The Netflix Question

Banks are evaluating tokenized money as a project decision. The evidence says it's may be more like a business model decision.

Jul 26, 2026

My bet is that if you walk into most banks running a tokenization workstream, then you will find some kind of a project. Should we offer a stablecoin, should we tokenize deposits, which chain, which custody vendor, which wallet? Senior leadership routes the question to project owners, the project owners scope an RFP, then a pilot, and the pilot competes for funding against every other initiative in the planning cycle. I mean to all of us bankers this process looks and feels responsible. But I do wonder if it might be the most expensive category error available to a bank right now, because whats arriving is really not a project, and treating it as one almost guarantees you’ll answer the wrong question well.

I wrote earlier this year that tokenization is not a product, it’s infrastructure. I still believe that, but I’ve come to think “infrastructure” is also incomplete. My challenge is that the infrastructure framing keeps a bank inward focused, debating its own architecture while the actual transformation happens to its customers, its revenue pools, and its funding model. So let me have a go at laying the argument out properly. First, what is moving, then why the cash leg has to follow, and then the part that doesn’t fit in either the product box or the infrastructure box.

The asset side is already moving

Larry Fink made the call explicit at the New York Times DealBook summit at the end of 2022

“the next generation for markets, the next generation for securities, will be tokenization of securities.”

BlackRock backed it with product. BUIDL, its tokenized money market fund, launched on Ethereum in March 2024, and it became the largest tokenized Treasury fund within six weeks, and was the first to cross $1 billion in March 2025. It is now multichain, holds around $2.5 billion, and the category has grown crowded enough that Circle’s USYC has since taken the top spot.

When you think about the institutional roll call it has actually stopped being interesting because its stopped being surprising. Franklin Templeton’s BENJI is in production. Apollo has tokenized credit through Securitize, KKR has tokenized private equity through the same platform, and Hamilton Lane has tokenized private markets funds. DTCC moved its Tokenization Service from concept into live production this month, with a scalable launch targeted for October. JPMorgan’s Tokenized Collateral Network is moving repo collateral, and Kinexys, its broader blockchain platform, processes over $5 billion per day. Goldman, Citi, BNY, and HSBC have all built tokenized issuance and settlement infrastructure. Citi projects $4 trillion in tokenized assets by 2030, and BCG’s number is higher.

Funny. Because no single name on that list matters that much, but the breadth does. Treasuries, money market funds, equities, private credit, private equity, repo collateral, and structured products are all tokenizing at the same time, at the largest asset managers in the world, on a timeline measured in years rather than decades. This was the epiphany for me! The speed of whats happening in capital markets. Bankers who treat that as a separate system from the one they operate in are defending a very strange position, because the cash that settles those markets sits in their institutions.

The cash leg follows or the trade breaks

When a tokenized asset settles in seconds onchain, its payment leg has to settle on the same timeline or the trade does not happen atomically. I think that asymmetry has only three possible resolutions.

Prefunding works in principle. The buyer parks dollars in an escrow account days before the trade, capital sits idle, the trade is slow, and the operational overhead is real. Most institutional tokenized trades currently do this, but its exactly what BlackRock, Apollo, and the others are trying to engineer out of their products.

Principal risk is the alternative to prefunding and thats off the table. The buyer pays and waits for the asset, or the seller delivers and waits for payment. This is the failure mode the entire post Lehman settlement reform agenda was built to eliminate, and reintroducing it at institutional scale has gotta be a nonstarter.

That leaves tokenized cash on the same rail as the tokenized asset, settling atomically. The asset and the payment clear in a single transaction or neither does. This really is the only resolution that scales, and it is what every major institutional design now assumes. BUIDL redeems into USDC through Circle. JPMorgan’s collateral network settles against the bank’s own deposit token. Project Agorá demonstrated tokenized commercial bank deposits clearing against tokenized central bank reserves in May 2026. The cash leg cannot stay on ACH and Fedwire while the asset leg moves to atomic onchain settlement. Either tokenized cash exists at scale, or institutional tokenization stops. Institutional tokenization really does not look like its stopping anytime soon.

So far, this is the infrastructure argument. The reason it’s still incomplete is that I believe infrastructure transitions of this kind don’t stay infrastructure transitions, and I reckon there’s an interesting precedent that illustrates how.

