Even with the OCC’s conditional approval, Upstart should have an uphill battle to become a bank with a bank holding company.

1. Where the application stands

2. What the OCC required, and what it did not

The OCC imposed three special conditions, enforceable under 12 U.S.C. 1818, together with its standard pre-opening requirements.

Equally notable is what the decision does not contain:

• No CET1 floor and no tier 1 risk-based floor. Condition 2 names only the leverage and total risk-based ratios. While this is consistent with recent bank approvals, it may not be appropriate in this case.
• No limits on brokered or intermediated deposits, although these are described as the initial funding source.
• No condition specific to AI underwriting beyond the general requirement to adopt a model risk management policy, and no fair lending conditions.
• No published operating agreement, in contrast to some earlier fintech approvals (for example, the OCC’s 2022 approval of the SoFi Bank transaction, which required adherence to an operating agreement). LendingClub’s 2020–21 acquisition of Radius Bancorp, the closest marketplace-lender precedent, shows the reverse pattern. The OCC’s published conditions required $410 million of day-one bank capital, including at least $250 million from LendingClub, a three-year written Operating Agreement, and OCC no-objection to a new Chief Credit Officer and Chief Compliance Officer, but set no leverage or CET1 percentage. The ratio terms were instead set as commitments that LendingClub disclosed in its SEC filings: until the Operating Agreement expired in February 2024, LendingClub and LendingClub Bank each committed to maintain a CET1 ratio of 11.0%, a tier 1 risk-based ratio above 11.0%, a total risk-based ratio above 13.0%, and a tier 1 leverage ratio of 11.0%. The commitments ran to the Federal Reserve as well as the OCC, and they applied at the holding company as well as the bank.

The description of the decision letter is based on the text of Corporate Decision #1382. Confidential supervisory terms, if any, would not appear in it. The LendingClub precedent shows that the binding constraints can sit outside the published conditions: its capital ratio requirements did not appear in the OCC’s decision letter and were disclosed only later, as commitments in LendingClub’s SEC filings.

3. Why the OCC’s conditions are less demanding than they appear

3.1 The ratios, while in line with precedent, should be higher

Measured against capital conditions in recent de novo and charter approvals, a 12 percent leverage floor is unremarkable for a monoline, sponsor-dependent model. The FDIC has gone materially higher where a bank’s business is concentrated and dependent on its parent.

The OCC’s 15 percent total risk-based floor is the more distinctive element; it is the only explicit risk-based floor found among the recent fintech approvals reviewed. But a risk-based floor is only as strong as the risk weights beneath it, and for the bank as proposed the assets beneath it are mostly unsecured consumer loans with nonprime credit scores, which the standardized approach risk-weight at a flat 100 percent regardless of borrower credit quality. The floor does not address all the exposures that make the Upstart enterprise risky.

The flat weight matters because Upstart’s lending is concentrated among borrowers the agencies would classify as less than prime. Upstart’s platform offers unsecured personal loans at APRs of up to 35.99 percent, along with auto loans and HELOCs, and Upstart states that its model approves substantially more borrowers than a traditional model based only on credit scores, meaning many of its borrowers would be declined under conventional underwriting. An academic study of roughly 900,000 Upstart loans originated from 2014 through early 2021 found an average credit score at origination of 653 and an average APR of 22 percent. Upstart’s own access-to-credit materials feature borrowers with FICO scores below 660, and in valuing its Level 3 loans at December 31, 2024 it used a weighted-average credit risk rate of about 16 percent. Some third-party marketplaces report a higher average score for their channels, so the mix varies by product and channel. On the public evidence, however, a large share, and likely most, of the personal loans the bank would purchase would fall at or below the 660 FICO line the agencies use to identify subprime borrowers.

