A non-custodial stablecoin off-ramp just shipped — and it exposes the half of the chain merchants never fixed

What shipped on September 18, 2026

The announcement is short enough to quote almost in full. Stable.com, the financial platform operated by the global payments group Unlimit, connected to the Polygon Open Money Stack. Holders of USDT and PayPal USD (PYUSD) on Polygon can now initiate a bank transfer, in their own name, directly from their own wallet — without first depositing those tokens into an account the platform controls.

Polygon Labs CEO Marc Boiron framed it exactly the way a merchant would: "People should not have to surrender control of their assets simply because they want to send money to a bank account." Unlimit CEO Kirill Eves added the usual industry line about money moving freely across currencies, accounts and networks. The plumbing behind the fiat leg is Unlimit's payout network, which the company says reaches 180+ countries and 150+ currencies, as reported by FinTech Global and Electronic Payments International.

The Open Money Stack itself isn't new. Polygon Labs unveiled it in January 2026 as a modular framework for stablecoin payments, opened a technical preview in June, and published a general overview on September 9. Polygon's own documentation describes it as "fiat on-ramps and off-ramps, wallet infrastructure, compliance, stablecoin orchestration, and onchain settlement in one open, vertically integrated platform," with six layers — wallets (custodial and non-custodial), ramps and cash access, cross-chain interoperability, the chain itself, stablecoin orchestration, and agentic payments. Licensed ramp coverage includes bank transfer, debit card and cash at 50,000+ retail locations, with KYC and AML built in. Polygon puts average transaction cost at roughly $0.002.

So the headline is modest: one platform, two tokens, one chain. The implication is not modest at all.

The exit was the last place custody was still compulsory

Think about the full life of a dollar that reaches a business through crypto rails, and ask at each step: who could stop it?

Receiving. If you use a non-custodial payment processor, nobody. The customer's payment lands at an address you control. There is no processor balance, no payout schedule, no account review standing between the sale and the money. We've called this "the settlement is the payment" for two years because it's the literal mechanic.

Holding. Also nobody, if the keys are yours. That's the entire argument for a self-custodial stablecoin wallet, and it's why the recent wave of bank-issued tokens with freeze and clawback functions — U.S. Bank's USBDC pilot being the cleanest example — reads so differently to a merchant than to a treasurer.

Cashing out. Here the honest answer, until this week, was: somebody. To turn stablecoins into a bank balance you sent them to an exchange or a fiat gateway, where they became a number in that company's ledger, subject to that company's compliance queue, withdrawal limits, maintenance windows and risk appetite. You had built a chain with exactly one custodial link, and it was the link you had to use every single month.

That is what the Stable.com integration attacks. Not the license — a fiat payout still requires a regulated institution, and no amount of self-custody changes that. What it removes is the balance transfer: the window during which your money sits in someone else's account for operational rather than legal reasons.

What that one custodial hop actually costs

Businesses underprice this step because most months it's boring. The cost shows up in the tail.

First, timing risk on the one day you can't afford it. Payroll is Friday, the deposit is "under review," and support replies in 48 hours. Nothing was stolen, nothing was frozen in the legal sense — the money simply wasn't yours to move for two days.

Second, counterparty risk you didn't price. A balance at a platform is an unsecured claim on that platform. We went through the mechanics in counterparty risk in stablecoin payments and the operational version in what happens when your crypto payment provider shuts down. The size of the claim doesn't matter on the good days and is the only thing that matters on the bad one.

Third, re-verification friction. Every custodial hop is a place where someone can ask you to prove yourself again, at a moment of their choosing. The pattern is spreading to the payout side too: in August, Deel began requiring workers electing stablecoin payouts to confirm ownership of the destination address before funds move — the subject of our piece on wallet verification for stablecoin payouts. That's a reasonable control. It's also one more gate between you and your money.

Fourth, and least discussed: the hop constrains how you operate. If cashing out is slow and annoying, you cash out in big infrequent batches. Big infrequent batches mean larger balances sitting in one place for longer — the exact concentration you were trying to avoid.

Why the custodial hop existed in the first place

It's worth being precise here, because "they removed custody from the off-ramp" can be read as a bigger claim than anyone is making.

