Why private credit exists, explained

Written 3 September 2026 · All explainers

How to read the markers. [D] — verified from a source retrieved for this note, listed under Sources. [I] — industry-understood; not formally disclosed by a named party. [S] — our own inference or judgment. Every figure is either sourced or marked as an estimate.

The short version

A bank lends out money that mostly belongs to its depositors. Because depositors can ask for it back at any time, the regulator makes the bank hold a slice of its own money — equity, called capital — against every loan, sized by how risky the loan is. The cost of holding a loan on a bank’s balance sheet is set by a rule, not only by the borrower, and when the rule changed after 2008 the cost of some loans rose enough that they moved to a different kind of lender.

That lender is a private credit fund. It has no depositors, so nobody can run on it, and no regulator sets a risk-weighted capital ratio for it. Its money is locked up by its investors for years. It is also almost all equity, which is expensive, so it only pays to hold loans with a wide spread: the unrated, mid-sized, buyout-financed loans the new bank rules made dearest for a bank to keep.

The loan changed shape on the way across: floating rate, a handful of lenders, held to maturity, a private rating or none, often one “unitranche” instead of a stack. And the circle has closed: banks now lend to the funds and partner with them. The loan left the bank; the bank’s exposure to it did not entirely leave, and how much remains is, by the regulators’ own account, only partly measured.

---

1. What a bank is, and why it has a capital ratio

A bank takes deposits and makes loans. The loans are its assets; the deposits are most of its liabilities; the gap is its own money, its capital. A bad loan is charged to capital first, so capital is the buffer between the loan book and the depositors.

Deposits can leave in a day and loans cannot, so a regulator sets a minimum thickness for the buffer, per unit of risk. Each asset gets a risk weight, a percentage that scales the loan by how likely it is to lose money. Risk-weighted assets (RWA) are loans times weights; the capital ratio is capital divided by RWA.

Under the Basel Framework, the international standard national rules copy, the minimums since 1 January 2023 are Common Equity Tier 1 (CET1 — ordinary shares and retained profits, the purest capital) at least 4.5% of RWA; Tier 1 at least 6%; total capital at least 8% (RBC20.1). On top sits a capital conservation buffer of 2.5% of RWA in CET1 (RBC30.2) and a countercyclical buffer set nationally between 0% and 2.5% (RBC30.9); the largest banks carry a further systemic surcharge. Under the standardised approach a loan to a company with no external rating carries a 100% risk weight; a loan to an AAA-rated company carries 20% (CRE20). [D — BIS, see Sources]

For a £100 loan to an unrated mid-sized company:

Source: derived from BIS RBC20.1, RBC30.2, RBC30.9 and CRE20, retrieved 3 September 2026. [D for the ratios; the arithmetic is ours]

So what: the bank must fund £7 to £13 of every £100 of this loan with shareholders’ money, and shareholders demand far more than depositors. That £7–£13 is the loan’s capital charge, and the rest of this piece turns on it. Raise the ratio or the risk weight and the loan gets dearer to hold, with nothing changed at the borrower.

---

2. What changed: three rules, three kinds of loan

Basel III, from 2010. After 2008 the Basel Committee raised the quantity and quality of capital, put common equity at the centre, and added the leverage ratio and the two buffers above. The standards were endorsed at the G20 Seoul summit in November 2010 and issued that December. [D — BCBS history] This touched every loan on every bank’s book: the same risk weight now needed more, and better, capital behind it. [S]

The 2013 US leveraged-lending guidance. On 21 March 2013 the Federal Reserve, FDIC and OCC issued the Interagency Guidance on Leveraged Lending. Its examples of leveraged lending include loans that fund buyouts, or where total debt exceeds 4.0× EBITDA (earnings before interest, tax, depreciation and amortisation — a proxy for operating cash flow). It says leverage above 6× total debt to EBITDA “raises concerns for most industries”, and that supervisors expect a borrower to be able to repay 50% of total debt over five to seven years. [D] This targeted one kind of loan: the buyout loan to a heavily indebted company. A New York Fed staff study found the guidance cut leveraged lending by large, closely supervised banks and “triggered a migration of leveraged lending to nonbanks”, which then borrowed more from banks. [D — Kim, Plosser and Santos, 2017]