This has a capital consequence the OCC’s condition does not address. The interagency Expanded Guidance for Subprime Lending Programs (2001), still in effect (apart from the recent removal of references to reputation risk), applies to institutions whose subprime lending programs equal or exceed 25 percent of tier 1 capital. Upstart Bank would likely cross that threshold soon after opening if it purchases Upstart-originated loans as its business plan contemplates. The guidance identifies subprime borrowers by characteristics that include a credit bureau score of 660 or below, depending on product and collateral. It states that the minimum capital requirements generally apply to portfolios with substantially lower risk profiles than subprime programs. It also directs examiners to expect, as a starting point, that an institution will hold capital against subprime portfolios of one and one half to three times the amount appropriate for non-subprime assets of a similar type. Consumer loans carry a 100 percent risk weight, so a well-capitalized bank, which must hold total capital of at least 10 percent of risk-weighted assets, needs about 10 cents of capital per dollar of ordinary consumer loans. Applying the guidance’s multiplier to that baseline yields 15 to 30 cents per dollar of subprime loans, or capital of 15 to 30 percent of the subprime portfolio. The OCC’s 15 percent floor therefore sits at the bottom of the range the guidance treats as a starting point, and only if the higher capital normally expected of a de novo bank is ignored.

3.2 How can Upstart meet the condition?

The OCC regulates the national bank, and Condition 2 is measured on the bank alone. Upstart’s most leveraged credit exposure, its committed-capital and co-investment arrangements, sits outside any bank today, at Upstart Holdings and its nonbank subsidiaries and vehicles. At June 30, 2026, Upstart reported maximum exposure to loss under those arrangements of $1,294 million, comprising $285 million of cash, $315 million of restricted cash, $534 million of beneficial interests, $108 million of line-of-credit receivable and $52 million of loans. Maximum exposure was $932 million at December 31, 2025 and $1,078 million at March 31, 2026. Some of the arrangements are derivatives, under which Upstart periodically pays or receives cash as credit performance deviates from initial expectations; others are debt-like, an upfront investment repaid through a share of collections that varies with performance against expectations. Upstart has not said whether these positions, or the new ones its committed-capital agreements will generate, would be held in the bank or outside it. Asked on its second-quarter 2026 earnings call how much capital it would put into the bank, its chief financial officer said the company had not disclosed the specifics of its capital plan but was “sufficiently capitalized at Upstart today to launch the bank.” That is a statement that the parent can fund the bank’s initial capital, not that the bank or the group would meet the OCC’s ratios.

On the same call, Upstart said it expects to move “the bulk to all” of its originations to Upstart Bank soon after launch. If the bank becomes the originator and the loans continue to be sold under committed-capital arrangements, the first-loss positions those sales create must be retained by the bank or by an affiliate. That leaves Upstart three ways to meet the 15 percent condition, and each has a cost.

Hold the exposures in the bank and capitalize them. At 15 percent, a first-loss position risk-weighted at 1,250 percent requires about $1.88 of capital per dollar of exposure. The $404.1 million of initial capital would therefore support only about $215 million of such exposure, before any capital is allocated to the bank’s loans. Holding the $546 million of beneficial interests reported June 30, 2026¹, would require about $1.0 billion of capital on its own, and holding the full $1,294 million of maximum exposure would require about $2.4 billion, against consolidated book equity of $797 million. On this path, the stated initial capital is far too low, and Upstart would need to raise significant new equity.

Keep the exposures outside the bank. This is the arrangement under which $404.1 million could plausibly satisfy Condition 2, and it is consistent with a bank whose stated initial business is buying Upstart loans funded by brokered and intermediated deposits. But it does not remove the exposures from regulatory capital; it moves the question to the Federal Reserve. If Upstart Holdings becomes a bank holding company, Regulation Q applies on a consolidated basis, and Section 4 estimates a consolidated total risk-based ratio of roughly 8.0 percent (Scenario A) to 3.9 percent (Scenario B), below the 10 percent well-capitalized level and, in Scenario B, below the 8 percent minimum. The Board must also weigh the company’s financial resources and its ability to serve as a source of strength to the bank. Concentrating the first-loss book at the parent is precisely what makes those findings harder to reach.

