Stablecoins, Autonomous Payments and the Dollar’s Next Act
The rise of programmable money and payments that execute without humans are not two separate stories. A full picture of why almost everything, still, happens directly or indirectly in dollars.
There’s a way of reading economic history that explains more than it seems: every great leap was, at its core, a leap in transactionality. In how much, how far, and how often we can buy and sell. Beyond moving goods and information, the railroad and the telephone multiplied the number of possible deals. Let alone the arrival of computers, and even more so smartphones. And every time transactionality grew, money had to evolve to keep up: from metal coins to the check, from the check to the card, from the card to the instant transfer, and from the instant transfer to programmable money. It might seem obvious to jump ahead and conclude right here that less friction means more transactions. And more transactions, more commerce.
Today we’re living through two chapters of that same race at once. On one side, money became programmable: stablecoins can be “bills or coins” that a piece of software can move with a signature, with no banking hours and no borders. On the other, something else has started showing up more and more often: payments that execute without a human pressing a button. They look like separate topics, but they’re joined at the core, and they lead to a single underlying question: how much transactionality can the system handle, and what does it take to sustain it?
This article connects two dots that touch on crucial undercurrents in this great geopolitical war for control of the world’s financial center. It starts with stablecoins, moves on to the problem opened up by a network that records everything, and ends with autonomous payments. This is where transactionality runs out of ceiling…
Spoiler: almost everything, from one end to the other, ends up settling in dollars.
Part I :: Money That Became Programmable
What a stablecoin is (and isn’t)
A stablecoin is a crypto asset created to maintain a fixed parity with a reference asset, which in practice is almost always the US dollar. The “stable” part was meant to evoke stability, since perceived volatility is precisely what it tries to avoid. That’s why it would be more accurate to describe it as pegged, in reference to its link with the underlying asset. And for the same reason, the “coin” part is debatable too: is it really a currency, or a token? Beyond the legal implications, the underlying can be any asset whatsoever, not necessarily legal tender; and the design mechanism may not even involve a link to an underlying asset at all, sustaining the price or the supply relationship algorithmically instead. A clarification is worth making about these last ones: algorithmic stablecoins also have an anchor, since they aim to be worth “x” — just without any asset backing them; what they lack, then, is not the peg but the backing. With all those nuances, the core idea remains simple. But underneath that definition there’s a map with four very different models, and it’s worth knowing them before trusting any of them.
Fiat-backed stablecoins are the dominant model: for every token there is, in theory, a dollar, a euro or another currency held by the issuer. USDT, USDC, the new USAT or PayPal’s PYUSD work this way. Crypto-backed ones, like DAI or GHO, hold themselves up by over-collateralizing with other crypto assets and are governed by code via smart contract, with no single issuer. Those backed by other (non-fiat) assets put gold or Treasury bonds behind the token; XAUT and PAXG present themselves as “digital gold,” transferable in an instant. And algorithmic ones have no backing in the traditional sense: what an algorithm ultimately tries to do is hold the price. That last model had a major failure as a precedent, and it’s worth understanding why.
In May 2022, TerraUSD (UST) was the third-largest stablecoin in the world by market capitalization. There was no bank or reserve holding the dollars it represented: it used a two-token system (UST and LUNA) that minted and burned against each other to hold the peg, and it attracted capital with a protocol, Anchor, that promised up to 20% yield. When confidence broke, a “death spiral” was unleashed: to prop up UST, LUNA was issued, which sank LUNA, which sank UST further. In seven days, between May 9 and 15, some US$60 billion and over 90% of the market cap evaporated. The design’s justification rested on classic supply-and-demand thinking and on interest in an ecosystem (Terra) that was growing exponentially, not just in capitalization but in development too. The lesson got etched into the regulation that came afterwards: if there’s no underlying asset behind it (and let’s add: a liquid one), it isn’t stable; it’s a promise that lasts as long as the euphoria does.
