Product makers have two ways to change the future: either tweak an existing product or start from scratch. Uber founders Garrett Camp and Travis Kalanick could have taken the incremental road—say, by approaching a local San Francisco cab company with a few ideas about how to make its dispatch system a tad more efficient. Instead, they created an entirely new product that has changed the way we consume transportation, in the process doing tremendous damage to the taxi industry.

Money has never had an Uber moment. It has always evolved through a series of tweaks to the existing model, or the money stack as I like to call it. An existing layer in the stack is refined, or a new one added on top of an old one. But the money stack itself never gets torn down and rebuilt from scratch. It’s the same stack that we had 500 years ago.

Zoom forward to 2038. It’s fun to imagine what the monetary system might look like if some sort of Uber moment were to upend everything. I don’t think that’s going to happen exactly. Given its long legacy, the money stack will probably still be around in 20 years. But there may be big changes ahead, including a reduced role for banks, the rise of fintech companies, and innovations like bitcoin that exist outside the traditional money system.

First, I want to flesh out what I mean when I say that money is a stack. Like the earth’s crust, over time one layer of monetary strata has been piled upon another layer piled on another. At the base of the money stack are central banks. They issue deposits and banknotes. All of the transactions initiated on other parts of the stack ultimately rely on the base.

Central banks weren’t always at the bottom of the stack. Through most the last millennia, precious metals in the form of coins served as the base layer. Paper instruments like notes and bills of exchange were erected on top of the coin layer. Paper was convenient—note holders no longer had to lug bulky metal around from place to place. If skeptical, they could always bring their banknotes back to the bank-teller for redemption into underlying gold or silver coins.

The banknote layer was originally dominated by private bankers, but by the 1800s and the early-1900s, government-owned banks—central banks—had pretty much co-opted the issuance of notes from the private sector. Central bankers then proceeded to slowly cut away their connection to the underlying coin level. The first snip occurred when World War I forced most nations to stop redeeming their notes and deposits with gold. The last strand was severed in 1971 when President Nixon closed the dollar’s gold window to other central banks.

If central banks occupy the bottom of our modern money stack, directly on top of them is the banking layer. Commercial banks plug themselves into the base layer of the stack by opening accounts at the central bank. This connection allows banks to transfer central bank balances amongst each other in real-time. Banks in turn issue their own IOUs, or deposits, to their customers. These IOUs repackage many of the base layer’s capabilities for use by individuals and businesses. Customers willingly pay bankers for this privilege since members of the public are not permitted to have a central bank account.

Take wire transfers. When I ask Citi to wire funds to a friend who banks at Wells Fargo, Citi doesn’t actually move my deposits to Wells Fargo. It calls up the Federal Reserve, where it has an account, and orders the central bank to transfer funds to Wells Fargo, which also has an account at the Fed. Only when the transfer of “base money” is complete will Wells Fargo credit my friend’s account. Under the hood, Citi and Wells have leased us their bottom-level access to the central bank.

A fancier re-packaging of central bank services is provided in the form of debit and credit card payments. When a card payment occurs, funds don’t actually move from the buyer’s account into the seller’s account. The debit and credit card systems are really just communications networks. Over the course of a day, these networks receive and process millions of transaction messages. The amounts in these messages are batched, tallied, and offset against each other in the evening, each bank being provided with an owed or owing amount. The next morning banks then go down to the bottom of the stack and use their accounts at the Federal Reserve to settle up. Only then do funds get credited to the accounts of sellers.

A whole ecosystem of non-bank financial firms have plugged themselves into the banking layer. In the U.S., this top-most layer is composed of financial technology firms, or fintechs, such as PayPal, Dwolla, Venmo, and Square Cash. In other parts of the world, mobile money operators are significant players in the non-bank level of the stack, including M-Pesa in Kenya, Paytm in India, and Alipay in China.

Take PayPal, for instance. It keeps an account at Wells Fargo, a commercial bank, re-packaging Wells Fargo’s transactional services for PayPal customers. Remember that Wells Fargo is in turn repackaging the Fed’s services. So a customer with a PayPal account stands at the dizzying top of the stack, consuming services distantly provided by the central bank via two different intermediaries.

Having worked through the stack and some of its history, I’m going to make some guesses about what it might look like in 2038. I see three basic themes playing out: 1) a move to unstack parts of the stack, 2) a gradual improvement to the base banknote layer rather than its removal, and 3) a constant shrinking of the timing gap between the communication of a transaction and its settlement.

Unstacking the stack will become more common as non-banks that populate the third-layer skip bank intermediaries and directly connect to central banks. This is already underway in some parts of the world. In the U.S., the Office of the Comptroller of the Currency is championing a fintech banking charter. And a few months ago the Bank of England allowed Transferwise—a remittance provider—to open an account, the first ever non-bank to be granted the privilege of UK central bank access.