Restructure the arrangements. In theory, Upstart could move toward whole-loan sales without retained first loss, or cap and run down its committed-capital exposure. While Section 4 shows this is more capital-efficient than raising equity against the exposure, sporadic misses from the performance of UPST’s collateral is why these arrangements have become increasingly necessary. Management describes co-investment as the price of the longer, 24-month funding commitments it values most, so reducing it would trade capital relief for less stable funding, the risk Section 5 identifies as correlated with credit stress.

Upstart has not said which path it will take. On the public numbers, the bank can meet the OCC’s 15 percent condition with $404.1 million only if the first-loss book stays outside it, and that outcome relocates the capital problem to the holding company application rather than solving it. The realistic choices are to raise substantial new equity, to shrink the exposures, or to ask the Federal Reserve to approve a holding company that is not well capitalized on a consolidated basis, and that’s a nonstarter.

3.3 Fair-value accounting removes a buffer and accelerates losses

Upstart carries its loans, beneficial interests, retained securitization notes and residuals, and its line-of-credit receivable at fair value, predominantly Level 3. That has two capital consequences. First, there is no allowance for credit losses, so there is no allowance-based tier 2 capital. Second, changes in expected losses and in discount rates flow through earnings, and therefore through CET1, immediately. At June 30, 2026, Level 3 fair-value assets totaled roughly $1.9 billion, about 2.4 times book equity of $797 million.

A clarification is warranted. Upstart’s equity is invested in assets whose value depends almost entirely on management assumptions. Although the capital rule deducts after-tax gain-on-sale resulting from a securitization and assigns a 1,250 percent risk weight to the portion of any credit-enhancing interest-only strip not already deducted, it does not otherwise discount equity for the estimation risk in Level 3 assets. That estimation risk should have been addressed by a significantly higher risk-based ratio requirement than the OCC has required.

3.4 Funding is intermediated and unconstrained

The bank’s initial funding through deposit brokers, placement networks and other intermediaries is precisely the funding profile the FDIC scrutinizes in assessing risk to the Deposit Insurance Fund. The OCC decision describes it but places no limit on it. That leaves the question to the FDIC.

4. Consolidated risk weighting

If the Federal Reserve approves Upstart Holdings as a bank holding company, Regulation Q (12 CFR Part 217) applies on a consolidated basis. The Small Bank Holding Company Policy Statement (12 CFR Part 225, Appendix C) is unavailable: Upstart’s total assets of $3.2 billion exceed the $3 billion threshold, and the policy statement also excludes companies with a material amount of SEC-registered debt or equity outstanding.

The following mapping applies the standardized approach to the June 30, 2026 balance sheet. The classifications are the author’s judgment based on public disclosures; they are not Upstart’s.

Before securitization exposures, risk-weighted assets are about $1.48 billion. The committed-capital arrangements meet the definition of securitization exposures: Upstart absorbs credit losses on investor-funded pools up to a contractual cap, which tranches the credit risk. For a thin first-loss position on a pool of 100 percent risk-weighted consumer loans, SSFA produces 1,250 percent whenever the detachment point is at or below the pool’s capital charge, and 1,250 percent applies regardless if the due-diligence and data requirements for SSFA are not met.

Two observations follow. First, leverage is not the constraint; risk-based capital is. Under Scenario B the consolidated company would fall below the 8 percent Regulation Q total capital minimum. Second, a 1,250 percent risk weight equates to dollar-for-dollar capital only at an 8 percent requirement. At the OCC’s 15 percent level, each dollar of first-loss exposure would require about $1.88 of capital, which is why these positions cannot efficiently sit inside Upstart Bank and why their treatment at the holding company is central to the Federal Reserve’s review.

Equity required

The table shows the estimated new common equity needed to reach each total risk-based threshold, assuming proceeds are held as cash at a 20 percent risk weight (which slightly increases RWA). In principle, part of each shortfall above the tier 1 requirement could be met with tier 2 subordinated debt rather than common equity, but regulators are unlikely to accept a de novo bank capital plan that relies on holding-company debt, so the table treats the full amount as common equity.