As for size, the market is no small thing. According to CoinGecko, stablecoins add up to around US$314 billion*, roughly 14% of all crypto market capitalization, with traded volume hovering around US$137 billion per day. And the dollar dominates by far: of that total, around US$307 billion — more than 97% — are dollar stablecoins. Euro stablecoins barely reach some US$900 million (EURC at the head) and gold ones, about US$175 million. There are versions in pounds, yen, francs, reais and even Argentine pesos (we’ll come back to this), but the vast majority of the world’s “stable digital money” is denominated in US dollars.
*Data as of 6/5/2026 — Source: CoinGecko
From niche tokens to infrastructure: how they climbed the ranking
There’s a simple way to see how much, and how, everything changed: look at the market-cap ranking of the largest crypto assets. In 2014, when Tether was born, stablecoins didn’t even register. According to a CoinGecko study that takes a snapshot every June 1 between 2014 and 2026, the first stablecoin entered the top 10 only in 2020: USDT showed up in third place with 3.9% of the top 10’s capitalization and never left. Not that USDT was new: it had existed since 2014 and was already the most important stablecoin; what changed is that only then, by total capitalization, did it cross into the top 10. USDC joined in 2021. By 2026, USDT weighs in at around 8.3% and USDC at 3.3%, and between them — plus newer names like USDe, USDS or the stablecoin tied to the Trump family — stablecoins occupy close to 11.6% of the top 10, a category that six years earlier simply didn’t exist in those tables.
The deeper takeaway goes beyond a percentage: it means that an ever-larger share of crypto stopped being speculative by design. This, admittedly, spirals into a whole set of arguments about the term speculation and its association with the crypto ecosystem: volatility gets framed as the sole characteristic of risk (but also of opportunity), when everything actually depends on knowledge and on what you’re investing in. That’s what we use analysis for, and why we avoid herd behavior. Thanks to volatility, Bitcoin and Ethereum always offered chances to buy or re-buy below the current price. But the point we want to make is this: from a single stablecoin in 2014 to more than 150 today, “stable crypto money” went from oddity to infrastructure. And that infrastructure is, precisely, what everything that comes next is going to run on.
And the climb didn’t stop. By mid-2026, in the middle of a sharp market downturn, something happened that would have sounded unthinkable years ago: Tether came to match — and at moments surpass — Ethereum in market capitalization, fighting it for second place in the ranking (both hovering around US$186–187 billion). The mechanics are telling: while ETH plunges with the market, USDT not only barely moves thanks to its link to the underlying asset (a price the market accepts, for now), but on top of that, selling other cryptos usually happens against the stablecoin Tether issues. That in turn drives up demand for it — so if it loses the peg, it does so by rising above one, not falling below it. Supply and demand, in the end. Today, among the 26 largest crypto assets, six stablecoins coexist: USDT, USDC, DAI, USD1, USDe and PYUSD. “Stable crypto money” is no longer a satellite of crypto — it’s one of its largest masses.
Coinbase’s own CEO, Brian Armstrong, summed it up on June 5, 2026, when in a post on X, in the middle of the price slump, he said that Bitcoin going down doesn’t mean crypto is going down; stablecoins, derivatives and prediction markets are on the rise.
https://x.com/brian_armstrong/status/2063027072461935084?embedable=true
His point is exactly this article’s point: the crypto ecosystem is, right now, much broader than Bitcoin and touches every area of finance. And stablecoins are one of the clearest proofs of that.
Looking inside: the code, the control and the trust
Now then… how exactly does a stablecoin work? How is one created? The question first takes us to a distinction: while a native token is the main asset of a blockchain (the currency the network runs on), a non-native token is built on top of an existing network, through a smart contract and standards like ERC-20 on Ethereum or SPL on Solana. A stablecoin like USDT is precisely the latter: a program that lives on a blockchain but is not the network’s own currency. And like any program, it does or executes exactly what its code says — no more, no less.
And what does it do? At the base, for cases where there’s a link to collateral, for every token in circulation there should be an underlying asset in custody. The token is minted when someone deposits dollars, and burned when they withdraw them. It can also happen that crypto assets are minted and held in the contract and later transferred — it all depends on how the contract is parameterized.
