A year after Visa released its white paper on on-chain lending, it has finally transitioned from a payment network to a credit network

Author Charlie

On September 8, U.S. time, just as everyone was slowly returning to the office from the Labor Day holiday, Visa got everyone energized.

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

If you only see the keywords "Visa, stablecoins, on-chain lending," this news can easily be interpreted as Visa simply embracing stablecoins once again. After all, in recent years, from USDC clearing, stablecoin cards, to various on-chain payment experiments, Visa has frequently appeared in similar news.

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

Because the product launched today is almost a sequel to that white paper a year later.

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

The most interesting aspect of these cases at that time was that they began to answer a long-standing question troubling crypto lending: If on-chain lending can only rely on over-collateralization, meaning you need to deposit $150 to borrow $100, it can certainly become an efficient leveraged trading market, but it is very difficult to truly enter the realm of corporate financing in the real world.

The reason companies need credit is that they have future cash flows but lack sufficient cash today. If a company already has $150 in liquid assets, its need for that $100 loan is completely different from a company relying on future income for turnover.

Visa's white paper last year actually pointed out another path: don't just focus on crypto collateral on-chain, but turn the payment receivables, cross-border cash flows, and clearing obligations from the real business world into assets that on-chain lending institutions can understand, verify, and control.

Rain needs to complete Visa card clearing every day, thus requiring short-term turnover funds; Credit Coop turns future card receivables into programmable collateral and repayment sources; Huma Finance further extends a similar model to cross-border payment financing, supplier payments, and trade financing.

At that time, Visa was more like an observer. It compiled these new models into an industry research report, telling everyone that the value of stablecoins should not be limited to payments.

A year later, Visa has stepped into this map itself.

This is also the truly noteworthy aspect of today's news.

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

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

At the same time, Credit Coop's Spigot smart contract controls the corresponding clearing receivables. Once the money comes back, it will first repay the loan according to a pre-set order, and only the remaining portion will return to the borrower's account.

If you strip away the crypto packaging, this is not a brand new financial invention.

It is essentially still structured financing. Traditional finance already has mature mechanisms like lockbox accounts, cash pooling, accounts receivable financing, and borrowing bases.

The real change is not that financial principles are being reinvented, but that a set of processes that previously required banks, lawyers, account control agreements, extensive manual reconciliations, and periodic reports is now becoming a programmable credit system that can read real transaction data daily, automate borrowing, automate repayment, and operate around the clock.

Because of this, I actually think it is more important than many DeFi products that attempt to "reinvent finance."

Mature financial markets have never lacked clever credit structures; what is truly lacking is how to lower a set of financing infrastructures that only large enterprises can afford to a level where a fintech company that has only been established for a few years and needs only a few million dollars in turnover funds daily can also use it.

This is precisely a very realistic problem faced by the stablecoin card project now.

Suppose a card company’s users spent $1 million today; it needs to pay this amount according to Visa's clearing schedule, but the corresponding funds may not have synchronized back to the company’s account. This time lag creates a liquidity demand that needs to be resolved every day.

For mature card issuers like JPMorgan or Capital One, this is hardly a topic worth discussing. They have their own balance sheets, can obtain bank credit, and can engage in large-scale accounts receivable financing, and once the business matures, they can even securitize.

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

It may only need two to three million dollars in funds daily, yet it needs to borrow frequently, repay frequently, and clearing still occurs on weekends. The scale is too small for a large bank or private credit fund to set up a complete credit, legal, account control, and post-loan management system specifically for it; yet the growth is too fast, and relying solely on cash on hand will severely drag down capital efficiency.

It’s not that traditional finance can’t do it.

It’s that it’s not profitable to do this business.

What Credit Coop has found is the gap that traditional financial cost structures 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, with over 3,000 borrowings and more than 9,000 repayments, and currently reports no defaults.

Rain is one of the earliest and most representative users. It started using this financing arrangement to support daily Visa clearing in 2023, and the cumulative clearing scale involved has approached $2 billion.

Looking back at Visa's white paper from last year, this growth rate is indeed 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 is now publicly disclosed is a real payment cash flow at the billion-dollar level.

So the real change happening today is not that Visa suddenly discovered on-chain lending, but that a set of models it observed and studied in last year's white paper has begun to transform into infrastructure that Visa itself participates in building.

At this point, looking at the relationship between Credit Coop, Rain, Huma, and Visa, you will find that they are not simply competitors, nor is it a matter of who will replace whom, but rather they occupy different positions on the value chain.

Rain is closest to the source of payment business.

It helps fintech, crypto wallets, and other platforms issue Visa cards, responsible for card projects, fund management, authorization, clearing, and a series of infrastructure. The core asset it accumulates is not a specific lending protocol, but an increasingly large real payment scale.