Reducing the exposure is more capital-efficient than raising equity against it. At a 15 percent target, each dollar of first-loss committed-capital exposure unwound reduces required capital by about $1.88 and also releases the collateral posted against it. But this tactic would be a complete about-face for how the company has needed to operate in order to sell its loans at a price that justifies its business model.

5. The unexpected-loss problem

The disclosed loan sensitivities are modest. For the $1,028 million of held-for-sale and held-for-investment loans (excluding the consolidated securitization), the 10-Q reports the following effects on fair value:

These figures understate enterprise risk for three reasons.

1. The leverage is in the beneficial interests, and they are not sensitivity-tested in the filing. About $519 million of hybrid instruments references $6.32 billion of principal and about $27 million of derivatives references $2.93 billion. Each additional percentage point of cumulative loss on roughly $9.2 billion of referenced principal implies about $90 million of loss to Upstart, up to its contractual maximum. The absence of a sensitivity table for these positions is itself a disclosure gap a supervisor should pursue.
2. Credit and funding risk are correlated. When credit deteriorates, third-party funding contracts, and Upstart has historically ended up holding more loans on its own balance sheet, as in 2022. The balance sheet tends to grow when marks are falling, significantly challenging the company’s ability to scale.
3. Spread risk hits capital before defaults do. Under fair-value accounting, wider market spreads reduce asset values immediately, with no allowance to absorb them. For the six months ended June 30, 2026, Upstart recorded $76.8 million of loan charge-offs and fair-value losses and reported net income of $9.9 million: thin earnings with which to absorb a shock.

The loss record does not argue for a lighter touch. Upstart does not publish cumulative net loss as a percentage of original principal; it reports performance only relative to its own expectations, and the figures are self-reported. At its AI Day on May 14, 2025, Upstart showed cumulative losses running above expectation for every origination half from 2022 through 2023: roughly 59% above plan for H1 2022, 48% for H2 2022, 34% for H1 2023 and 17% for H2 2023, narrowing to about 5% for each half of 2024. These figures are read from chart labels and are approximate. The 2023 vintages, originated after Upstart tightened its underwriting, also ran above plan, and the Q1 2026 10-Q states that the Q1 2023 through Q1 2024 vintages are still forecast to underperform target returns, with only Q2 2024 onward in line. Rating surveillance points the same way. In its February 2026 and September 2026 reviews of Upstart securitizations, KBRA downgraded one class each time, citing an expected majority loss of principal within 12 months. KBRA did not name the class in its public releases; UPST 2022-1 Class C, rated CCC (sf) since November 2023 and included in the September 2026 review, is the likely candidate, but this is unconfirmed. Deal-level cumulative net loss for UPST 2023-2 through 2026-4 is not publicly available.

Concentrated exposure, a capped but highly leveraged first-loss position, correlated funding risk and immediate mark-to-market are the characteristics unexpected-loss capital is meant to cover. The OCC’s bank-level conditions are not sufficient to meet these concerns.

6. Why the FDIC and the Federal Reserve are unlikely to be as lenient

6.1 The FDIC

• Statutory mandate. Section 6 of the Federal Deposit Insurance Act requires the FDIC to consider, among other factors, the adequacy of the institution’s capital structure, its future earnings prospects and the risk it presents to the Deposit Insurance Fund. The OCC does not apply the last factor; the FDIC alone decides whether to insure.
• Precedent for higher leverage floors. Square (20 percent) and Ford and GM (15 percent) show the FDIC setting leverage floors well above the OCC’s 12 percent where a bank’s business is concentrated and parent-dependent. Upstart’s model is both.
• Parent support agreements. The FDIC required capital and liquidity maintenance agreements and parent company agreements in Square, Nelnet, Ford and GM. A comparable commitment from Upstart Holdings would raise the question the OCC’s conditions avoid: whether a parent whose capital is invested in Level 3 first-loss positions can support the bank in the same downturn that impairs those positions.
• Funding. An order could limit brokered and intermediated deposits or require a liquidity plan that the OCC decision does not.

