Will agentic commerce stay on the rails?

Two questions dominate almost every payments conversation right now: does agentic commerce disintermediate the card networks, and do stablecoins become the "money layer of the internet" and replace the networks? Our view is that both concerns are more valuation overhang than fundamental reality. Agentic commerce and stablecoins are real and reshaping parts of the payments stack, but consumer behavior, merchant economics and how the networks have positioned themselves in emerging standards all suggest the through-cycle risk to network revenue is smaller than the market is pricing. 

The core bear case is that once AI agents do the buying, they will route real purchases to stablecoins and bypass card rails. We think this logic inverts on closer inspection. A consumer's agent optimizes net of rewards and buyer protection, not merchant’s cost or convenience, and rewards (not earned on stablecoin transactions) plus chargeback rights (which are absent in irrevocable stablecoin payments) tilt the math decidedly toward cards. Protection arguably matters more, not less, when a bot holds the wallet. On the merchant side, steering to a cheaper rail requires offering a discount at least as large as the consumer's rewards-plus-protection value, but the large merchants most able to deploy steering agents already have the lowest acceptance costs, often well below the reward hurdle. The switch is uneconomic where it is most technologically feasible — with the largest merchants — which is why consumer-facing stablecoin checkout has largely stalled at pilots. 

Meanwhile, stablecoins are increasingly a new settlement layer that runs over the same network rail, not around them. Visa now references more than 160 stablecoin-linked card programs, with stablecoin settlement at a ~$7B annualized run rate and growing 50%+ sequentially. Notably, the volume today appears concentrated in settlement, cross-border payments, treasury flows and payouts rather than consumer checkout transactions, underscoring our view that stablecoins are currently augmenting network infrastructure more than displacing the card rails in consumer payments.

A recently launched stablecoin consortium called Open Standard includes 140+ firms across networks, payments, tech, banking and crypto, and points to a coin layer commoditizing rather than concentrating into a single issuer, which is tougher on stand-alone stablecoin issuers than on the networks. Notably, some large digital wallet and merchant-services players were left out of the consortium, a data point worth watching, particularly for firms that issue their own stablecoin. 

Genuine second-order risks remain: which agent controls checkout (we tend to think the “customer is always right” approach points to the consumer's agent), on-chain take-rate mix, thin-acceptance emerging market substitution and open regulatory questions on stablecoin yield (notably the CLARITY Act). But changing consumer payment behavior at scale has always required heavy incentives, and those incentives tend to erase the very economic advantage that motivated the attempted switch. Net-net: agentic commerce and stablecoins represent more opportunity than threat for the networks in our view.