The Blockbuster category error

Ok, so a big hat tip to Elizabeth St-Onge of TD, who riffed on Blockbuster and Netflix with me. I think the analogy is good here because it describes a specific failure of categorization rather than just a generic story about innovation, and perhaps the history is a little more pointy than the version we all usually hear.

Blockbuster and Netflix were both in the video rental business, and for its first two years Netflix even ran Blockbuster’s model by mail. So pay 4 bucks a title, due dates, late fees. Then in September 1999 it did a new thing when it introduced a flat monthly subscription with no due dates and no late fees, and by early 2000 it had dropped per title pricing entirely. Now, that was not only a distribution decision. It feels to me like it was a repricing of the whole customer relationship, and it was aimed squarely at the most profitable line in Blockbuster’s P&L. In 2000, Blockbuster collected nearly $800 million in late fees, around 16 percent of its revenue! Friction income.

When Blockbuster finally killed its own late fees in January 2005, walking away from hundreds of millions a year in the name of competing, the customers it was trying to keep had already learned there was a better deal elsewhere. Streaming, when it arrived in 2007, was actually the second act, and within a decade of it Netflix was making the movies rather than shipping them. Blockbuster, which understood the rental business better than anyone alive, went bankrupt in 2010.

I think its worth noting the order of operations here, because to me it’s the part of the analogy thats super interesting. The business model attack came first, on identical infrastructure, with DVDs in the mail. The infrastructure shift came second and finished the job. To call the whole thing an infrastructure transition really misses it I think, and I do wonder if the same misfiling is happening in banking, with the same sequence. Late fees were revenue the customer paid for friction they didn’t choose. And if you haven’t picked up the theme yet, so is float.

Lets have a go at running the parallel properly. Theres no doubt the rail change is real, atomic settlement instead of batch, always on instead of banking hours, programmable instead of static. But I think the rail change triggers a behavior change, and that behavior change is where bank P&Ls actually live. Probably the most basic question in commercial banking is why does a corporate keep idle cash in a non interest bearing account? The honest answer is friction. Moving money is slow, expensive, and operationally risky, so the idle balance is rational. Atomic settlement has the potential to delete that friction. If a corporate has $100 million sitting in an account for an hour, and it can move into a tokenized money market fund and back with no settlement risk, it will. If you meet this with skepticism, then ask yourself if you were advising a large corporate on liquidity, under what scenario would you recommend they leave that money idle? Honestly its hard to find someone with a good answer to that!

Now lets follow the consequence inside the bank. If stable idle balances disappear from the large corporate portfolio, the funds transfer pricing models that allocate value across every business line stop describing reality. The deposit franchise economics that subsidized commercial banking for a century get repriced, and this is already visible in the data. Non interest bearing deposits at US banks have fallen more than 30% since the March 2022 peak, before tokenized alternatives paid a single basis point of yield. Money market funds hold $7.6 trillion. A 2026 Tradeweb survey found 25% of corporate treasurers moderately or very interested in tokenized money market funds, in a market where fewer than 5% of treasurers held any digital asset the year before. I’m afraid that the repricing is happening whether banks act or not. Late fees went away for Blockbuster too, the difference is Netflix replaced them with something bigger.

That’s the category error I’m worried about in most bank tokenization programs. They’re project programs, run by project people, scoped as project decisions. But actually the right question list looks nothing like a list of project milestones. Which of our revenue pools exist because settlement is slow? Which client behaviors change when money moves like information? What replaces the income we lose, and are we building toward those pools or defending the old ones? Thats a war game about the identity of the institution, and I’m not sure how many banks are running it.

When the customer is an agent

There’s a second behavior change stacked on top of the first, and I think its has the potential to compound everything above. The economic actors that operate on these rails will increasingly not be humans.

Take the owner of ten dry cleaners. They’re never going to learn what a tokenized deposit is, and really they don’t need to. They will tell an AI agent to optimize their cash. So pay suppliers as efficiently as possible, collect receivables as fast as possible, and make sure no dollar sits idle when it could be earning. The agent executes against those instructions continuously, and an agent has no sentiment, no inertia, and no relationship with a branch manager. It will sweep every idle balance every hour of every day, because thats literally what it was told to do. Small businesses, the customers with the least treasury sophistication, have the most to gain from this, which inverts the usual adoption logic that says complexity arrives at the top of the market and stays there.