The “Basel III endgame” and its US reversal. The Basel Committee finished its reforms in December 2017 with an output floor limiting how far internal models can lower a bank’s capital. [D — BCBS history] The US agencies proposed to implement them on 27 July 2023 for banks with $100 billion or more in assets, estimating a 16% aggregate increase in CET1 requirements. [D — Federal Reserve] That proposal did not survive. On 19 March 2026 the same agencies issued three replacement proposals under which overall capital “would modestly decrease” — a law-firm reading of the impact tables puts aggregate CET1 requirements about 4.8% lower for the largest banks — and the corporate risk weight falls from 100% to 95%. Comments closed 18 June 2026; the Federal Reserve’s June 2026 supervision report still lists them as proposals. [D — Federal Reserve press release, fact sheet, June 2026 report; Sullivan & Cromwell for the percentage] The Vice Chair for Supervision gave the reason plainly: the earlier approach had “pushed activity into the less-regulated non-bank sector”, and the new one aims to “reduce incentives for traditional lending activities … to migrate outside of the regulated banking sector”. [D — Bowman statement, 19 March 2026] The UK waited: on 17 January 2025 the PRA delayed Basel 3.1 to 1 January 2027, citing US uncertainty. [D]

So what: for mid-market buyout loans the rule that mattered was the 2013 guidance, not Basel III as such. Basel III raised the price of every loan a little; the guidance made one category a supervisory problem, and that category is where private credit grew. The 2026 US proposals are the first reversal since 2008, and the regulator names private credit as the reason.

---

3. The fund’s balance sheet against the bank’s

A private credit fund is a limited partnership. Its investors (limited partners, LPs) commit money the manager calls as loans are made, and it “is typically locked up until loans are repaid, often for five to seven years”. Most of its loans never trade. [D — New York Fed] No depositor, so no run, so no risk-weighted capital ratio. The fund is, in effect, all capital.

That is not the same as no leverage. The fund borrows in three main ways:

• Subscription lines — one- to three-year bank loans secured on the LPs’ uncalled commitments, around 200 basis points over benchmark, used to lend today and call LP money later. The FSB calls them “the most common form of bank financing to private credit funds”. [D — FSB, May 2026]
• NAV facilities — three- to five-year loans secured on the fund’s portfolio, around 200–400 basis points over benchmark. [D — FSB]
• The BDC rule. A business development company (BDC) is a US fund vehicle. Since the Small Business Credit Availability Act, enacted 23 March 2018, Section 61(a) of the Investment Company Act 1940 lets a BDC lower its required asset coverage from 200% to 150% with board or shareholder approval: per $100 of equity it may borrow $200 rather than $100. [D — SEC staff guidance; Proskauer] Bank loans are now about 40% of BDC debt, up from 20% a decade earlier. [D — FEDS Notes, August 2026]

Why is the fund the cheaper holder of some loans and not others? Because the two lenders have opposite cost structures. [S — our framing, consistent with the BIS and Kansas City Fed findings below] The bank’s funding is mostly cheap deposits, so it wins wherever the capital charge is small: rated, liquid, low-spread lending. Its problem is the tail — the unrated, 6×-levered buyout loan where the risk weight is 100%, supervisors ask questions, and the capital charge eats the spread. The fund’s funding is mostly dear equity, so it loses on low-spread lending; but it pays no extra charge for risk. A wide-spread loan is, to a fund, simply a high-yielding asset; to a bank it is a high-capital one.

The BIS found that stricter post-crisis bank regulation “has spurred lending by private credit funds”, and that the cost-of-capital gap between BDCs and banks narrowed by about 200 basis points between 2010 and 2019, mostly because BDC equity got cheaper and BDCs borrowed more. [D — BIS Quarterly Review, March 2025] The Kansas City Fed adds that private credit funds lend “into markets with much higher credit risk, as indicated by higher loan spreads and higher loss given default”. [D] So what: the loan moved to where the cost of risk was lowest, and that was not the lender with the deposit franchise.

---

4. What the new holder does with the loan

Five differences from the loan a bank syndicated in 2007, each following from section 3.