A lot of people repeat that “on the blockchain, code is law” and that a contract, once deployed, is immutable. That’s a half-truth. The code of the USDT contract cannot be rewritten. But the state — the balances, the blacklists, the variable that flags the contract as deprecated — does change, or can change. All it takes is a call, via transaction, from whoever holds sufficient rank under the roles assigned in the base smart contract. If you look inside the contract you find functions like addBlackList, removeBlackList and destroyBlackFunds: the issuer can freeze an address, remove it from the list, or outright destroy the tokens it holds. USDT even has an upgrade mechanism: an upgradedAddress variable and a deprecated flag which, if activated, redirect every transfer to a new contract. This isn’t exclusive to Tether: USDC can freeze and wipe balances, and almost all major centralized stablecoins have some kind of pause button. And this is independent of the blockchain it runs on and of that chain’s level of decentralization. That said, it’s worth stressing that not all non-native tokens — the ones created via smart contract — have these functions or similar ones. It all depends on what that piece of software says, and on accepting it. In fact, one of the exceptions is DAI, which delegates that control to MakerDAO’s governance rather than to a company.
The paradox is worth naming, because it organizes everything that follows. Bitcoin was born to eliminate trust: concretely, by its purpose or end goal, so that you wouldn’t have to believe any bank or any issuer. Stablecoins travel the reverse path. They “solve volatility” by reintroducing exactly that trust: you trust that the issuer holds the underlying assets, that it won’t freeze your wallet, that the audit tells the truth. This is no minor point — it is de facto what makes them usable at scale, and also what makes them, ultimately, controllable.
The issuers: a story with enormous numbers
Behind the most-used “digital dollar” there’s a company with a balance sheet that surprises. According to its year-end 2025 attestation (audited by BDO), Tether had more than US$186 billion of USDT in circulation and reserves of nearly US$193 billion. What’s remarkable is where those reserves sit: more than US$141 billion in US Treasury bonds, which places it among the largest holders of American debt on the planet. At a moment in history when many countries are shedding — or have already shed — these bonds, the company led by Paolo Ardoino ranks above countries like South Korea, around 17th place if it were compared with nations. Add some US$17 billion in gold (it buys up to two tonnes per week) and US$8 billion in bitcoin. The result? More than US$10 billion in profit in 2025, with a margin its CEO describes as close to 99%.
Read in reverse, the figure is political as much as financial: a private stablecoin issuer became one of the largest sources of financing for the US government. Every time someone in an emerging market buys a digital dollar to protect themselves from their local currency, that dollar ends up, in good measure, buying US Treasury debt. We’re not demonizing what’s happening — just describing the status quo, and one of the main reasons the US government has given a positive nod to accelerating crypto adoption. Global transactionality gets recycled into bonds. And to round off the centralization theme, it’s worth knowing that 99% of those Treasuries are held in custody in a single place: Cantor Fitzgerald.
Why they exploded: digital dollarization
If stablecoins grew this much, it wasn’t because of speculation in the crypto trading and/or scalping scene, but out of a very concrete and very human need: protecting value. And nowhere is it as visible as in Latin America.
The Argentine case is textbook. With a peso that lost close to 95% of its value against the dollar in five years and inflation that topped 200% in 2023, people aren’t looking to invest: they’re looking not to lose. That’s why Argentina is, per capita, the world leader in adoption: more than 70% of crypto purchases in the country are stablecoins. It’s not an isolated phenomenon. Across the region, stablecoins already account for close to 40% of crypto transfers; in Colombia they’re around 52% of purchases, in Brazil close to 90% of crypto activity is tied to them. Analysts call it, accurately, digital dollarization.