6.2 The Federal Reserve

• Consolidated capital applies. As shown in Section 4, Regulation Q applies to Upstart Holdings on a consolidated basis, and on reasonable assumptions the company would be below well capitalized, and possibly below minimum, without equity or restructuring.
• Statutory factors. Section 3(c) of the Bank Holding Company Act requires the Board to consider the financial and managerial resources and future prospects of the company and the bank.
• Source of strength. Section 38A of the FDI Act (12 U.S.C. 1831o-1) requires a bank holding company to serve as a source of financial strength to its depository institution subsidiary. That obligation is hard to credit to a parent with the risk profile described above unless the committed-capital book is restructured, capped or capitalized.
• Commitments. The Board can condition approval on commitments regarding capital levels and permissible activities, which is the natural vehicle for limits on first-loss arrangements.

6.3 Common authority

Each agency has a reservation of authority to require capital beyond the regulatory minimums when those minimums are not commensurate with an institution’s risk (12 CFR 3.1(d) for the OCC, 217.1(d) for the Board and 324.1(d) for the FDIC). Examiners may also challenge Level 3 valuation inputs; a required markdown flows through GAAP and reduces CET1 directly. The agencies do not need to disallow equity to address its quality. They can require more of it, constrain the activities that put it at risk, and secure parent commitments.

7. The case for the OCC’s approach, and why it is not persuasive

A fair assessment should engage the strongest counterarguments.

• The 15 percent total risk-based floor is well above the 10 percent well-capitalized threshold, and the agencies have not historically imposed CET1 conditions. True, but a floor that is high relative to the bank’s own risk weights is not high relative to the enterprise’s risk, and the bank’s viability depends on the enterprise.
• The capital rule already assigns 1,250 percent to first-loss securitization exposures. Also true, and that is why the consolidated ratios look as they do. The point is that the OCC’s bank-level condition does not engage with those exposures at all; the other agencies must.
• Upstart intends to keep selling about 95 percent of its originations, so the bank will carry little credit risk. Selling loans to investors is precisely what generates the committed-capital exposure: the risk does not leave the group, it moves to the holding company in leveraged form or stays at the bank where the FDIC should require significant capital to offset its risk of insuring this potential bank’s insured, brokered deposits.
• The current agencies are approving charters quickly. The FDIC approved Ford and GM in January 2026, and the OCC has accelerated fintech reviews. But those FDIC approvals came with 15 percent leverage floors and parent support agreements, which supports this paper’s expectation rather than undercutting it.

8. What to watch

• Upstart’s Q3 2026 Form 10-Q: updated committed-capital exposure, beneficial-interest balances and any disclosure about regulatory capital planning.
• The FDIC order: the leverage floor, any risk-based or CET1 floor, any capital and liquidity maintenance agreement, and any limit on brokered deposits.
• The Federal Reserve order: commitments on consolidated capital and on committed-capital and co-investment arrangements.
• Capital raising: whether the $404.1 million is funded from existing parent resources or new equity, and whether any tier 2 instrument appears in the plan.
• Timing: the OCC’s 12-month capital and 18-month opening deadlines (about July 2027 and January 2028).

Methodology and limitations

• Balance sheet figures are from Upstart’s Form 10-Q/A for the quarter ended June 30, 2026. The Q3 2026 filing will change them.
• Risk-weight assignments are the author’s; Upstart does not disclose regulatory capital calculations. Upstart does not disclose counterparty names, attachment and detachment points, seniority, tenor or triggers for its committed-capital structures. Without those terms SSFA cannot be calculated, so the choice between 1,250 percent and SSFA is a judgment; any position above first loss would receive a lower SSFA risk weight. The company, or lenders to the counterparties, could supply the missing terms.
• For derivative-form arrangements, the exposure amount may be measured on a notional rather than fair-value basis.
• Other intangibles and deferred tax asset deductions are not disclosed and were not estimated; including them would increase the equity needed.
• The precedent review covers prominent fintech, industrial bank and community-bank approvals, not every de novo order since 2016, and does not assess non-public operating agreements or supervisory terms.