1. Floating rate. A benchmark (SOFR in dollars, Euribor in euros) plus a fixed spread; the LPs are promised a spread, and the fund’s own borrowing floats too. [I — industry standard; the BIS and New York Fed both describe private credit as floating-rate]
2. Direct negotiation, few lenders. The New York Fed defines private credit as “a loan that is negotiated directly between a borrower and a small group of nonbank lenders”. [D]
3. Held to maturity. The originator keeps the loan until repaid; most private loans “don’t trade in secondary markets”. [D — New York Fed; BIS] A syndicating bank keeps a sliver; a fund keeps it all.
4. Covenants and private ratings. One holder can write and enforce covenants — the BIS calls these loans “covenant-heavy”, built on cash flow rather than collateral — and can use a private rating, or none. [D — BIS; FSB]
5. Unitranche. One lender, one loan, one blended rate, instead of senior and junior tranches from different lenders. The borrower buys speed and one counterparty. [I — industry-understood; the FSB cites demand for “tailored financing solutions and fast execution”]

The contrast is the broadly syndicated loan (BSL): bank-arranged, agency-rated, sold to many investors and CLO vehicles, traded daily, covenants loosened as the buyer base widened. [I] The two markets now swap borrowers: in 2025, per PitchBook LCD data reported by trade press, $34.1 billion of direct-lender loans refinanced into the syndicated market and $36.9 billion went the other way, both records since tracking began in 2022. [D — see Sources]

---

5. Size and share

Every “private credit is $X trillion” headline uses a different definition. The table gives the definition and date with each number.

All retrieved 3 September 2026. [D]

The order of magnitude is settled and the number is not: $1.5tn to $2.5tn is the honest range, and the spread is definition, not error — the FSB’s US figure (~$1tn) sits below the Federal Reserve’s because it counts only bilateral lending to medium-sized companies from regulatory returns. Private credit is now the size of each of the two public markets it grew out of. [D — FSB] The share of buyout financing, the number most quoted, is deliberately absent: the LCD series is not among the sources listed.

---

6. The circle closes: banks lend to the funds

The loan left the bank’s book. The bank then lent to the fund that holds it.

The measured part. Y-14 data from the largest US banks show $95 billion of committed credit lines to private credit vehicles at end-2024 — $56 billion to BDCs, $40 billion to private debt funds — up 145% in five years, 56% drawn. [D — FEDS Notes, May 2025] Across FSB members the reported figure is about $220 billion, “less than 0.5% of total bank assets” where separately identified; commercial data suggest $270–500 billion. [D — FSB, May 2026] Why banks do it is the same arithmetic in reverse: a loan to a diversified fund carries a lower risk weight than a loan to the fund’s borrower. The Kansas City Fed estimates a 29.2% return on equity on bank loans to private credit funds against 7.9% on ordinary commercial lending, with “the lower risk weight treatment of loans to private credit funds” as the main driver. [D — August 2025]

The partnerships, from the banks’ and platforms’ own releases: Citi and Apollo (26 September 2024), a $25 billion programme in which Citi originates and Apollo, Athene and Mubadala fund; J.P. Morgan (24 February 2025), $50 billion of its own balance sheet plus about $15 billion from co-lenders; Overland Advantage, a BDC in which Wells Fargo sources from its middle-market clients and holds a minority stake while Centerbridge underwrites, about $7 billion underwritten since May 2024. [D]

What is measured, and what is not. The IMF’s April 2024 chapter warned of “multiple layers of leverage, often hidden by reporting gaps” from borrower to fund to end investor. [D] The ECB’s supervisory chair wrote in June 2025 that banks “face challenges when aggregating exposures across business lines or counterparty types”, that valuations “are often based on data provided by the funds themselves”, and that banks cannot systematically see when they co-lend to a company alongside a fund they also finance. [D — Buch, ECB, 3 June 2025] The FSB says uncertainty around bank exposure “is relatively large”, with commercial estimates “more than twice” its members’ data. [D] The Federal Reserve’s May 2026 report judged redemption risk “limited and manageable” but recorded that in Q1 2026 accepted redemptions from perpetual BDCs exceeded new inflows for the first time. [D]

So what: the capital rule moved the loan to a place where it is not risk-weighted, and the bank’s remaining exposure — a line to the fund, a share in a partnership, a position in the same borrower — is measured piecemeal by different authorities using different definitions. The loan is safer for the bank than in 2007. Whether the system is safer is the question each of these reports says it cannot yet answer.