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 treated as credible assets by lenders; how lenders can gain sufficient control; how money is lent out and when it is collected back; and if the borrower encounters issues, whether the funds entering the system can first repay debts before being freely allocated by the company.

Thus, Credit Coop increasingly resembles a programmable structured financing platform for payment receivables.

Huma is another direction.

From initially providing stablecoin loans to payment companies to later integrating with Arf, Huma is clearly developing towards a broader PayFi market. It not only provides funds for card settlement but also aims to cover various short-cycle, real payment cash flows such as cross-border payments, trade financing, and supplier payments, connecting these assets with on-chain and institutional funds.

When Visa's white paper was disclosed last year, Huma's monthly transaction scale was about $500 million, with active liquidity around $140 million, of which active PayFi loans approached $100 million. By August of this year, Huma disclosed that its cumulative transaction scale had exceeded $17 billion, with over $1 billion added each month.

Thus, Credit Coop and Huma will certainly have increasing overlap in the future.

Both are essentially doing the same big thing: how to turn relatively certain future payment cash flows into funds that can be used today.

However, their entry points at this stage are still somewhat different.

Credit Coop is more focused on delving into the very specific scenario of Visa card clearing, tightening the clearing data, receivables control, and repayment processes.

Huma, on the other hand, is expanding outward, hoping to incorporate more different types of payment scenarios into PayFi and further connect to broader funding sources.

One is more vertical, the other is more horizontal.

In the short term, Rain appears to be their common customer and partner. The larger Rain's payment scale, the more turnover funds it needs, and the more it hopes for a richer variety of financing sources, rather than relying on just one.

The real entity standing above these companies is Visa.

Because Visa possesses something that is difficult for others to replicate: who has actually conducted how many transactions, how much needs to be cleared today, and whether past clearings have been completed on time.

This is the ground truth of payments.

In traditional finance, credit analysis fundamentally spends a large part of its costs on verifying information.

A company claiming to have $10 million in revenue each month will not convince a lending institution just because the CEO confidently states it in a meeting room. They will require bank statements, financial reports, audits, accounts receivable details, and will constantly verify and periodically require borrowers to resubmit data.

This is the most basic issue in credit business: borrowers always know their real situation better than lenders.

Thus, the truly expensive cost for lending institutions is often not coming up with a sophisticated model, but confirming whether the data fed into the model is real.

What Visa is doing today is making this step suddenly simple.

Credit Coop does not have to rely entirely on Rain or other card projects to tell it "how many transactions we have today" or "how much we should receive tomorrow." Instead, it can directly read Visa's settlement records after authorization.

If this direction continues to develop, a large part of the credit analysis process that originally required manual verification will gradually transform into data exchange between machines.

The significance of this matter may even be greater than "loans are issued in USDC."

It also allows DeFi to move from over-collateralized loans to cash flow loans, presenting a more pragmatic path for the first time.

The most successful lending model in DeFi in the past was actually very simple: I don't know who you are and don't want to investigate whether you are reliable; as long as you put up $150 worth of assets here, I can confidently lend you $100. If the price drops, the smart contract automatically sells the collateral.

This model is very suitable for trading markets because it greatly reduces credit risk and operational costs.

But it is not suitable for the vast majority of real-world enterprises.

When a company truly needs a loan, it is often precisely because it has future income but does not have enough money today. What it needs is to convert future cash flow into today's liquidity, rather than exchanging more existing assets for less cash.

This is also the issue that so-called RWA and institutional DeFi have been grappling with over the past few years.

How to do low-collateral or even no-collateral loans?

Some have tried to create on-chain identities, some have developed credit scoring systems, and some have established reputation systems based on wallet history. But these methods will soon encounter a huge challenge: real business activities are too complex, and a wallet address cannot explain a company's true operational situation.

Visa today provides a narrower but more realistic answer.

There is no need to first solve the on-chain credit for everyone in the world.

As long as Visa itself knows that this settlement obligation is real, and Credit Coop can control the corresponding receivables, then this real cash flow itself is sufficient to become an important basis for credit.

So I like to summarize this matter with a sentence:

The data of traditional finance begins to connect with on-chain funds.

But going a step further, it is not just a matter of funds.

Because this system also makes data, funds, and repayment processes programmable. Daily real transactions can be read, limits can be adjusted with business changes, borrowing can occur automatically, and repayments can be made automatically according to rules once income returns.

A loan that originally had no bank willing to undertake due to its small scale and cumbersome operations now has the opportunity to become economically valuable.

This is the true value that I believe stablecoins create here.

Not because they are "decentralized," but because they can operate 24 hours a day and are inherently connected to smart contracts, making the time granularity of cash flow and business flow closer.