Sources

• OCC Corporate Decision #1382, Upstart Bank, N.A. (July 23, 2026). https://www.occ.gov/topics/charters-and-licensing/interpretations-and-decisions/2026/cd1382.pdf
• Upstart Holdings, Form 8-K (March 10, 2026). https://www.sec.gov/Archives/edgar/data/1647639/000164763926000036/upst-20260310.htm
• Upstart Holdings, Form 10-Q/A for the quarter ended June 30, 2026. https://www.sec.gov/Archives/edgar/data/1647639/000164763926000066/upst-20260630.htm
• Upstart press release: conditional OCC approval (July 23, 2026). https://ir.upstart.com/news-releases/news-release-details/upstart-receives-conditional-approval-occ-establish-upstart-bank
• FDIC Order: Square Financial Services (2020). https://www.fdic.gov/sites/default/files/2024-03/pr20033a.pdf
• FDIC Order: Nelnet Bank (2020). https://www.fdic.gov/news/press-releases/2020/pr20034a.pdf
• FDIC Order: Varo Bank, N.A. (2020). https://www.fdic.gov/resources/supervision-and-examinations/bank-applications/decisions/varo-bank-na-draper-ut.pdf
• FDIC press release: Ford Credit Bank and GM Financial Bank (January 22, 2026). https://www.fdic.gov/news/press-releases/2026/fdic-approves-deposit-insurance-applications-ford-credit-bank-salt-lake
• FDIC Order: GM Financial Bank (2026). https://www.fdic.gov/gm-order-and-statement.pdf
• OCC conditional approval: Erebor Bank (2025). https://occ.gov/news-issuances/news-releases/2025/nr-occ-2025-101a.pdf
• FDIC Order: Houston Bank and Trust (2025). https://www.fdic.gov/bank-examinations/houston-bank-and-trust-bank
• Federal Reserve SR 20-16, Supervision of De Novo State Member Banks. https://www.federalreserve.gov/supervisionreg/srletters/SR2016.htm
• SoFi Technologies, 2022 Annual Report (OCC operating agreement). https://www.sec.gov/Archives/edgar/data/1818874/000181887423000073/sofitechnologies2022ars.pdf
• OCC Conditional Approval #1258, Radius Bank conversion and merger into LendingClub Bank, N.A. (December 30, 2020). https://www.occ.treas.gov/topics/charters-and-licensing/interpretations-and-decisions/2021/ca1258.pdf
• LendingClub Corporation, Form 8-K (January 19, 2021), reporting Federal Reserve Bank of San Francisco approval on January 15, 2021. https://www.sec.gov/Archives/edgar/data/1409970/000140997021000003/lc-20210115.htm
• LendingClub Corporation, Form 10-K for 2023 (capital ratio commitments under the OCC Operating Agreement). https://www.sec.gov/Archives/edgar/data/1409970/000140997024000009/lc-20231231.htm
• American Banker, “LendingClub exits OCC agreement, sees path for faster growth” (February 2024). https://www.americanbanker.com/news/lendingclub-exits-occ-agreement-sees-path-for-faster-growth
• 12 CFR Part 3, OCC capital adequacy standards (including §§ 3.2, 3.22 and 3.42). https://www.ecfr.gov/current/title-12/chapter-I/part-3
• 12 CFR Part 217, Regulation Q (including § 217.42). https://www.ecfr.gov/current/title-12/chapter-II/subchapter-A/part-217

1. The $546 million is the fair value of beneficial interest assets on the balance sheet. The $534 million cited above is the beneficial-interest component of Upstart’s maximum exposure to loss under its committed-capital and co-investment arrangements (Form 10-Q, Note 4). The two figures are measured on different bases. ↩︎