---

Caveats

• The US Basel III endgame is a proposal, not a rule. The 19 March 2026 proposals closed for comment on 18 June 2026; the Federal Reserve’s June 2026 supervision report still describes them as proposals, and no final rule was retrieved. Any statement about their effect is about a proposal.
• The 4.8% CET1 reduction comes from a law firm’s reading of the agencies’ impact tables, not from the press release, which says only “modestly decrease”. The 2023 figure of +16% is the agencies’ own.
• The pre-Basel III minimum common-equity ratio (widely reported as 2%) could not be verified from a retrieved primary page and is therefore not stated as a figure.
• The capital charge table is a floor, not a bank’s actual charge. National implementations, G-SIB surcharges (not retrieved in this session), stress-test buffers and internal models all move the number. The direction of the argument holds; the £7–£13 is the Basel floor for one loan.
• Lock-up periods disagree. The New York Fed says LP capital is locked “often for five to seven years”; the Federal Reserve’s May 2026 report says private debt funds are “locked up from 7 to 10 years”. Both are quoted; the difference is fund type and vintage.
• US private credit size disagrees by source. FSB ~$1tn versus Federal Reserve $1.34tn–$1.4tn; the definitions differ as set out in the table. The share of LBO financing is deliberately not stated because the LCD pages could not be retrieved.
• The 2025 LBO counts and refinancing flows are PitchBook LCD figures relayed by trade press, not retrieved from PitchBook.
• Unitranche, SOFR/Euribor pricing and the description of the broadly syndicated loan are marked [I]: industry-understood, with no single disclosing source.
• “Cheaper holder” reasoning in section 3 is [S] — our inference from the capital rules and the BIS and Kansas City Fed findings, not a measured cost comparison.
• Nothing here is a view on whether private credit loans are good or bad assets, or on what the 2026 US proposals will do to the flow if adopted.

Sources [D]

All retrieved and confirmed to load on 3 September 2026.

Basel Framework (BIS)

• RBC20 — Calculation of minimum risk-based capital requirements — paragraph 20.1: CET1 4.5%, Tier 1 6%, total 8% of RWA, effective 1 January 2023.
• RBC30 — Buffers above the regulatory minimum — 30.2: 2.5% CET1 conservation buffer; 30.9: countercyclical buffer between 0 and 2.5%.
• CRE20 — Standardised approach: individual exposures — corporate risk weights: 20% AAA to AA-, 100% unrated, 150% below B+.
• History of the Basel Committee — Basel III endorsed at Seoul, November 2010, issued December 2010; post-crisis reforms completed December 2017 with an output floor.

US capital rules and guidance

• Agencies request comment on proposed rules to strengthen capital requirements for large banks — Federal Reserve, 27 July 2023 — $100 billion threshold; estimated 16% aggregate CET1 increase.
• Agencies request comment on three proposals to modernize the regulatory capital framework — Federal Reserve, 19 March 2026 — overall capital “would modestly decrease”; comments due 18 June 2026.
• Fact sheet: Proposals to modernize the regulatory capital framework — 19 March 2026 — corporate risk weight cut from 100% to 95% under the standardised approach.
• Statement by Vice Chair for Supervision Bowman — 19 March 2026 — activity “pushed … into the less-regulated non-bank sector”; aim to “reduce incentives for traditional lending activities … to migrate outside of the regulated banking sector”.
• Supervision and Regulation Report, June 2026 — Regulatory developments — the three March 2026 proposals listed as proposals with comments due 18 June 2026.
• Banking agencies release Basel III, GSIB surcharge and revised standardized approach proposals — Sullivan & Cromwell, March 2026 — law-firm reading of the impact estimates: about 4.8% lower aggregate CET1 for Category I–II banks.
• SR 13-3: Interagency Guidance on Leveraged Lending — Federal Reserve, 21 March 2013 and the guidance text (PDF) — definition examples (buyout proceeds; total debt above 4.0× EBITDA); 6× “raises concerns”; 50% repayment over five to seven years.
• Macroprudential Policy and the Revolving Door of Risk — Kim, Plosser and Santos, New York Fed Staff Report 815, May 2017 — the guidance reduced bank leveraged lending and triggered migration to nonbanks, which increased their bank borrowing.
• PRA announces a delay to the implementation of Basel 3.1 — Bank of England, 17 January 2025 — to 1 January 2027, citing US uncertainty; full implementation 1 January 2030.

