When a crypto payment provider shuts down overnight, custody decides who loses money

What happened: a crypto payment provider shut down mid-checkout

On July 29, 2026, Kulipa — a Paris-based company that issued white-label stablecoin payment cards for wallets and fintechs — ceased operations. Not "announced a wind-down for September." Ceased. Cards that had worked at breakfast were declining by the afternoon.

Kulipa never had end users of its own. It sat in the middle of the card value chain, handling the backend for other people's card products: processing, fraud management and settlement, so a wallet company could ship a card without becoming a card company. The Block reported on April 1, 2026 that it had raised a $6.2 million seed co-led by Flourish Ventures and 1kx, with White Star Capital and Fabric Ventures participating — bringing total funding to $9.2 million — off the back of 120,000+ cards issued since its February 2025 launch, 20 signed customers, and claimed 70% month-over-month transaction growth. Founder and CEO Axel Cateland described the company as "closing the last mile between stablecoins and everyday spending."

Four months later the last mile closed itself. Subsequent analysis by PANews attributed the shutdown to solvency problems and catalogued the fallout: the website offline, around 20 client platforms affected, no notice period. The market it served was not collapsing around it — the same reporting cites crypto card transaction volume of roughly $7.8 billion in May 2026, up about 230% year over year (per The Kobeissi Letter). A growing segment does not save an individual provider in it.

The part worth studying: why not one user lost a balance

Here is the detail that should change how you evaluate payment infrastructure.

Solflare co-founder Vidor Gencel posted publicly the same day that the issuing partner was winding down over solvency issues, apologised for how users found out — "a lousy way to find out," if your card declined at a till — and then made the only point that mattered: funds are safe and there is nothing you need to do.

Why? Because the card was self-custodial end to end. No deposit. No top-up. No balance parked with the card provider. The card spent straight from the user's own wallet at the instant of purchase. Gencel contrasted that with the standard model, where the issuer holds your balance — and when that issuer fails, your money is frozen behind a bankruptcy estate while you stand in line as a creditor.

Ready (formerly Argent) was built the same way, which is why both companies could state within hours that user assets were untouched. Roughly 120,000 cards stopped working. Zero balances were lost. The difference between an outage and a loss was one architectural decision made long before anyone knew there would be a shutdown.

Now translate that to a merchant's balance sheet

A consumer whose card dies has a bad afternoon and reaches for a different card. A business whose payment provider dies has a very different week, because a merchant's relationship with a processor is not "spend from my wallet" — it is "you hold my revenue for a while, then you send it to me."

Settled-but-unpaid revenue. Between the sale and the payout, someone else is holding your money. Every day of that cycle is a day of exposure to a company whose balance sheet you cannot see.

Rolling reserves. Higher-risk categories routinely have a percentage of volume withheld for months. That is, by construction, a large balance sitting with a third party.

Payout batches in flight. If you also pay out — affiliates, contractors, creators, winners — a provider failure freezes the outbound side too, and your obligations to those people do not pause because your provider did.

Operational whiplash. Even where money is eventually recovered, the immediate cost is real: checkout down, staff improvising at the counter, customers told to come back, support tickets, and an emergency migration to a replacement you had no time to evaluate.

None of that requires fraud or bad faith. It requires only that a company holding your money runs out of its own.

Why "we're regulated and well funded" is the wrong reassurance

The instinct after a shutdown is to look for a bigger, better-capitalised provider. That instinct is reasonable and insufficient, for three reasons.

Funding is a lagging indicator. Kulipa closed a seed round in the same year it stopped operating. Money in the bank at announcement time says something about investor conviction months ago; it says nothing about runway, unit economics, partner-bank costs or a compliance shock next quarter.

Risk hides in layers you never signed with. Modern payment stacks are stacked. Your provider may depend on an issuer, which depends on a BIN sponsor or partner bank, which depends on its own regulator's mood. Kulipa's clients did not choose Kulipa's dependencies — they inherited them. When you accept a custodial provider, you accept its whole supply chain.

Contracts allocate liability, not access. A strong agreement may determine who is eventually entitled to what. It does not put money in your account on the morning the platform goes dark. Recovery through an estate is measured in quarters.

Which is why the durable question is not how strong is this company? but what does this company's failure actually do to me? That answer is fixed by architecture, and you can check it before you sign anything.

What non-custodial actually means for a merchant

Payzum is a non-custodial, crypto-only payment processor. The customer's payment goes directly to a wallet the merchant controls. Payzum never holds, pools or controls the money — the settlement is the payment. There is no merchant balance in a Payzum account, no payout cycle to wait through, and therefore no pot of your revenue exposed to Payzum's corporate fortunes.

Run the Kulipa thought experiment on it. If Payzum stopped existing tomorrow, you would lose the tooling: the checkout, the dashboard, the QR generation, the webhooks, the reporting. That is a genuine operational problem and a migration you would rather not do. What you would not lose is a single unit of settled revenue, because none of it was ever in transit through us. Your Tuesday sales are in your wallet on Tuesday.

