Half of all firms take two quotes before choosing a lender.

Where do base rates go from here?

Up, on the evidence of the curve rather than the commentary. The Federal Open Market Committee held at 3.50% to 3.75% on 29 July 2026, its fifth consecutive hold, on a 9 to 3 vote in which all three dissents were for a hike. That was the first time since September 2016 that three policymakers dissented in the same direction.

The forward market agrees. Three-month SOFR is priced at 3.90% at the end of 2026 and 4.04% at the end of both 2027 and 2028 (Blue Gamma, 4 August 2026), against a three-month Term SOFR fix of 3.76% and a one-month fix of 3.65% on the same day. The eleven basis point gap between the one-month and three-month fixes is the term curve pricing a move inside the next quarter. The Committee's own June projections put the 2026 median at 3.8% and the 2027 median at 3.6%, both above where the range sits today.

Named house views for 2027 span roughly 100 basis points, from two cuts (Goldman Sachs Research, 9 June 2026) to a next move that is a hike, advanced to December 2026 (JPMorgan, 30 July 2026). That spread is itself the finding. Nobody with a published 2027 call is confident, and the market is charging borrowers for that.

The borrower-facing conclusion is unglamorous. Terming out floating-rate debt now is a decision that does not require a view on the Fed. Waiting for cuts is a position, and the curve is currently pricing against it.

What does the market actually charge by size?

Between 125 and 175 basis points more for being small, at the same seniority. A borrower below $10m of EBITDA is quoted S+550 to S+750 for senior non-bank unitranche debt; a borrower above $25m of EBITDA is quoted S+425 to S+575 (SPP Capital Partners, Market At A Glance, July 2026). On a $20m facility, 150 basis points is $300,000 a year, and none of it buys the smaller borrower anything.

That premium is a function of size, packaging and competition rather than of credit. The most precise academic reading available says the same thing about dispersion generally: pricing differences across lenders do not appear to be explained by risk, over a third of firms behave as if they do not comparison shop, and half of all firms appear to obtain only two quotes before picking a lender (Amiti, Kashyap, Kovner and Weinstein, NBER Working Paper 34870, February 2026).

Two published grids disagree and the difference matters when a borrower is reading them. A portfolio index whose median company carries $60.6m of last-twelve-months adjusted EBITDA prints core middle-market unitranche at S+475 to S+550 (Lincoln International, Private Credit Snapshot, as of 1 May 2026), well inside the placement survey above. They are measuring different populations. For a founder-owned business below $10m of EBITDA, the placement survey is the honest number.

Leverage has moved with price, and against the small end hardest. Total debt for sub-$10m EBITDA issuers now clears at 2.50x to 3.25x, down from 2.50x to 4.00x in July 2025, a loss of three-quarters of a turn in a year, while the senior band lost half a turn. Above $25m of EBITDA, total leverage still clears at 5.00x to 6.50x. Minimum equity is 40% of capitalization with at least 60% of it new cash.

All-in, the spot unitranche yield is 9.00% to 9.75%, with swap-adjusted yields up roughly 47 basis points quarter over quarter to 9.47% to 10.22% (Valuation Research Corporation, Private Markets Trends Q2 2026, July 2026). The comparator a quarter earlier was 8.9% all-in at S+4.75% to S+5.50% (Lincoln International, Q1 2026, published 7 May 2026). Note which number moved: forward base rates, not spreads, are driving most of the increase in all-in cost.

Are the banks retreating?

Not in 2026, and this is the most commonly misstated fact in the market. The net percentage of domestic banks tightening commercial and industrial standards was 0.0% for large and middle-market firms and 1.8% for small firms in the July 2026 senior loan officer survey, down from 8.1% and 6.6% in April. The Federal Reserve's own summary language is that banks reported basically unchanged standards for firms of all sizes and eased or left basically unchanged all queried terms.

On price the banks are competing hard. A net 26.8% of banks narrowed spreads over their cost of funds to large and middle-market firms, and a net 14.5% to small firms. Commercial and industrial standards are easier than their historical midpoint since 2005, the only loan category of which that is true, and bank commercial and industrial loans stood at $2,894.2bn in June 2026, up 8.0% year on year after being flat through 2025.

The retreat that is real is structural rather than current, and it is specific to leveraged lending. Direct lenders still finance roughly 90% of buyout transactions below $500m of enterprise value, individual bank hold sizes in middle-market leveraged lending have contracted from $75m to $100m down to $30m to $50m, and several large regional banks have exited the segment entirely since 2024 (ABF Journal, 19 March 2026). A borrower whose deal looks like a sponsor buyout meets a different bank from the one a relationship borrower meets.

Regulatory relief is directional and not bankable. On 19 March 2026 the federal banking agencies proposed cutting the corporate exposure risk weight from 100% to 95%; the comment period closed on 18 June 2026 and there is no final rule and no compliance date. Do not underwrite 2027 capacity on a proposal.

Is 2027 the maturity wall?

No. 2028 is. Some $39bn of Morningstar LSTA index loans mature in 2027 against $230bn in 2028 (LevFin Insights, CR TrendLines July 2026, as of 30 June 2026), and maturities through the end of 2027 have already fallen to $32bn from $62bn at the end of 2025, with $129bn pushed into 2029 and beyond (PitchBook LCD, 17 July 2026). Total leveraged loan market outstandings were $1.57tn at the end of June 2026.

