On September 8, in American time, as everyone slowly returned to the office from Labor Day holiday, Visa got everyone's attention.

It announced a new on-chain lending model: providing VisaNet's clearing data to on-chain lending institutions, allowing them to use stablecoins to provide the necessary clearing funds for the rapidly growing stablecoin card projects daily.

If one only sees the keywords “Visa, stablecoin, on-chain lending,” this news can easily be interpreted as Visa simply embracing stablecoins once again. After all, in the past few years, from USDC clearing, stablecoin cards, to various on-chain payment experiments, Visa has often appeared in similar news.

But my first reaction upon seeing this news was to recall the white paper titled “Stablecoins Beyond Payments: The Onchain Lending Opportunity” that Visa released a year ago.

Because the product announced today is nearly a sequel to that white paper one year later.

When I read that report last year, what impressed me most was not the well-known DeFi lending protocols like Aave and Morpho, but several very specific company cases in the latter half: Credit Coop, Rain, and Huma Finance.

The most interesting part of these cases at that time was that they began to answer a question that had troubled crypto lending for a long time: If on-chain lending can only rely on over-collateralization, meaning you have to put in $150 to borrow $100, it can certainly become an efficient leveraged trading market, but it is very hard to genuinely enter real-world business financing.

The reason enterprises need credit is that they expected cash flow in the future, but do not have enough cash today. If a company already has $150 of assets that can be liquidated at any time, its need for that $100 loan is completely different from a company that relies on future income to keep its operations running.

Visa's white paper pointed out another way last year: Do not just focus on crypto collateral on-chain, but instead, turn payment receivables, cross-border capital flow, and clearing obligations from the real commercial world into assets that on-chain lending institutions can understand, verify, and control.

Rain needs to complete Visa card clearing daily, so they need short-term working capital; Credit Coop turns future card receivables into programmable collateral and repayment sources; Huma Finance further expands similar models into cross-border payment financing, vendor payments, and trade financing.

At that time, Visa was more like an observer. It organized these new models into an industry study, telling everyone: the value of stablecoin should not just remain in payments.

Today, a year later, Visa has stepped into this map itself.

This is also what makes today's news truly noteworthy.

What Visa brought out this time is not another blockchain payment channel, but one of its most valuable and hardest to replicate assets: the actual clearing data of VisaNet.

With customer authorization, Credit Coop can directly obtain daily Visa clearing documents from projects, and then combine these real payment and clearing records with borrowing and repayment history on-chain to determine how much money can be borrowed, when to disburse, and whether the loan has been repaid on time.

Meanwhile, Credit Coop's Spigot smart contract manages the corresponding clearing receivables. Once the money comes back, it will first repay the loans according to a predetermined order, and only the remaining part will return to the borrower's account.

If you take away the crypto packaging, this is not some new financial invention.

It is essentially still structured finance. Traditional finance has long had mechanisms such as lockbox accounts, cash pooling, receivables financing, and borrowing bases.

The real change is not the reinvention of financial principles, but rather that the processes that used to require banks, lawyers, account control protocols, a large amount of manual reconciliation, and regular reports can now gradually evolve into a programmable credit system that can read real transaction data daily, automate borrowing, automate repayments, and operate around the clock.

For this reason, I actually find it more important than many DeFi products that attempt to “reinvent finance.”

Mature financial markets have never lacked clever credit structures; what they truly lack is: how to reduce a set of financing infrastructure that originally only large enterprises could afford to a level that a fintech company established for just a few years, needing only a few million dollars in working capital daily, can also use.

This is precisely a very realistic problem currently facing the stablecoin card project.

Let’s say today a card company’s users swipe out $1 million; it needs to pay this amount according to Visa’s clearing timeline, but the corresponding funds may not yet have synchronized back to the company's account. This time difference in between creates a daily liquidity need that must be resolved.

For mature issuing institutions like JPMorgan or Capital One, this is hardly a discussion point. They have their own balance sheets, can obtain bank credit, and can do large-scale receivables financing, and after their business matures, they can even securitize.

But a rapidly growing stablecoin card company is in a completely different situation.

It might need only two or three million dollars in funds daily, but it needs to frequently borrow and repay, and even settlements happen on weekends. The scale is too small, making it not worthwhile for a large bank or private credit fund to specifically build a full set of credit, legal, account control, and post-loan management systems for it; yet growth is too rapid to rely solely on the cash on its balance sheet, as it would severely drag down capital efficiency.

This isn’t that traditional finance can’t do it.

It’s that doing this business isn’t cost-effective.

What Credit Coop has found is the gap that traditional finance's cost structure cannot cover.