Of course, there is also a very worthy research question: if these companies nurture their customers to maturity, will the customers still need them?

Karta is a very interesting case.

Karta initially obtained this type of settlement financing through Credit Coop, continuously borrowing and repaying small amounts, accumulating its credit record in real business. In June of this year, it subsequently received a $125 million institutional credit line.

This path has given me a completely different understanding of on-chain credit compared to "disrupting traditional banks."

It may not necessarily be a substitute for traditional credit; rather, it could be a kindergarten for traditional credit.

A newly established company, without three years of audited financial statements, without a sufficiently large asset scale, and without the qualifications for large institutions to invest significant manpower for due diligence, can first use on-chain financing to get started, accumulating thousands of real borrowing and repayment records over two to three years.

Once the data is sufficient and the scale is large enough, it can then "graduate" into the traditional institutional financing market with lower costs and larger limits.

If this path is ultimately proven valid, it will create a very interesting division of labor.

Companies like Credit Coop will be responsible for serving those customers that traditional financing currently finds too small, too new, or too troublesome, helping them establish verifiable credit histories; banks and private credit funds will then take over once these customers grow in scale.

This is actually very similar to the development of the payment industry.

A mature financial ecosystem is usually not about one company swallowing all upstream and downstream players, but rather many companies interlocking at different stages.

Rain is responsible for card issuance and payment traffic, Credit Coop handles credit control and post-loan management, Huma provides broader PayFi liquidity, traditional banks and private credit provide cheaper large funds, while Visa offers networks, data, and rules.

Everyone will compete, but also need each other.

The difference between "competition is greater than cooperation" and "cooperation is greater than competition" is something I have always felt is very important for understanding the fintech ecosystem in Europe and America. Truly valuable companies do not necessarily do everything; instead, they find the segment of the value chain where they have the greatest comparative advantage and become the part that is hard for others to bypass.

From this perspective, Visa's actions today have a larger strategic significance.

For the past decade, the crypto industry has loved to tell a story: blockchain will eliminate payment intermediaries like Visa.

The current outcome may seem completely the opposite.

Visa does not necessarily have to continue controlling the channel through which every dollar truly moves.

Some money can flow through Solana, some through Ethereum, some as stablecoins, and some may be the bank's own tokenized deposits. Visa itself has also increasingly taken the initiative to connect with these new funding networks.

But Visa can try to control another layer of more valuable things: trust and data.

Who is the real consumer?

Who is the legitimate AI agent?

Was this transaction authorized?

How much can this payment voucher spend?

How many payments has a card program actually made today?

Has it completed settlements on time for the past 365 days?

When this information comes from Visa, even if the money itself does not necessarily run on the traditional VisaNet track, Visa can still become a very important layer in the financial system.

It is gradually transforming from a payment network into a trust and data network.

And this connects to another topic I have been thinking about recently: Agentic Finance.

In the past year, when people talked about Agentic Commerce, the most easily imagined scenario is: in the future, AI can help me buy things.

I tell the AI, I'm going to New York for three days next week; based on my itinerary, find a hotel that costs no more than $500 per night, preferably close to the client's office. It searches, compares, checks reviews, and finally helps me book.

Going further, it can even contact the hotel, change flights, book restaurants, and finally complete the payment with a virtual card authorized by me.

These capabilities are already very close to us.

Karta's own AI concierge is already doing similar things. It can search, make calls, send emails, handle travel needs, and generate one-time virtual cards to complete transactions.

Visa has also done a lot of infrastructure work around Agentic Commerce in the past year.

The core issues are simply a few: how do merchants know that the entity accessing the website is a legitimate AI agent and not a malicious bot; how to prove that this agent is indeed authorized by the user; can it make payments; how much can it spend; and if something goes wrong, who is responsible.

These issues address Agentic Payments.

However, if AI agents in the future do not just help me buy plane tickets but also start helping companies operate, then "how to pay" will quickly become an insufficient question.

The real financial issues that companies face every day are never just "should I spend this $100."

Rather, it is where the money comes from.

Suppose in the future, a company's AI agent manages cloud service costs, advertising spending, vendor payments, travel, and procurement simultaneously. Today, there is $700,000 in the account, but $1.2 million needs to be paid in the afternoon; tomorrow, $900,000 in receivables will come back.

What should it do?

Should it borrow $500,000 or $600,000 today?

Which of the two credit lines should it choose?

One lending institution has a slightly lower interest rate, but the other can disburse funds in real-time on weekends; which is more cost-effective?

Should today's incoming cash be used to pay off debts immediately, or should a portion be kept as a safety cushion?

If business volume grows by 20% next month, should the credit limit be automatically adjusted upward?