A bank transfer is a regulated act. It has to originate from a licensed institution, on behalf of an identified person, with AML screening against the beneficiary and the source of funds. None of that goes away. Stable.com is still doing KYC. Unlimit's licensed network is still the thing that touches the banking system.

What was never legally required is the part everyone assumed was: that your tokens first become a deposit liability of the platform. That step is an artifact of how these products were built — take custody, credit a ledger, then sell from the ledger — not a requirement of the rails. Pull it out and you get the same regulated payout with a much shorter window of exposure. Polygon's documentation is explicit that its wallet layer supports both custodial and non-custodial models; this integration is what choosing the second one looks like in production.

That's the structural lesson, and it generalises well beyond one partnership: in payments, most custody is a design decision dressed up as a compliance requirement. Card acquiring holds your money because the model is built on reversibility and rolling reserves, not because a regulator wrote "hold the merchant's funds for three days." The same logic is why a crypto processor that takes custody is making a choice, not obeying a rule.

Read the boundaries before you rebuild your treasury on it

We'd rather be the ones to say this: the announcement is narrow, and a business should treat it as a promising first implementation rather than a finished rail.

• Two stablecoins, one chain. USDT and PYUSD on Polygon. If you settle in USDC, or on Base, Solana, Arbitrum or anywhere else, this specific route doesn't reach you today.
• No published economics. The announcement didn't include transaction fees, processing times, per-country availability or which banks are supported. Those four numbers are the ones that decide whether a route is usable, and none of them are public yet.
• KYC and AML still apply. Non-custodial means your assets stay in your wallet until the transfer executes. It does not mean anonymous, and it does not mean unconditional — a payout can still be refused.
• The provider is mid-transition. Unlimit rebranded its business accounts operation as Stable.com in September and hasn't published a completion date for migrating remaining European clients. Early platforms move.
• It's still a dependency. A non-custodial off-ramp is a better dependency than a custodial one. It is not the absence of one. Keep a second route.

None of that makes it unimportant. Direction of travel is the story: a year ago the industry's answer to "how do I get from stablecoins to a bank account" was uniformly "give us the tokens first." That answer is now contested, by a major chain and a licensed payments group, in public.

The money-in side is the half you already control

Here is the part a business can act on this week, and it's the opposite end of the same chain.

Everything above is about the exit. But the exit only matters in proportion to how much you have to use it — and the first question is where the money lands when a customer pays you. With Payzum, it lands in a wallet you control. There is no Payzum balance, no payout schedule, no settlement delay, because there is nothing to settle: the customer's transaction to your address is the settlement. Payzum never holds, pools or controls the funds.

The concrete capabilities, and nothing more than the real ones:

• Online: hosted checkout (redirect, modal or inline), no-code payment links and buttons, invoices with expiry and overpayment detection, donations and tip jars, and recurring subscriptions. Drop-in compatibility with existing e-commerce plugins, snippets and webhooks.
• In person: POS with a fresh QR per sale, physical terminals, PIN-protected cashiers and per-cashier and per-terminal analytics. No acquirer, no card-network fees, no chargebacks.
• Paying out: mass payouts by CSV (BTC, LTC, DOGE) plus EVM stablecoin payouts on Polygon, Arbitrum, Optimism, Base, BNB Chain and Avalanche. We compared the chains on cost and confirmation in mass payouts on Polygon vs Arbitrum vs Base.
• Volatility: accept whatever the customer holds, with optional auto-conversion to USDC or USDT so what you keep is a dollar-denominated balance.
• Developers and agents: a REST API with API keys, signed webhooks, an integration playground, and x402 so AI agents can pay USDC on Base per API call, directly to your wallet.

Networks: Bitcoin, Ethereum, Solana, Polygon, Base, Arbitrum, Optimism, BNB Chain and Avalanche. Typical confirmations run about 0.4 seconds on Solana and around 2 seconds on Base and Polygon.

The honest boundary. Payzum is crypto-only. We accept crypto and settle in crypto. We are not an off-ramp, we do not move money into bank accounts, and we have no commercial relationship with Stable.com, Unlimit or Polygon Labs. We're writing about their announcement because it matters to the same merchants we serve — the fiat leg is a separate decision you make with a separate provider, and we'd rather you make it with clear eyes than take a recommendation from us.

