Upstart Holdings, the AI-driven consumer lender, has received conditional approval from the Office of the Comptroller of the Currency to form its own national bank, Upstart Bank, N.A., and expects to open it in early 2027, pending federal deposit insurance and Federal Reserve sign-off on the holding company. When it does, the new bank will become what Upstart calls the "primary originator" of its loans, moving most or all of that work in-house and displacing the banks that serve as the lender of record today. Upstart paid those originating banks about $11.2 million in premium and trailing fees in the first half of this year, fees it would not owe on loans its own bank makes. The institutions that buy Upstart's loans, meanwhile, keep their place in the chain.

American Banker framed the story as a question of who loses, and the answer is more precise, and more revealing, than it first appears. The partners Upstart is dropping and the partners it is keeping are divided along a clean line, and that line exposes what the fintech-bank partnership model always was underneath: not a durable symbiosis, but a temporary arrangement that was always going to end this way once the fintech grew up.

The arrangement Upstart is leaving

For most of its life Upstart has operated as a marketplace rather than a lender. It couldn't originate loans directly across the country, because it lacked a bank charter and the ability that comes with it to lend nationally and export interest rates without navigating a patchwork of state licenses. So it did what a generation of fintechs did: it partnered with chartered banks that acted as the lender of record, originating loans on paper using their charters, while Upstart supplied everything that actually distinguished the product, the AI underwriting models, the customer acquisition, the servicing, and captured the economics. The partner bank collected a fee for the use of its charter.

That arrangement let Upstart behave like a national lender without being a bank, and it scaled: the company worked with more than a hundred partner banks and credit unions. In 2025, the loans Upstart's platform produced ended up split roughly 64% with institutional investors, 26% with lending partners, and 10% on Upstart's own balance sheet. The originating banks were one layer of this machine; the buyers of the loans were another.

Why the partnership was always a bridge

Here is the uncomfortable truth about the lender-of-record layer: the partner bank contributed very little beyond its charter. It did not build the underwriting models, which are Upstart's. It did not acquire the borrowers, which Upstart does. It did not, in most cases, hold much of the risk, since the loans were sold onward. What it provided was a regulatory privilege, the legal capacity to originate, rented out for a fee. And renting out a regulatory privilege is a commoditized, low-value service, one a fintech pays for only as long as it cannot obtain that privilege for itself.

Which means the relationship was structurally temporary from the start. The fee existed only because Upstart lacked a charter, so the moment Upstart could get one, the partner's role, and its fee, was destined to evaporate, because the fintech had only ever been paying for access to a charter it can now own outright. The partnership was never a marriage; it was a lease, useful to the tenant right up until the tenant could buy the building. Upstart's charter is the tenant buying the building, and the lease ending.

Who loses maps onto what each partner actually provided

This is why the division between the losers and the survivors is so clean. The originating partners lose because their entire contribution was the rentable charter, the part that is inherently disintermediable. The loan-buying partners keep their role because buying loans is a genuine, value-adding function: it means supplying capital, bearing credit risk, and earning a return, and Upstart's own small bank does not replace it. Upstart will still need buyers for the vast majority of the loans it makes, since it plans to keep only a fraction on its own balance sheet and to keep selling to capital partners, as a Stephens analyst noted in observing that the new bank complements the loan-buying business even as it displaces the origination one.

So the split is not arbitrary. It maps exactly onto whether a partner was providing a commoditized regulatory service or real economic value. Rent out a privilege, and you are disintermediated the instant your customer can obtain that privilege directly. Provide capital, and you endure, because capital is a genuine contribution rather than a rentable formality. The question of who loses when a fintech becomes its own bank turns entirely on which of those two things a partner was actually selling.

This is vertical integration, not simple cost-cutting

It would be a mistake to read the vanishing $11.2 million in fees as pure savings, because Upstart is not eliminating a cost so much as internalizing a function. Bringing origination in-house means Upstart now shoulders what the partner banks used to carry: the capital requirements of a chartered bank, the ongoing supervision of the OCC and the Fed, the work and cost of gathering deposits, and a heavier compliance burden. The company is trading a fee paid to partners for the obligations of being a bank itself.