BDC leverage rule

• Staff responses regarding business development companies and Section 61(a) — SEC — Section 61(a) amended in 2018 to permit asset coverage of 150% rather than 200%, with board or shareholder approval.
• Spending legislation contains long-awaited reforms for BDCs — Proskauer, 2018 — enacted 23 March 2018 in the Consolidated Appropriations Act; 200% ≈ 1:1 debt to equity, 150% ≈ 2:1.

Size, structure and bank linkages

• Bank Lending to Private Credit: Size, Characteristics, and Financial Stability Implications — Federal Reserve FEDS Notes, 23 May 2025 — US $1.34tn and global ~$2tn at 2024 Q2; $95bn committed lines ($56bn BDCs, $40bn funds), 56% drawn.
• The Price of Bank Funding Behind Private Credit: Evidence from BDCs — FEDS Notes, 7 August 2026 — bank loans ~40% of BDC debt, up from 20% a decade earlier; ~90% as credit lines.
• Financial Stability Report, May 2026 — Federal Reserve (PDF) — Box 4.1: ~$1.4tn, 10% of US non-financial corporate debt, about one-third of below-investment-grade debt ex bank loans; funds locked 7–10 years; Q1 2026 redemptions exceeded inflows; risk “limited and manageable”.
• The global drivers of private credit — BIS Quarterly Review, March 2025 — >$2.5tn, ~87% US; stricter bank regulation “has spurred lending by private credit funds”; BDC–bank cost gap narrowed ~200bp 2010–2019; floating-rate, directly negotiated, held to maturity, covenant-heavy.
• Report on Vulnerabilities in Private Credit — FSB, 6 May 2026 (PDF) — $1.5–2tn at end-2024, US ~$1tn; comparators; ~$220bn member data vs $270–500bn commercial; subscription lines “most common”, ~200bp; NAV 200–400bp; private ratings; data gaps.
• NBFIs in Focus: The Basics of Private Credit — New York Fed Teller Window, 17 October 2025 — definition; ~$1.3tn ($800bn funds, $500bn BDCs) at December 2024; LP lock-up five to seven years; loans do not trade.
• Banks and Private Credit: Competitors or Partners? — Kansas City Fed Economic Bulletin, 6 August 2025 — 29.2% vs 7.9% return on equity; lower risk weight on loans to funds as the main driver; funds lend at higher spreads and loss given default.
• Global Financial Stability Report, April 2024, Chapter 2 executive summary — IMF (PDF) — “multiple layers of leverage, often hidden by reporting gaps”; recommendation for closer supervision of funds, investors and leverage providers.
• Hidden leverage and blind spots: addressing banks’ exposures to private market funds — Claudia Buch, ECB Banking Supervision blog, 3 June 2025 — aggregation challenges; fund-supplied valuations; no systematic look-through to co-lending.
• Preqin Global Report: Private Debt 2025 (PDF) — AUM $1.50tn at end-2023; forecast $2.64tn by 2029 at 9.88% CAGR; ex-renminbi funds.
• 2025 private credit growth driven by LBO comeback — Alternative Credit Investor, 9 January 2026 — trade-press relay of PitchBook LCD: 214 direct-lending LBO deals and $81.4bn in 2025 (248, $72.9bn in 2024); $34.1bn and $36.9bn cross-market refinancings.

Bank–fund partnerships (primary releases)

• Citi and Apollo announce $25 billion private credit, direct lending program — Citi, 26 September 2024 — Citi originates; Apollo, Athene and Mubadala fund.
• J.P. Morgan increases direct lending commitment to $50 billion — 24 February 2025 — $50bn balance sheet plus ~$15bn co-lenders; >$10bn across >100 deals since 2021.
• Overland Advantage finances ~$4B across 18 transactions in 2025 — PR Newswire, 29 January 2026 — Wells Fargo sources and holds a non-controlling minority stake; Centerbridge underwrites; ~$7bn since May 2024.