed. The market should treat it like a trade with clear invalidation levels.
Things that can break it:
- Regulators force "agent identity" into the banking mold: If every agent must be tied to a fully verified legal entity with bank-style controls, banks could retain the edge, and crypto becomes just a backend settlement option.
- Big AI platforms keep payments closed: If the main agent ecosystems (model providers, app stores, enterprise platforms) internalize billing with credits and netting, on-chain payments may be optional, not default.
- Security failures become systemic: Agents with wallets are a new attack surface. If agent-driven exploits and draining events become routine, enterprises will push back hard.
- UX does not actually improve: If using stablecoins remains operationally painful (bridges, fees, chain selection, key management), agents may still settle in fiat behind the scenes.
The near-term tell will be whether agent payment flows show up as repeatable, measurable usage, not just demos. Volume matters, but so does retention.
Market positioning: where to watch for confirmation
Armstrong's "boom" call is ultimately about on-chain payment velocity. For traders and builders, the watchlist is straightforward:
- Stablecoin settlement growth: Are stablecoins gaining share as a transactional medium, not just parked liquidity?
- Agent-friendly wallet standards: Session keys, programmable limits, automated compliance checks.
- On-chain merchant tooling: Subscription billing, metered usage, refunds, dispute resolution equivalents.
- Regulatory catalysts: US stablecoin frameworks and clearer rules around custody, identity, and permitted payment activity.
If those boxes start getting checked, the narrative shifts from "crypto is for humans and speculation" to "crypto is infrastructure for autonomous commerce." That is the kind of demand curve that can quietly turn into a cycle.
Takeaway: Armstrong is betting that AI agents will be the new payments demographic, and they will route around banks by default. The bullish confirmation is simple: stablecoin payments that look like cloud billing, always on, automated, and global. The bearish invalidation is equally simple: regulation, closed platforms, or security failures that keep agents on trad rails. Keep your eyes on stablecoin settlement, wallet automation, and whether agent commerce moves from prototypes to production.