On top of that comes the other big use case: remittances. The region received some US$174 billion in 2025, and sending money through traditional channels costs on average 6% of the transaction; with stablecoin infrastructure, it drops to 1–2%. For a migrant sending US$300 a month, that difference is a full month’s grocery cart. With more than 40% of the region’s population unbanked, and instant payment systems like PIX or SPEI already mainstream, the jump to a dollar that lives on your phone is almost natural. Which shows how advancing technology effortlessly breaks through any institutional or cultural barrier.
And here comes the nuance that often gets lost: not everything is the dollar. Local-currency stablecoins have started to appear — Ripio launched wARS, pegged to the peso, in late 2025, and plans more — designed precisely for everyday payments and transfers within the region. Same technology, different unit of account. But, for now, the dollar wins.
The lineage and the regulated play: USDT0, USAT and the ARGt case
The family tree helps explain where all this is going. From Realcoin to Tether, from the Omni layer on Bitcoin to expansion as an ERC-20 on Ethereum and the multichain rollout, until arriving in 2025 with two “new products”: USDT0, an omnichain version for moving across networks without fragmenting liquidity, and USAT, the regulated version.
USAT hit the market on January 27, 2026. It’s the first product Tether designed to fit inside the new US stablecoin law, and the difference is fundamental: Tether doesn’t issue it — Anchorage Digital Bank does, the only crypto bank with a national charter in the US, supervised by the OCC. The reserves are held in custody by Cantor Fitzgerald, with monthly attestations (the first one reported just US$17.6 million). But why does USAT matter if USDT already exists? Because there’s a rather relevant detail: the USDT we’ve always known is not legally available to US customers. USAT is the vehicle for competing, within the law, with Circle’s USDC for the regulated on-chain dollar.
In Argentina, the ARGt episode showed the other face of regulation. This 1:1 “digital peso,” issued by Twin and offered in Belo’s app, was halted by the CNV in March 2026: by promising up to 32% annual yield, it qualified as a security (an investment contract), not a means of payment. On April 15 the CNV lifted the measure, but with one condition: removing every reference to yield. The line the regulator drew — payment versus investment — is exactly the same one that structures the US law. But there’s an important nuance that shouldn’t be missed: the CNV did not require audits of the backing. Even so, Argentina is in the middle of its regulatory process, with a sandbox extended in time and a public consultation mechanism open to keep reviewing everything tied to an ecosystem that still has many angles left to cover.
The big regulatory shift in the US: GENIUS and CLARITY
The GENIUS Act was signed into law on July 18, 2025, and it’s the first major US crypto law. It requires payment stablecoin issuers to obtain a license, to keep 100% of reserves in cash and Treasury bonds, to publish monthly attestations (with an annual audit if they exceed US$50 billion), and it forbids them from paying interest on balances. This is where USAT comes from. A detail that’s often confused: the law is already enacted, but it doesn’t take full effect until around January 2027 (or 120 days after the final rules).
The CLARITY Act targets the rest of the board: it defines what is a security and what is a commodity, and splits oversight between the SEC and the CFTC. The House of Representatives passed it in 2025 and the Senate Banking Committee advanced its version on May 14, 2026 (15 to 9), but it still has a long road ahead in a midterm election year. The main sticking points are how much a stablecoin balance can be “rewarded,” and the conflicts of interest tied to the Trump family’s crypto businesses. Together, the two laws promise to shrink the gray zones and to define the stablecoin as a means of payment, with the rest of the crypto market getting a framework of its own.
Part II — A Network That Remembers Everything
The privacy problem
If we’re going to transact more and more, it’s worth knowing where all of that gets recorded — especially in the age of automation. The answer is somewhat uncomfortable: on a public blockchain, 100% of operations are written down and permanent. There’s no delete button, no hide button, once the information lands on the chain. What always looked like an advantage and a strength is now starting to show a flip side the crypto community had been warning about since Bitcoin’s launch.