Borrowers bought that time deliberately. Amend-and-extend activity ran $106bn year to date through June, up 26% year on year, including $27bn in June alone and $39bn of institutional volume in the second quarter, and private-equity-backed borrowers drove 74% of maturity-extension amendments (PitchBook LCD, 17 July 2026). The quality of what was extended improved sharply: BB minus or higher rose to 30% of amend-and-extend volume in 2026 from 11% in 2025, while B minus fell to 27% from 44%.

Three things spoil the reprieve. The opportunistic refinancing pipeline has largely exhausted itself, with refinancing issuance down 40.6% year on year in the first quarter, because the borrowers who could refinance into tighter spreads already did. The $40bn 2028 software maturity wall has barely moved through the first half of 2026. And borrowers of the 2021 and 2022 vintages carry roughly 0.9x more leverage and roughly 0.4x less adjusted cash interest coverage than at underwriting, which keeps their refinancing risk elevated through 2027 (Valuation Research Corporation, Q2 2026).

The lower middle market is where refinancing is hardest rather than merely dearer. New capital there commonly demands lower pro forma leverage, legacy lenders exiting at a discount, debt converted into subordinated or equity securities, or new equity as a condition precedent (SPP Capital Partners, July 2026). Those are ownership outcomes, not pricing outcomes, and they are decided long before the maturity date.

How much stress is actually in the system?

Less than the worst number suggests and more than the best one does, because the same market currently prints four different default rates and the definition drives most of the gap. Payment default runs 1.5% by volume (KBRA DLD, May 2026), documentation default 2.51% (Proskauer Private Credit Default Index, 28 July 2026, across 716 loans and $195.6bn), covenant default 3.1% (Lincoln International, as of 31 March 2026) and a liability-management-inclusive measure 6.0%, a record, of which over half of second-quarter events were maturity extensions rather than missed payments (Fitch Ratings, 30 July 2026).

The global regulator reaches the same conclusion independently: outright defaults run around 1%, rising to around 5% once selective defaults are counted (Financial Stability Board, Report on Vulnerabilities in Private Credit, 6 May 2026). Anyone quoting one of those four numbers without its definition is quoting a preference rather than a fact.

The size cut is counter-intuitive and currently favours the small borrower. Defaults below $25m of EBITDA fell to 1.9% in the second quarter from 2.3%, while the $25m to $49.9m band rose to 3.4% and was the only band to deteriorate (Proskauer, 28 July 2026). The offsetting point is that valuations are thinnest at the small end: 12% of borrowers in the $10m to $20m EBITDA bracket were marked below 90% of par against 3% above $100m of EBITDA (ACC and AIMA Quarterly Update, as of March 2026).

The signal worth watching is not the default rate at all. Payment-in-kind interest sat on 10.6% of loans and represented 8.9% of total interest income in the first quarter of 2026, and what the lender calls bad payment-in-kind, meaning credit-driven rather than structured, was 5.9% of all loans against 6.4% in the fourth quarter of 2025 and 2.5% at the end of 2021, with loan-to-value on those loans up 33.5 points to 76.0% in a year. Lincoln International calls that a shadow default rate. Lender takeovers of $24.2bn in 2025 plus $15.2bn year to date in 2026 compare with $13.6bn across the preceding three years combined.

Recoveries are genuinely contested and both readings are current. One dataset of restructurings observed through business development companies puts recoveries near 50 cents (Octus, 11 May 2026); a ratings cohort puts them at 70% to 90% with minimal realized losses (Fitch, 6 March 2026). They measure different populations. Print both or neither, and never average them. The Financial Stability Board's own summary judgment is the one to hold onto: private credit remains untested to a prolonged economic downturn.

What should a borrower do before 2027?

Start the process earlier than the maturity requires, and get more than two quotes. Those two moves are inside a borrower's control and the evidence says most borrowers make neither. Everything below is the market a borrower who starts late will meet instead.

Documentation is tightening decisively in private credit even where price is not. 98% of surveyed lenders say their underwriting became notably stricter since the start of 2026; the share expecting looser documents fell from 33% to 4% in a year while the share expecting tighter documents rose from 13% to 56%; and 55% would not provide payment-in-kind flexibility on a new buyout (Houlihan Lokey, Q2 2026 Private Credit Survey). Original issue discount is moving too: lenders quoting 99 or tighter fell from 75% to 56% year on year, and quotes at 98.5 more than doubled from 18% to 42%.

Sector matters more than it did. Software carries 75 to 100 basis points above the standard matrix on one lender survey and 150 to 300 basis points above comparable non-software credits on another, with interest-only periods cut from three years to two or less and cash controls now standard. That is the single largest sector adjustment in the market.

The one clearly expanding source of lower-middle-market capacity is the licensed small business investment company program. Thirty-six new licences were issued in the 2026 fiscal year through a licensure date of 15 June 2026, twenty-five of them at the 2.00x leverage tier, and the Investing in All of America Act, signed on 21 May 2026, lifted the standard debenture leverage ceiling to the lesser of 200% of private capital or $250m, raised aggregate commonly-controlled leverage to $475m, and exempted rural, manufacturing and critical-technology investments from the leverage cap. For a sub-$25m EBITDA borrower that is where new lenders are actually appearing.

Finally, plan against the forecast rather than on it. The most direct named call for middle-market pricing into the second half of 2026 is that spreads tick up 25 to 50 basis points, particularly for lower-quality credits and difficult sectors (PitchBook, quoted in Capstone Partners, 18 May 2026). No named source publishes a 2027 spread forecast for high yield, loans or direct lending. The rating agencies publish full 2027 default outlooks in November and December. A borrower whose 2027 plan depends on knowing those numbers does not have a plan.