That is a classic build-versus-buy decision playing out over the arc of a company's growth. At small scale, renting a charter is cheaper and simpler than becoming a bank, so fintechs buy the service. At large scale, the fees and the operational friction of coordinating a hundred partners start to outweigh the burdens of owning a charter, and building becomes the better deal, especially once a charter unlocks direct access to consumer deposits as a cheaper and more stable source of funding than repeatedly selling loans wholesale. Upstart has simply crossed the scale threshold where owning beats renting.

The trade is real, and not obviously free

None of this makes the move a guaranteed win, and the tradeoffs deserve to be stated plainly. Becoming a bank subjects Upstart to capital requirements and to prudential supervision that can constrain how fast it grows and how it underwrites. It has to compete for deposits in a crowded market and absorb the operational complexity of running a regulated institution. And there is a risk specific to Upstart's history: its AI underwriting has proved sensitive to the macroeconomic cycle, struggling notably when interest rates rose, and funding more loans with its own deposits on its own balance sheet concentrates on Upstart the credit and funding risk it previously distributed to partners and buyers. The company's stated intention to keep selling most loans to investors softens this, making the new bank a hybrid rather than a full balance-sheet lender, but the direction is toward Upstart holding more of its own risk. In exchange for the funding advantages and control that a charter brings, it is giving up some of the asset-light flexibility that defined the platform model, and whether that is a good trade will come down to execution.

The bellwether, and the regulatory irony beneath it

The most consequential part of this story is what it signals for everyone else. If Upstart's most successful peers follow the same path, and the wave of recent charter approvals for fintechs like Nubank, Mercury, and others suggests many will, then the small banks that built businesses around renting their charters to fintechs face an existential problem. Their fee income depends on fintech partners who cannot or will not become banks themselves, and their best partners are precisely the ones now graduating to their own charters. The rent-a-charter model works only so long as charters are hard for fintechs to get.

And here is the irony. The fintech-friendly regulatory posture that is being celebrated across the industry, with the OCC approving charters at a faster clip than it has in years, is the very thing dismantling the partner-bank business. By making charters easier for fintechs to obtain, regulators are removing the scarcity that gave partner banks their reason to exist. Those banks' role depended on the charter being a scarce privilege worth renting; make the privilege abundant, and the rental business collapses. The regulatory generosity that helps the fintechs is, at the same time, a slow undoing of the banks that served those fintechs through the years when charters were out of reach.

Seen whole, Upstart becoming its own bank looks at first like a simple story of a company maturing from a platform into a lender. Its deeper lesson is about the arrangement it is leaving behind. The fintech-bank partnership was a lease on a privilege, not a partnership of equals, and leases end when the tenant can buy. The partners whose only contribution was a signature on the loan were always going to be written out the moment the fintech could sign for itself, while the partners who put up capital remain, because capital was never a rentable formality. Upstart will not be the last tenant to buy the building, which is good news for fintechs and, through lower costs, potentially for their borrowers, and a quiet reckoning for the banks whose business was collecting rent on a privilege that is no longer scarce.

Primary sources

  1. American Banker and National Mortgage News (reporting by Carter Pape) for Upstart's plan to move most or all originations into a national bank it aims to open in early 2027, the displacement of the banks that serve as lender of record, the $11.2 million in premium and trailing fees paid to originating banks in the first half of the year that Upstart would not owe on its own bank's loans, the continued role of loan buyers, and the 2025 funding split of 64% institutional investors, 26% lending partners, and 10% Upstart's own balance sheet.
  2. American Banker and National Mortgage News for Stephens analyst Kyle Joseph's observation that the new bank complements Upstart's existing loan-selling business while cutting both ways for partners.
  3. Finovate, Banking Dive, and PYMNTS for the OCC's conditional approval of the de novo charter on July 23, 2026, the Delaware headquarters and branchless digital model, the pending FDIC deposit-insurance and Federal Reserve holding-company applications, proposed bank CEO Annie Delgado's remarks, and the broader wave of fintech charter approvals under the current OCC, including Nubank, Mercury, Valt, and Erebor.
  4. Yahoo Finance and Simply Wall St for the trade-off between funding and cost efficiencies and the tighter regulatory scrutiny and capital requirements of becoming a bank holding company.
  5. Yahoo Finance and Upstart's own statements for the goal of lowering the cost of capital through direct deposit access, the reliance on more than 100 partner banks and credit unions, and CTO and incoming CEO Paul Gu's comments.
  6. The Motley Fool for the characterization of the shift from a platform middleman to a lender and the role of deposits as a new funding source.