Your address is a pseudonym. It doesn’t carry your name or a specific ID, but it records everything you do in a single place, forever. And there’s an entire industry dedicated to linking those addresses to real people: Chainalysis does on-chain intelligence for governments and works with the FBI, the DEA and Europol; Arkham uses AI to tie wallets to people and companies, and bills itself as the “Bloomberg of crypto.” With a single point of contact — an exchange with your ID, a purchase with your name — your whole history can be reconstructed, and as a byproduct, you get linked to other actors who also start losing their pseudo-anonymity. The veil slips not just for you but for anyone who transacts with you. And the trail grows the more transactionality there is, shaping a system that is at once an extremely powerful compliance tool and a form of surveillance growing exponentially thanks to new technologies.
Proving without showing
Is there a way out? One idea is gaining strength, and it consists of not hiding the data, but proving it without showing it. That’s the idea behind zero-knowledge proofs.
To keep it simple, the classic example is age. You want to prove “I’m over 18” without handing over your date of birth. You generate a cryptographic proof of the fact — with schemes like zk-SNARKs (Groth16), which produce tiny proofs that verify in milliseconds —; the system validates it and grants you access, without ever seeing the data. The proof is at once private (it reveals nothing extra) and sound (it can’t be forged). This isn’t theory: the European identity wallet (EUDI, under eIDAS 2) already requires selective disclosure, and it fits the GDPR’s data minimization principle. If you can prove without handing over the data, there’s no personal database to leak or to safeguard.
The other face: when the math fails
Zero-knowledge proofs are as powerful as they are hard to implement well, and it’s best not to romanticize them. The most recent case made that crystal clear. Zcash, the crypto that uses these proofs to offer fully private transactions, carried a critical flaw in its “Orchard” circuit for years: a poorly defined constraint allowed false data to be fed into a cryptographic operation and, with that, to mint unlimited amounts of counterfeit currency undetectably. The vulnerability was live from Orchard’s activation in May 2022 until it was urgently patched in early June 2026.
What’s unsettling is unsettling twice over. On one hand, it had survived years of scrutiny from some of the best cryptographers in the world. On the other, the system’s own privacy properties meant there was no way to prove cryptographically whether anyone exploited it before the fix: the same math that protects the user makes the counterfeiter invisible. The flaw was found by a security researcher, Taylor Hornby, using cutting-edge AI tools alongside traditional methods; the developers are now proposing a network upgrade so that anyone can verify that Zcash’s supply was not tampered with.
The moral isn’t “privacy is bad,” but something more sober: in these systems, a single error in an equation can be worth a fortune, and sometimes you can’t even know whether someone collected it. The challenge will be resolving that dilemma.
Here an uncomfortable circle closes back to Part I. In March 2026 the FATF published a report flagging stablecoins as the asset most used in illicit transactions: according to Chainalysis, they concentrated close to 84% of illicit on-chain volume in 2025, and TRM Labs estimated some US$141 billion received by illicit entities (nearly half tied to a sanctioned Russian token). The recommendation? AML rules, attention to transfers between non-custodial wallets, and (verbatim) evaluating tools such as wallet freezing and restricting certain smart contract functions. In other words: the regulator is asking for exactly the freeze and blacklist buttons from the beginning of this article. Privacy and control pull on the same rope.
Tornado Cash: is writing code a crime?
The rawest collision between privacy and the law has a very well-known name — to the point that it’s the inevitable example in any article or conference on the subject. Tornado Cash is a non-custodial mixer that uses these same techniques to cut the trail between the origin and destination of funds. The US sanctioned it in 2022 and withdrew the sanction in 2025, after a court ruling. But its co-founder, Roman Storm, went to trial: on August 6, 2025, a jury found him guilty of operating an unlicensed money-transmitting business (up to 5 years) and hung on the two most serious charges — money laundering and sanctions violations — which add up to 40 years.
The prosecution requested a retrial for October 2026; the defense holds that the hung jury proves the doubts around criminalizing the mere act of writing code that also has legitimate uses. Even Vitalik Buterin published an open letter in January 2026 asking for clemency. The paradox: the Department of Justice itself declared in 2025 that “writing code is not a crime”… and still pushes for the second trial. The question will haunt anyone who builds tools for autonomous systems: is the person who writes a tool responsible for what others do with it?