From the data Visa announced today, this is no longer a small-scale experiment. Credit Coop's related model has completed over $2.5 billion in clearing financing, experiencing more than 3,000 borrowings and over 9,000 repayments, with no defaults reported so far.

Rain is one of the earliest and most representative users. It has been using this financing arrangement to support daily Visa clearing since 2023, with the total clearing scale now approaching $2 billion.

If we look back at Visa's white paper from last year, this growth rate is actually quite noteworthy.

At that time, Visa disclosed that by September 2025, Rain had borrowed and repaid over $175 million in USDC through Credit Coop, while Credit Coop's monthly loan scale had just exceeded $30 million. In less than a year, what has now been disclosed is a real payment cash flow at the billion-dollar level.

So the real change happening today isn’t that Visa suddenly discovered on-chain lending, but that a set of models it observed and studied in last year's white paper is beginning to become an infrastructure that Visa itself is participating in building.

Looking at the relationship between Credit Coop, Rain, Huma, and Visa now, we can see that they are not simple competitors and that none will replace the other; rather, they occupy different positions in the value chain.

Rain is closest to the source of payment business.

It helps fintechs, crypto wallets, and other platforms issue Visa cards while managing card projects, funds, authorizations, clearances, and a series of foundational infrastructure. The real core asset it accumulates isn’t a particular lending protocol, but an increasing scale of real payments.

In other words, Rain is where cash flow is generated.

Credit Coop stands behind the cash flow.

It addresses the problem of whether these funds that occur daily and will return in the future can be considered trusted assets by the lenders; how lenders can obtain sufficient control; how money can be lent out, and when it must be returned; if the borrower runs into problems, whether the funds entering the system can first pay off the debt before reverting to the company's discretion.

So Credit Coop is increasingly resembling a programmable structured finance platform for payment receivables.

Huma is another direction.

From early stablecoin lending to payment companies, to later integration with Arf, Huma is clearly developing towards a broader PayFi market. It is not only providing funds for card settlements but hopes to cover various short-cycle, real cash flows such as cross-border payments, trade financing, and vendor payments, and connect these assets to on-chain and institutional funds.

When Visa disclosed its white paper last year, Huma’s monthly transaction scale was about $500 million, with active liquidity around $140 million, of which active PayFi loans were nearly $100 million. By August of this year, Huma disclosed that its cumulative transaction scale had exceeded $17 billion, increasing by more than $1 billion each month.

Thus, the boundaries between Credit Coop and Huma will increasingly become blurred.

Credit Coop cannot forever only do Visa card settlements, and Huma will not be content with only cross-border payment financing. Both companies will likely expand into more merchant receivables, card clearing, and corporate payment scenarios in the future.

However, I don’t think it must end up being that only one survives.

Source of capital, customer acquisition, data, credit judgments, funds control, and post-loan management are fundamentally different capabilities. In the future, it is more likely to form a multi-layered collaboration, similar to today’s payment industry, rather than one protocol swallowing them all.

Finally, and this is what I’m most interested in, Agentic Finance is likely to mature a year or two later than Agentic Payments, but the value it ultimately creates may be greater.

Having an AI agent spend $20 for you, we already need to solve authorization, fraud, and disputes.

Having it decide to borrow $2 million for a company is a completely different trust threshold.

So the real breakthrough for Agentic Finance won’t be when ChatGPT, Claude, or Gemini suddenly become particularly good at shopping.

The true watershed will be when companies become willing to turn their financial policies into rules that machines can understand.

How much can be borrowed at most.

Which funds can be used as collateral.

How much cash must be retained in the account.

Under what circumstances can the credit limit be increased.

What level of interest rate can be accepted.

What must be prioritized for repayment after receiving funds.

When these things become rules that machines can read and execute, the AI agent can truly begin to engage with a company's balance sheet.

It does not need to “own” the company’s money.

It just needs to have the authority to make decisions.

And credit, at its essence, is a form of authorization about the future.

A year ago, Visa was only observing how Credit Coop, Rain, and Huma were trying to turn payment flows into credit through a white paper.

Today, a year later, it is beginning to incorporate its own VisaNet data into this system.

This step is, of course, still far from a world where AI can borrow money, repay money, and manage a company’s balance sheet by itself.

However, the direction has become much clearer than last year.

In the past few years, the biggest story around stablecoins has always been payment.

What is worth watching next might be credit.

Because payment addresses how money moves.

Credit determines whether future money can be used today.

If this step really works, the second half of stablecoin will no longer just be payments.

But rather, it will involve balance sheets.