I think we’re starting to see glimpses of this now. Mercury has shipped a command line interface to the bank account, along with a server built for AI agents to call it, and if there’s a command line, there will be agents driving it. I’ve connected an LLM to QuickBooks and a bank account to reconcile transactions myself, it works today. And the killer app may not come from a bank or a fintech at all. SAP sits in the back office of essentially every Fortune 500 company, and every receivable in that network is someone else’s payable. If SAP nets those flows across its installed base on tokenized rails, the volume that banks believe they intermediate gets settled before it ever touches a bank. Banks like to think they sit at the center of corporate payment flows. The ERP vendors actually do.

The pattern is not new. Every new technology gets used first to do the old thing faster, and then someone does a net new thing. Mobile phones were portable phone calls for a decade before nobody used them for calls at all. It feels like tokenized money is in its faster phone calls phase, cheaper cross border, faster settlement, and the net new phase, autonomous treasury, machine speed liquidity, conditional payment flows wired into physical logistics, is where the new revenue pools form. Vantage Bank’s POC work this year surfaced what that looks like at street level when a Texas trucking company that pays drivers the moment a delivery is confirmed in Mexico told the bank this lets them retain their best drivers, and a manufacturer said faster supplier payment wins them better suppliers. I love that because its a competitive position story rather than a cost story, and its exactly the kind of value a project committee never finds because no customer knew to ask for it.

The existential math

The cleanest test for whether a bank should treat this as existential is to compare it against the product misses that weren’t. Plenty of large banks sat out P2P payments for years while Venmo took the market, then spent the better part of a decade clawing it back through Zelle, and it stung, but it was survivable, a missed feature, and the banks carried on. Business models don’t miss you the way products do. If I were on a bank board, then the question about tokenized money I’d be asking management is whether this one is survivable if they’re wrong? I’m not sure it is.

Jamie Dimon is a good example. He has been publicly skeptical of crypto for a decade, and for that same decade JPMorgan has been building Kinexys, JPMD, and the Tokenized Collateral Network anyway. Skepticism and a hedge are not contradictory, in fact they’re a pretty rational pair. If there is even a 10% chance the bears are wrong about this, you cannot be out, because the downside of being out is being the next Blockbuster.

And the uncomfortable structural fact is that the banks with the most to lose are not all moving. The mega banks are building, and four of them are now standing up a shared tokenized deposit network with The Clearing House, targeted for 2027. The super regionals are joining consortiums like Cari Network and Project Keystone, often both at once, which tells you they’re hedging across rails because they can’t afford to bet on one. But the middle is waiting, and the long tail of community banks will get whatever their core vendor ships, whenever it ships. Meanwhile the combined market value of the Big Three core vendors has fallen from roughly $140 billion to about $60 billion in two years, which means the layer those banks are waiting on is itself under a bit of stress. A bank whose tokenization plan amounts to “whatever the core vendor ships” has outsourced its future to a balance sheet under pressure.

I think banks end up choosing among three plays. Either become an infrastructure provider, the financial market infrastructure of the tokenized system, or become the white glove aggregator that assembles tokenized products into something customers actually want and will pay for, or become a specialist liquidity provider into the new flows. All three feel viable. What is not viable is treating the choice as a project backlog item and revisiting it next planning cycle, because the FTP model ain’t going to wait for the roadmap.

Netflix makes far more money now than it ever did renting videos. The revenue pools that died were replaced by bigger ones, captured by the firm that read the shift as a change in what the customer was buying rather than a change in how the tape got delivered. I reckon the same will be true here. Settlement revenue, float income, and correspondent fees will shrink, and machine speed treasury, tokenized asset servicing, agent facing financial products, and the infrastructure underneath all of it will grow. The money doesn’t disappear, it moves to whoever rebuilt around the new behavior.

Blockbuster knew about Netflix the whole time. It watched the subscription model take its late fees, it watched streaming take the rest, and in 2000 it passed on buying the company for $50 million. Knowing wasn’t the constraint. The constraint was that it kept filing an extinction level business model shift under new distribution formats, one more project decision in a quarterly review.

If this resonated, the earlier pieces in this arc are Tokenization Is Not a Product and The End of Idle Money.

References

Tokenization and institutional adoption

The Blockbuster record

Deposits, treasurers, and funding

Settlement and policy

Bank, vendor, and tooling moves

Atomic Settlement

Discussion about this post

Ready for more?