We already have programmable dollars and a network that records everything. What’s missing is for paying to be as simple and automatic as loading a web page. That’s what x402 is about.
Part III — No Ceiling: Autonomous Payments
The human bottleneck
Back to the curve from the beginning. Human transactionality always hit the same limit: decision fatigue. We’re human, it’s that simple. A person cannot — and does not want to — approve a thousand micropayments per hour. To put it in perspective: Visa processes on the order of 293 billion transactions per year, and the world’s non-cash payments hover around 3.4 trillion annually. Sounds enormous, but it’s a ceiling on human action. When the one paying stops being a person, that ceiling disappears.
The internet learned to show and connect almost everything, but it never could charge natively. And the detail is almost poetic: around 1997, a code was reserved in the HTTP protocol — the language computers have spoken to each other since 1989 — called 402: Payment Required. It sat marked as “reserved for future use” for nearly three decades. If the famous 404 says “this doesn’t exist,” 402 always wanted to say “it exists, but pay up.” Nobody ever defined how. Until 2025, when Coinbase, and shortly after Cloudflare, switched it on.
How x402 works
The flow is elegant for being extremely simple, which doesn’t make its implications any less profound. The client requests a resource; the server responds 402 with the payment terms in the header (scheme, price, network, address); the client retries, attaching a signed authorization. There’s a very fine adjustment in this: it uses EIP-3009 (transferWithAuthorization), a gasless “check” with a fixed amount and recipient, so the server can’t overcharge or charge twice. A facilitator (Coinbase offers a free one, but anyone can run their own) verifies and settles the transaction on-chain, and the server returns the resource with a 200 OK and a receipt (X-Payment-Response) containing the transaction hash. Four steps, inside the same old HTTP. And on top of that it’s network-agnostic: most payments settle in USDC on Base or Solana, for speed and cost. (Once again: “the dollar,” so far.)
The full stack
x402 doesn’t work alone. It’s one layer inside an architecture that only recently clicked into place.
Underneath sits settlement (the stablecoins from Part I). On top, payment (x402). Higher up, identity and reputation: the ERC-8004 standard (“Trustless Agents”), which went live on Ethereum mainnet on January 29, 2026 — developed by people from MetaMask, the Ethereum Foundation, Google and Coinbase — gives every autonomous system an on-chain identity (an NFT with its “file”), a reputation registry and a validation registry, integrated with x402. And at the top, the coordination protocols those systems use to discover and talk to each other: A2A, MCP and Google’s Agent Payments Protocol (AP2). To round out the system, a rating layer is already emerging, and access is starting to be decided by it. This is very familiar by now — today we use thousands of services through ratings, from restaurant reviews to ride-hailing drivers. Even Satoshi Nakamoto foresaw, in Bitcoin’s very first code, the possibility of a marketplace with signed products and a base rating to build trust on.
It’s already happening
If you’ve made it this far, it’s worth underlining: this has already started. By mid-2026, x402 had accumulated some 165 million transactions across more than 480,000 agents, with an average value close to US$0.30 per call and total volume on the order of US$50 million. The concrete cases exist: GPU providers charging per inference, on-chain data services selling each query, marketplaces where systems buy and sell among themselves and APIs that generate code and charge a few dollars in USDC per call. The heavyweights came in: Stripe integrated x402 in February 2026, Cloudflare lets you put a toll on any site from the network edge, Google’s AP2 uses it for payments between systems, and x402 became native to Amazon Bedrock AgentCore, where an agent pays in USDC without human intervention and the operation settles on Base in about 200 milliseconds. Even MCP servers can expose tools that pay for themselves when invoked.
Given the obvious interest behind this protocol’s proliferation, it’s not one company’s bet: behind x402 already stand Coinbase, Stripe, Cloudflare, Amazon and Google.