What a fully non-custodial chain looks like, step by step

1. Own the destination. Create or connect the wallet you want revenue to arrive in. That address is the account — everything downstream inherits its custody model. Enable 2FA on the dashboard and keep the keys where your treasury policy says keys go.
2. Turn on acceptance where you sell. Hosted checkout or a payment link for online, the POS QR for the counter, invoices for B2B terms, subscriptions for recurring. Set optional auto-conversion to USDC or USDT if you don't want to hold volatile assets overnight.
3. Get paid, and watch it land. The customer pays; the transaction confirms in seconds; the funds are at your address. A signed webhook fires so your order system, ERP or accounting tool updates itself, and the audit log records every event.
4. Spend from your own wallet wherever you can. Supplier invoices in stablecoins, contractor and affiliate payouts by CSV or EVM batch, agent revenue arriving per call. Every dollar spent on-chain is a dollar that never needs an off-ramp at all.
5. Exit only what you must, through a route you've tested. For the fiat you genuinely need — rent, payroll, tax — use a licensed off-ramp, and prefer one that doesn't ask you to deposit first. Test it small, document the timing, and keep a second provider configured before you need it.

Who this changes something for

The businesses that feel this most are the ones where the exit hop is frequent, large, or politically fragile.

• Exporters and cross-border service firms. You invoice abroad and get paid in stablecoins in minutes instead of waiting on correspondent banking. The exit was where the speed advantage got handed back — a shorter custodial window is a direct improvement to working capital.
• Operators in high-inflation or capital-control markets. Holding a dollar-denominated balance you control is the point. Every day that balance is a claim on a platform instead of a token in your wallet is a day the strategy isn't actually running.
• Agencies, freelancers and contractors paid from abroad. Small, frequent conversions are precisely where custodial minimums, fees and review queues hurt most. See getting paid from abroad without a traditional bank account.
• Businesses with heavy payout volume. Affiliates, prize pools, creator revenue shares, contractor networks. If most of your outflow is stablecoins going out anyway, the off-ramp is a thin edge case rather than a core dependency — which is the healthiest place for it to be.
• Anyone who has already been de-risked once. If a bank or processor has closed your account before, you know the value of a chain with fewer places to be told no.

Custodial route vs non-custodial route — the comparison

The honest counter-arguments

"Self-custody moves the risk onto me, and I'm not a crypto security team."

Correct, and we won't pretend otherwise. Non-custodial means the failure modes change rather than disappear: lost keys, a mistyped address, an employee with too much access. Those are real, and a business that can't operate key management responsibly is genuinely safer with a custodian.

The trade is a known risk you control against an unknown risk you don't. Key management is an engineering problem with established answers — multisig, hardware signers, role separation, a documented recovery procedure. A platform's compliance decision about your account is not a problem you can engineer around, because you are not the one making it. Most businesses conclude that the first problem is the one they'd rather own.

"My accountant needs a bank statement. This is all academic."

You will still need fiat, and you will still need an off-ramp — that's precisely why we wrote about someone else's product today. The realistic goal isn't zero fiat. It's making the fiat leg small and scheduled rather than large and constant, so it becomes a routine treasury task instead of a chokepoint your revenue sits behind.

On the reporting side, the ground is moving in a useful direction: FASB's August 2026 proposal on whether stablecoins count as cash equivalents is exactly the question your accountant is asking, and it's being answered in public.

"If KYC still applies at the exit, what did self-custody actually buy me?"

Time and exposure. KYC is an identity check that happens once and persists. Custody is a window during which your money is someone else's liability — and that window is where balances get frozen, delayed, netted against fees, or caught in an insolvency. Shrinking it from days to the duration of a single transaction is a meaningful reduction, even though the identity check is unchanged.

There's a second answer that's easy to miss: optionality. Money in your wallet can go to a supplier, a contractor, a different off-ramp or a card programme — the a16z data we covered on stablecoin spending on crypto payment cards shows how much of it never converts at all. Money in a platform account can only do what that platform allows.