AI Could Be the Biggest Thing to Happen to Stablecoins. Or Is It the Other Way Around?
Stablecoins processed $33 trillion in transaction volume last year — more than VisaV and MastercardMA combined. And the companies behind that milestone are already building for a customer base that doesn't sleep, doesn't carry a wallet, and doesn't need a bank account: autonomous AI agents that pay other AI agents, millions of times a day.
Circle Internet GroupCRCL, Stripe, and a growing roster of crypto infrastructure firms are actively laying the rails for machine-to-machine commerce, betting that stablecoins, digital tokens pegged to a stable asset like the US dollar, are the only payment layer that can keep up with software running at internet speed.
Sixty-fold in five years
The stablecoin market reached $312B in capitalization by end-2025, up 60-fold from $5B in 2020, and it's stopped moving in lockstep with crypto prices. Circle's USDC supply sits near a record $78B even as Bitcoin and broader crypto markets remain well off their highs, a sign that stablecoins are increasingly used for actual payments, not just crypto trading.
Adjusted stablecoin volumes grew more than 90% year-over-year (YoY), and transaction velocity, how frequently tokens change hands, has climbed alongside it.
Within days of the GENIUS Act being signed into law in July 2025, which formally classified payment stablecoins as neither securities nor commodities, Goldman SachsGS, JPMorganJPM, Citi, Bank of America, and Morgan Stanley all addressed stablecoin strategy on their Q2 earnings calls simultaneously.
Per a recent Morph report, EY-Parthenon research found 54% of organizations not yet using stablecoins plan to deploy within six to twelve months.
Why AI agents need stablecoins
A legal AI agent fielding requests from other software agents might charge fractions of a cent per task. Running that through a traditional card network, with fixed fees and percentage-based pricing, wipes out the economics instantly.
"Microtransactions are a poor fit for traditional rails in terms of cost, latency, and programmability," Mark Palmer, an analyst at Benchmark-StoneX, told Bloomberg, adding that "AI agents would benefit from programmable money that can be embedded directly into software workflows without long settlement windows."
Circle CEO Jeremy Allaire draws a sharp line between AI shopping on your behalf and the actual opportunity. "The problem to solve is not how does an agent buy something on Amazon," Allaire said in an interview after Circle's February earnings call.
"The real opportunity is all of the things that AIs need to consume from each other." Think software agents paying for API calls, data pulls, or specialized services from other agents, at scale, across borders, without human sign-off at every step.
Sean Neville, who co-founded Circle alongside Allaire in 2013 and now leads AI-native financial platform Catena Labs, puts it plainly: "Stablecoins are effectively AI-native money.
If an AI actor is making a payment, it doesn't make a lot of sense for it to have a physical credit card, it makes a lot of sense for it to have money that moves at internet speed." Catena is explicitly building what Neville calls an "AI-native bank," with AI actors as the primary customers.
Building the pipes
Circle recently launched Arc, a high-throughput blockchain built specifically for stablecoin payments, and began testing "nanopayments", transactions that cost fractions of a penny, making machine-to-machine commerce economically viable for the first time.
Stripe, meanwhile, partnered with crypto venture firm Paradigm to build Tempo, a blockchain designed specifically for stablecoin payments that raised $500M at a $5B valuation.
Stripe has spent more than $1.1B acquiring stablecoin infrastructure overall, including the 2025 purchase of Bridge. Visa, Mastercard, UBS, and Shopify are all listed as Tempo partners.
Circle's stock has more than doubled over recent weeks following a blowout earnings report, and analysts at Bernstein have a $190 price target on the shares, implying ~60% additional upside from the $120 level at the time of their note.
Allaire's comments on agentic commerce are landing at a convenient moment. "Management did a pretty good job on commentary related to agent pieces, especially at this time in the equity market where AI is so disruptive to stocks," John Todaro, an analyst at Needham & Co., told CoinDesk. Per the same report, Bernstein's Gautam Chhugani rates Circle at outperform.
The gap between vision and volume
Coinbase GlobalCOIN-incubated x402, an open standard designed for agentic payments, reports $24M in total volume across ~94,000 buyers and ~22,000 sellers over 30 days, set against a global e-commerce market expected to reach $6.88T this year.
"I would not put stablecoins in the category of things that consumers are asking to pay with," Chris Donat, head of payments and fintech research at BWG Global, said in the same report.
Stablecoins also currently lack the fraud protection, dispute resolution, and credit extension that card networks bundle into every transaction.
The GENIUS Act covered stablecoins, but the broader market structure bill, governing which digital assets count as securities vs. commodities, remains stuck in Senate negotiations, with the White House ramping up pressure to get it passed before the legislative window closes.
Treasury Secretary Scott Bessent called on Congress to "finish the job" in a recent Wall Street Journal op-ed, warning that the US risks ceding its standing as a global financial leader without clear rules.
Per The Hill, senators Angela Alsobrooks and Thom Tillis have reached a bipartisan agreement in principle, but banking industry concerns about loopholes remain unresolved.
A Morph report forecasts AI agents as the single largest category of stablecoin transaction initiators by 2027, surpassing retail peer-to-peer transfers in volume within two years of mainstream agentic AI deployment.
Annual stablecoin settlement volume is projected to top $50T by end-2026. The infrastructure is being poured now; the question is how fast the machines show up to use it.