Why it changes the economics, not just the speed
This isn’t “paying faster.” It changes supply and demand at the same time. On the demand side, a system can query a hundred APIs an hour, each from a different provider, and pay only for what it uses. On the supply side, any microservice becomes sellable: a data point, a computation, a query that wasn’t worth charging for before. With fees of fractions of a cent, subscribing to everything stops making sense: you pay for exact usage. The subscription model — which existed precisely because charging small amounts many times was expensive and cumbersome — starts to become obsolete.
But there’s a leg that tends to be forgotten and will be decisive: the receipt. The payment already knows how to move on its own; the invoice still needs to learn to travel with it. Without a receipt there’s no control, and without control there’s no accounting or auditing possible in a system where the one spending may not be human. The payment and the document that justifies it have to travel tied together. That’s why I’ve submitted a pull request to the x402 open-source repository to address this problem. For those interested, the paper behind the proposal is available here: Verifiable Invoice Commitment: An EVM-Native Standard for Committing Fiscal Metadata to On-Chain Payments by Javier Mateos :: SSRN
There’s always a dark side
When systems start paying each other, new structures appear to organize them.
And five questions appear that nobody has finished answering:
- Laundering at machine speed. Automated stablecoins between non-custodial wallets: exactly what the FATF flags as the key vulnerability, with that 84% of illicit on-chain volume and the ~US$141 billion of 2025.
- Market manipulation without a lab. All three barriers fall at once — capital, knowledge, coordination. The flash crash of May 6, 2010, in which close to a trillion dollars evaporated in minutes and where a trader operating from his home (Navinder Sarao) played his part, gives an idea of what a swarm of systems could do today.
- Diffuse jurisdiction. Operator, facilitator, stablecoin and counterparty in four different countries. No major regulator — not the SEC, not the CFTC, not ESMA, not IOSCO — has issued specific guidance. A real vacuum.
- Dilution of responsibility. A hires B, B hires C, at machine speed. If C delivers a wrong data point and a company makes the costly decision, who answers for it? The Air Canada precedent (2024) — a tribunal held the company liable for what its chatbot promised a customer, rejecting the idea that the bot was a “separate entity” — points the way: whoever deploys the system carries what the system does.
- Geopolitics of intelligence. The data your system sees, it sends. A handful of companies, on another continent, end up deciding the price, the access and the policy over sensitive data belonging to local businesses.
Who pays, rules
The transactionality curve never stopped, because the system always wanted more. And since 1971, with the end of the dollar’s convertibility to gold, the current monetary system — built on trust and on issuance against debt — needs the economy to grow constantly. And for the economy not to stall, the engine of transactionality not only can’t stop: it has to accelerate. More than 37 trillion dollars of debt depend on it. Stablecoins were the latest great leap on the money side; autonomous payments are the next one, on the side of who executes. The underlying novelty is old and new at once: who pays, rules — and, for the first time, the one paying may not be a person. That’s why control stops being a formality and moves to the center: it’s not enough that something can pay; you have to be able to reconstruct what it paid, why, with whose authorization and under which rules.
How much is at stake? Depends on who you ask — and the dispersion of the forecasts is, in itself, a data point. Juniper Research projects US$1.5 trillion in global autonomous spending by 2030; McKinsey talks about US$3 to 5 trillion; Bain and Morgan Stanley, more cautious, calculate hundreds of billions for the US alone. Nobody knows the exact number. Everybody agrees on the direction.
And there’s one last reading left, the one that closes the arc. From one end of this curve to the other — from the Argentine saver buying digital dollars so as not to lose, to the system paying an API in fractions of a cent — almost everything settles in the same currency. Without tanks or treaties, stablecoins extended the dollar’s hegemony into the economy of machines. What’s missing isn’t technology: it’s clear, fundamental rules. You can’t regulate after every technological shift; you’re always behind. You have to understand the underlying engine and, instead of writing in a hurry, watch how money has already learned to move on its own.
Data verified as of June 2026 (CoinGecko, Tether/BDO, FATF, TRM Labs, Chainalysis, Coinbase, Juniper Research); market figures and regulatory status change fast.
