Three days on the floor in Amsterdam, and the word everyone had agreed on in advance was agentic: on the main stage, in every breakout, on the lanyards of firms selling protocols that did not exist eighteen months ago. The signal that stayed with me, though, was quieter than the one being broadcast. More in the corridors than from the lectern, I heard a market rearranging itself around a single question of who buys, who builds, who partners, and who is left explaining the delay. Three things came through more clearly than anything on stage.
The first is that the stablecoin race is on and is no longer speculative, and it is worth looking closely at what is actually changing hands, because the pattern is consistent. When Mastercard agreed in March to acquire BVNK for up to $1.8bn, it was not buying a coin. BVNK is a stablecoin payments stack: the licences, the on- and off-ramps, and the orchestration to move value between fiat and stablecoins across more than 130 countries, already running quietly behind names like Worldpay and Deel. Mastercard had the global fiat reach and no native way to settle on-chain around the clock; BVNK had the settlement layer and needed distribution. Folded into Mastercard Move, it gives the network 24/7 stablecoin settlement for acquirers and processors, and stablecoin checkout through its gateway. The same logic explains Coinbase's roughly $2bn run at the same company last November, which collapsed before Mastercard stepped into the gap.
Behind it the pattern repeats. Stripe's $1.1bn purchase of Bridge, now closed, bought a single API to receive, hold, convert and issue stablecoins, and through Bridge's Open Issuance the ability to let its own customers mint a stablecoin and fund cards in markets where dollars are scarce. Ripple's smaller acquisition of Rail, at around $200m, bought a platform already handling something like a tenth of global stablecoin payment volume, with the banking relationships to move fiat in and out, and so distribution for its own RLUSD. S&P counted at least fourteen such transactions in 2025. In every case the asset is the same. It is not the token but the execution and the licences: the rails, the permissions, the ability to move money across borders without breaking at the regulatory edge. That is what is scarce, and that is what commands the premium.
The regulatory edge is already doing the culling. MiCA's transitional window closes on 1 July, and the numbers tell the story plainly. By May, only 194 crypto firms across the EU held a full licence, in a market that counted more than three thousand registered businesses two years ago. France has said it will treat operating without a licence as a criminal matter. We have seen the shape of this before, in stablecoins themselves: Tether's USDT never met the standard and was quietly removed from the major European exchanges, while Circle's compliant USDC and its euro counterpart kept their place. When the rules bite, the licence becomes the business.
Which brings me to the question I would put to any founder reading this: why are you building? To compete and to grow, of course, but what is the plan beneath that? The incumbents, the fiat-rails generation who built the first wave, are now in their own race to buy, build or partner, and most of them will buy, because they do not have the time to build and AI is pulling their attention in three directions at once. You may feel safe inside a partnership, and you are, right up until your partner decides to build the thing themselves, at which point you become a line in someone's make-versus-buy analysis. The instinct is to hold out for a better valuation, and that instinct tends to misjudge the clock. This is less about chasing valuations than about chasing customers and reaching execution readiness while the window is open. The aim is not to be the last one standing; it is to be early.
The third thing I heard was less a conversation than a clock ticking under all the others.
The EU is rewriting the rulebook for how payments work across Europe. PSD3 is the directive and the PSR is the regulation alongside it; together they replace PSD2. The texts were agreed in April, with the substantive obligations landing largely in 2028: far enough away to feel abstract, close enough that the readiness work starts now. A handful of changes matter most. The payment-institution and e-money-institution regimes merge into a single authorisation. Non-banks gain more direct access to the rails, which erodes the very thing a bank partnership used to provide. Liability moves to whoever performs the authentication, with confirmation-of-payee becoming mandatory. And there is a tailwind for pay-by-bank, as mandatory open-banking interfaces make account-to-account payments materially more viable.
This is harder than it looks for anyone comfortable under PSD2 today. Existing licensees will have to re-authorise under the merged framework, and the bar around capital, safeguarding and governance steps up with it. For a smaller payment or e-money firm on a thin margin, that is a fixed cost arriving precisely as its principal advantage, privileged access held through a banking partner, is handed to everyone. Meeting the new standard is increasingly a question of scale, and a great many licensees do not have it. I would expect a wave of consolidation to follow the rulebook rather than precede it: the same pattern MiCA is now setting in crypto, arriving in payments a year or two behind.
Strip away the keynote language and the impact on payments is concrete and already being built into the rails. Both networks have shipped products. Mastercard's Agent Pay issues agentic tokens that bind a card credential to a specific agent, merchant and consent policy, so a model can complete a checkout without ever holding the card number; Visa's Intelligent Commerce does the equivalent with scoped tokenised credentials and issuer-side authentication. OpenAI, in conversation with Checkout.com, was candid that it wants the orchestration and intent layer. The genuine questions sit underneath. Authentication has to be rebuilt around mandates and cryptographic consent. Liability turns murky: for a validly issued token the issuer carries fraud liability as usual, but the consumer keeps chargeback rights, so a merchant can still face a dispute on a purchase the customer's own agent authorised. Verified agent identity becomes essential, and product data quietly becomes a payments problem, because an agent routes around anything it cannot read. Adyen put it plainly: getting the agent ready will be the story of the year, and most of that work sits upstream of the checkout.
Two pieces of plumbing are worth noting in the same breath. The UK Payments Initiative went live on 2 June, the first new UK payment scheme since Faster Payments in 2008, giving commercial variable recurring payments a common rulebook. And Adyen was selected as PSP for GOV.UK Pay, a three-year mandate worth up to £25.3m across roughly a thousand public-sector services, displacing Stripe on non-Crown flows. Neither made the headlines; both are the kind of infrastructure that quietly compounds.
The temptation at a week like this is to leave persuaded that everything has changed overnight. The truer reading is that the destination is no longer in doubt and the route is mostly unbuilt, and that the firms who will matter are deciding now whether they are buying that route, building it, or being bought along the way. That decision does not get easier the longer it waits.
I will be in London on 1 and 2 July for IGB Live. If our paths cross, even briefly, it would be good to connect.
If you are building or backing in payments and quietly weighing what readiness, or a sale, looks like from here, I am always open to a thoughtful conversation.
And — if you have read this far, consider it a quiet compliment. A few prior pieces that I think sit alongside today's thinking rather well:
Securities Offered through Wellesley Hills Securities. Member FINRA/SIPC