No longer dependent on banks as indirect providers of central bank services, fintechs could become even more potent competitors when it comes to payments. Come 2038, consumers may be just as likely to keep a payments account at the Apple or Starbucks as they are Wells Fargo.

An even more radical unstacking would provide the public with direct access to digital central bank balances. The public already has a physical means of accessing the bottom-most level of the stack: banknotes. But the relative prevalence of cash in retail payments has been steadily declining as consumers turn to electronic forms of payments. This has led some central bankers to consider the idea of introducing a central bank digital currency, or CBDC. A CBDC could be designed as a regular bank account or as a cash-like anonymous electronic token. Either way, a CBDC would provide the public with a direct digital connection to the base layer, potentially cutting banks and fintechs out of the stack.

Conversely, some economists believe that central banks should close their historical connection to the public by banning cash. Because the anonymity of cash facilitates crime and tax evasion, a cash ban would presumably make these illicit activities more difficult.

Both options—the banning of cash and the creation of a new form of cash—are quite radical. I think the most likely scenario is continuity. By 2038, paper money will still be the public’s sole connection to the base layer, although by then it will have become a niche monetary instrument that provides instant settlement and anonymity.

Incremental upgrades will ensure its continued relevance. Expect notes to slowly displace coins. The printing of notes on new substrates such as polymer increases their lifespan and decreases costs. This should allow central banks to put an end to the annoying trend of replacing small denomination notes with clunky coins. By 2038, the U.S. quarter could be a sleek banknote.

To help make cash become even more relevant, central bankers may start paying interest to banknote holders. Banknotes have always been burdened with a zero percent interest rate. But this needn’t be the case. Since all banknotes have a serial number, it is a simple matter to enact a weekly banknote lottery. Come 2038, the payouts that central bankers make on banknotes would make these old fashioned instruments at least as competitive as a high-yielding payments account at Amazon.

Finally, expect a continual shrinking of the gap between the communication of a digital payment message and the actual settlement of that payment. Take bill payments, for instance, which are frustratingly slow, sometimes taking several days to land in the biller’s account. When a bank receives a customer’s request to make a utility payment, that order is collected along with thousands of other orders into a bulk file. The file is transmitted at the end of the day to an Automated Clearinghouse, or ACH, which sorts and offsets these payments, providing each bank with a final amount owed or owing. Banks have typically only settled their debts a day or two later via transfers of central bank money on the base layer.

Payments gaps can be problematic. They increase the amount of planning that individuals and organizations must incur to ensure that they have sufficient funds in place to make ends meet. Anyone who needs to pay rent today but can only do so after an incoming ACH payment has been settled will understand the urgency of this problem.

In 2016 the U.S. introduced same-day ACH. This means that a payment order transmitted before 2:45PM can be settled that very day. As 2038 approaches, expect more of these sorts of incremental improvements to the ACH payments gap. Same-day ACH could become twice-a-day ACH, or hourly ACH. All of these gap reductions introduce more certainty into the payments stream.

Improvements needn’t be confined to the likes of ACH. Take SWIFT, the communications network responsible for processing cross border payments messages. Until last year a payment from Canada to Australia could take several days to settle. But SWIFT recently introduced new standards that require banks to ensure same-day settlement. By 2038, cross border payments could be up to same-minute settlement.

One potential spoiler to my 2038 scenario is bitcoin. Bitcoin exists entirely independently of the existing stack, putting it at odds with centuries of monetary innovation. Could we finally be witnessing money’s first Uber moment?

Local exchange trading systems, or LETS, may give us a hint. These are credit-based monetary systems formed by communities of like-minded individuals that began to pop in the 1980s and 90s. Rather than being built on top of existing monetary institutions like banks, LETs emerged as independent monetary islands, with their own currencies. The IOUs issued by LETs were secured by the collective social wealth of the community. They were denominated in novel units of account—like labor time. But LETS have never been able to attract more than a small core of regular users. They have simply been too awkward. Unless someone has a strong ideological motivation to sign up, most people will stick to the existing stack.

Bitcoin faces a similar problem. While a small group of devoted fans regularly transact in the stuff, it hasn’t become a mainstream payments medium. Lags and a complicated user interface may explain some of the mainstream’s reticence to buy stuff with bitcoin. But the biggest hurdle is that most people want stable money. Because bitcoin has no connection to the existing monetary stack, its price is irreparably volatile.

That doesn’t mean that bitcoin can’t be money’s Uber moment. But history shows that it’s a long shot. In the end, bitcoin could end up like LETS, a curious off-stack option that never attracts more than a dedicated group of regular transactors. In the meantime, the money stack will continue to be upgraded.