These are no longer payment issues.

This is cash management, credit, and balance sheets.

This is Agentic Finance.

I believe this is precisely where today's Visa news is most worth thinking about in the long term.

Because if future payment data can be read by machines, credit limits can be called by programs, and repayment rules are written into smart contracts, then what AI agents can ultimately manage is no longer just a single transaction, but an entire set of liquidity.

The significance of this step far exceeds "AI helping me buy coffee."

Payments determine how existing money moves.

Credit determines whether future money can be used today.

And most of the growth in the modern economy is not entirely driven by existing cash.

The housing market relies on mortgages, credit card spending relies on revolving credit, cross-border trade relies on trade finance, and corporate expansion relies on bank credit and bonds. The economic system we take for granted today is essentially built on various forms of "early use of future cash flows."

So if stablecoins ultimately only move one dollar from point A to point B faster, they change the payment infrastructure.

But if future cash flows can also be read, analyzed, financed, and automatically repaid in real-time, then it begins to enter not just payments but also balance sheets.

This is why I believe today's Visa product is much deeper than "Visa supports stablecoins."

Of course, we must seriously face a counterargument: does all of this really need blockchain?

The answer is actually not as simple as the crypto industry likes to imagine.

The entire process could theoretically be placed back into the traditional financial system. Visa provides real-time settlement data through APIs, banks dynamically adjust limits based on this data, disburse funds through instant payment systems, and then repay through automatic cash pooling.

As bank APIs, real-time payments, and tokenized deposits become increasingly mature, many capabilities do not necessarily have to run on public blockchains.

Moreover, just because smart contracts can control funds in an on-chain account does not mean that real-world bankruptcy laws, creditor priorities, accounts receivable ownership, and fraud issues automatically disappear.

Code can automatically distribute money.

Courts are still courts.

So today, this matter cannot simply be summarized as "DeFi is replacing traditional credit."

On the contrary, I believe it shows that the most realistic commercial value of blockchain is often not inventing a completely new financial system, but rather reducing operational friction to a sufficiently small level in areas where the cost structure of traditional finance cannot reach.

When a cash flow is already digital, high-frequency, and occurs around the clock, while the loan size is not large enough for banks to build a whole artificial system, programmable credit becomes particularly valuable.

This may not sound as sexy as "disrupting global finance."

But truly significant financial infrastructure often grows out of these rather mundane little problems.

In the coming year, I will pay special attention to three things.

First, will Visa gradually transform its connection with Credit Coop into a more open credit data layer?

If in the future not only Credit Coop but also Huma, banks, private credit institutions, and even more on-chain lending platforms can read Visa-verified payment and settlement records with customer authorization, then Visa's role will undergo a significant change.

It will not just handle payments.

It will start providing credit data.

By that time, VisaNet's transaction data could even become a new credit infrastructure.

Conversely, if more than a year later this model still only exists as an isolated case with Credit Coop and no more lending institutions join in, I will significantly lower my judgment on it.

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

Credit Coop cannot always just do Visa card settlement, and Huma will not be satisfied with just cross-border payment financing. Both companies are likely to expand into more merchant receivables, card settlements, and corporate payment scenarios in the future.

However, I don't think there can only be one survivor in the end.

Sources of capital, customer acquisition, data, credit assessment, fund control, and post-loan management are fundamentally different capabilities. In the future, it is more likely to form a multi-layered cooperative relationship like today's payment industry, rather than one protocol eating everyone else.

Third, and this is what I am most interested in, Agentic Finance is likely to mature one or two years later than Agentic Payments, but the value it ultimately creates may be greater.

Having an AI agent spend $20 for you already requires solving authorization, fraud, and disputes.

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

Therefore, the real breakthrough point for Agentic Finance will not be when ChatGPT, Claude, or Gemini suddenly become particularly good at shopping.

The real watershed moment is when companies are willing to turn their financial policies into rules that machines can understand.

How much can be borrowed at most.

What funds can be used as collateral.

How much cash must be kept in the account at a minimum.

Under what circumstances can the limit be increased.

What interest rate levels are acceptable.

How much money needs to be prioritized for debt repayment after receiving it.

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

It does not need to "own" the company's money.

It only needs to have the authorized decision-making power.

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

Last year, Visa was still just observing in a white paper how Credit Coop, Rain, and Huma were trying to turn payment flows into credit.

A year later, it has started to genuinely integrate VisaNet's own data into this system.

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

But the direction is much clearer than it was last year.

In recent years, the biggest story of stablecoins has always been payments.

What is more worth watching next may be credit.

Because payments solve how money moves.

Credit determines whether future money can be used today.

If this step is truly successful, the second half of stablecoins will no longer just be about payments.

But rather about balance sheets.