That is the whole point of the distinction, and it is worth being precise about it: non-custodial does not mean nothing can go wrong. It means the thing that goes wrong cannot be your money going missing with the vendor.

Concretely, the acceptance side looks like this:

• In person: a POS that generates a new QR per sale, so each transaction is its own request for its own amount. Any phone becomes a terminal, cashiers log in with PINs, and you get analytics per cashier and per terminal. No acquirer, no card-network fees, no chargebacks.
• Online: hosted checkout (redirect, modal or inline), no-code payment links and buttons, invoices with expiry and overpayment detection, recurring subscriptions, and donation/tip-jar flows — drop-in compatible with existing e-commerce plugins, snippets and webhooks.
• Volatility: accept what the customer holds, with optional auto-convert to USDC or USDT, so what lands in your wallet is a digital dollar. (See USDT vs USDC for payments.)
• Payouts: mass payouts by CSV (BTC/LTC/DOGE) and EVM stablecoin payouts on Polygon, Arbitrum, Optimism, Base, BNB Chain and Avalanche — so the outbound side isn't sitting in someone else's queue either.
• Networks: Bitcoin, Ethereum, Solana, Polygon, Base, Arbitrum, Optimism, BNB Chain, Avalanche. Typical confirmations: Solana ~0.4s, Base ~2s, Polygon ~2s.

Network fees on Base, Polygon or Solana are cents, and there is no acquirer in the path to add a percentage. // confirmar pricing actual

How it works, step by step

1. Point settlement at a wallet you already control. You supply the destination address (or one per network). Complete KYC in-product. You are not opening an account that holds your money — you are telling payment requests where to land.
2. Turn on the surfaces you sell through. POS for the counter, payment links for WhatsApp or Instagram, hosted checkout for the store, invoices and subscriptions for recurring clients. Choose networks and coins, and whether to auto-convert to USDC/USDT.
3. Take the payment. At the till, a fresh QR for that exact amount; the customer scans and confirms. Online, a hosted checkout or a link. Confirmation lands in seconds on Solana, Base or Polygon — and it lands in your wallet, not in a pending balance.
4. Reconcile. Signed webhooks push events into your systems, the REST API and integration playground cover anything custom, and audit logs, 2FA and per-cashier analytics let you close the day cleanly.

Notice what is missing from that list: a step where you wait for someone to release your funds.

Use cases: who this change actually protects

• The counter business with thin margins. A café, barbershop or auto shop cannot absorb a week without takings. With settlement landing in the merchant's own wallet per sale, a vendor outage means finding another way to generate a payment request — not chasing a receiver for last week's revenue.
• The cross-border service business. An agency, studio or consultancy invoicing clients abroad already knows the pain of money sitting in intermediaries. An invoice paid in USDC arrives in the wallet directly, with overpayment detection and expiry handled — no correspondent bank, no processor float.
• The operator that pays people out. Marketplaces, affiliate programs, creator platforms and prop firms carry obligations in both directions. Mass payouts from your own wallet mean a provider's problems never become your contributors' problems.
• The API provider selling to machines. If you monetise an API through x402, agents pay USDC on Base per call, and that USDC goes straight to your wallet. Payzum acts as the middleware in front of your existing endpoint — you configure the endpoint, your API key and a price, and payment is settled through an external facilitator (currently Coinbase's) before the paid call is proxied through. Per-call revenue with no accumulating balance anywhere.

Custodial provider vs non-custodial processor, when things go wrong

Common objections

"If Payzum can also disappear, what's the difference?"

Any company can disappear — that's the honest baseline, and it's exactly why the question is what a failure costs you, not whether one is possible. If a custodial provider fails, you lose the tooling and the money it was holding. If Payzum failed, you'd lose the tooling and migrate. The revenue is already yours, in a wallet whose keys we never had.

"Doesn't holding my own wallet just move the risk onto me?"

It moves the risk to a place you can actually control and insure against with process: hardware wallets or a multi-sig for treasury, a separate operating wallet, documented key custody, staff access rules. That's a security discipline with known best practice. Counterparty solvency is not something you can mitigate — you can only find out about it, usually late.

"We're small. Isn't this a problem for big merchants?"

Inverted. A large merchant has legal resources, negotiated terms and diversified processors. A five-person business with one provider and two weeks of settled revenue in flight is the one for whom a shutdown is existential. Kulipa's clients weren't reckless — they were normal companies that happened to build on the right architecture.

"Do I have to give up cards to do this?"

No. Payzum is crypto-only and doesn't replace your card acceptance; it adds a rail where the money reaches you directly and can't be reversed. Plenty of merchants run both and treat the non-custodial rail as the one that doesn't depend on anyone else's solvency.