Why does a loan now yield more than a bond?
Because the two markets no longer hold the same borrowers. On KKR's figures the US leveraged loan index yields 8.1 percent against roughly 7.3 percent for the US high yield index, a gap of about 100 basis points in the loan's favour.
KKR's credit team is explicit that the premium does not reflect better fundamentals on the bond side. It names three causes on the loan side: the software overhang, CLO technical pressure and documentation concerns that have accumulated in the loan market. In its words, the loan market is offering more yield today precisely because it carries more risk.
That is a different explanation from the one most borrowers carry in their heads. A floating-rate loan that reprices every quarter has usually been cheaper than a fixed-coupon bond with call protection, and the spread between them was read as the price of that flexibility. The 2026 gap runs the other way, and it is being attributed to what sits in each index rather than to rates.
The headline spread picture on the bond side also needs reading with care, and KKR reads it that way. High yield spreads sit near the 96th percentile of their post-crisis range, which looks stretched. Effective duration is roughly three years, near a fifteen-year low against a long-run average of about four. Adjusted for that, spread per unit of duration sits at only the 74th percentile. The bond market is not as expensive as its headline suggests, and the loan market is not as cheap.
Who is leaving the loan market, and where are they going?
The highest-rated leveraged borrowers, into bonds. BB-rated credit now makes up a record 57 percent of the US high yield market and 68 percent of the European market. The loan market, by KKR's description, is approximately 60 percent single-B.
The migration is visible over months, not years. In the US the BB cohort of the bond market grew from 1.19 trillion dollars in January 2026 to 1.25 trillion dollars in May, while the CCC bucket shrank from 213 billion dollars to 193 billion dollars over the same period. The bond market is getting larger at the top and smaller at the bottom at the same time.
New issuance tells the same story more sharply. BB-rated paper accounted for 39 percent of US high yield issuance year to date, an all-time high for any comparable period, and swung to a 59 percent share of month-to-date June supply from 30 percent in May. Single-B issuance over the same month collapsed to less than 9 percent, a fifteen-month low. Year-to-date pro forma volume of 189 billion dollars through early July was up more than 20 percent on the 157 billion dollars priced at the same point in 2025, after a seven-month high in April.
KKR's reading is that this is structural rather than a window. Its phrase is that transactions which would previously have been financed entirely in the loan market are increasingly requiring bond market participation to be completed, and it expects upcoming refinancings to include high yield in a more significant way. The queue is real: about 7 percent of the loan market matures within two years, more than 100 billion dollars of par that will have to decide which market it refinances in, alongside roughly 10 percent of the bond market on the same horizon, the highest share since before the financial crisis.
What is the bond market offering that the loan market is not?
Certainty, at a price. A non-call high yield bond fixes the cost of capital for a defined period, removes repricing risk and reaches an investor base that is not driven by CLO demand. KKR's judgement is that as CLO appetite has become more selective and direct lending terms have tightened, that execution certainty has become worth paying a modestly higher coupon for.
It is also a more senior market than its old name implies. First lien secured bonds are now 33 percent of the US high yield market, an all-time high, with total secured exposure approaching 36 percent in the US and 38 percent in Europe. The data centre financing market, now roughly 3 percent of the US index and almost entirely secured, is one of the newer contributors.
And it is a market with less of the sector that has hurt the others. Software is roughly 3 percent of the US and European high yield markets, against approximately 13 percent of the US leveraged loan market and more than 20 percent of US direct lending. When software credit repriced earlier this year, software loan bids fell to their lowest reading in four years and the gap to non-software loans widened to more than eight points by the end of the first quarter. The bond market, with a fraction of the exposure, held materially better. The companion article on this site carries that price split in full.
Disclosure sits underneath all of it. Because high yield bonds are issued under securities law, issuers generally carry offering memoranda, quarterly calls and in many cases full annual and quarterly filings. A bond investor can see more than a loan holder can, and is being paid less for the privilege.
| Measure | US high yield bonds | US leveraged loans | As at |
|---|---|---|---|
| Index yield | About 7.3% | 8.1% | Early August 2026 |
| Share rated BB (bonds) or single-B (loans) | 57% BB, a record | About 60% single-B | 15 July 2026 |
| Software and services share of market | About 3% | About 13% (direct lending: over 20%) | 1 May 2026 |
| First lien secured share | 33%, an all-time high | Secured by construction | 15 July 2026 |
| Effective duration | About 3 years, near a 15-year low | Floating rate, reprices quarterly | 15 July 2026 |
| BB share of new issuance, month to date June | 59%, from 30% in May | Not applicable | 30 June 2026 |
| Single-B share of new issuance, month to date June | Under 9%, a 15-month low | Not applicable | 30 June 2026 |
| Par maturing within two years | About 10%, highest since pre-crisis | About 7%, over 100 billion dollars | 30 June 2026 |
“Owners read the loan index as a neutral benchmark for what debt costs. It is not neutral any more. It is the price of whoever is still in it after the best credits have walked up a flight to the bond market, and a lender quoting off that number is passing a composition premium down to borrowers who had nothing to do with it.”
Does any of this reach a company too small to issue a bond?
Yes, in three ways, and this is our read rather than KKR's, which writes for allocators and says nothing about middle-market borrowers. A company of the size most owners here run is not a bond issuer and is unlikely to become one. It is still priced, directly or by reference, off markets that are being sorted above it.
The first is the benchmark. A lender that quotes a margin by reference to where the loan index or its sector comparables yield is quoting a number that now carries a premium for software overhang, CLO technicals and weak documentation. None of those is a fact about the borrower in front of it. It is worth asking a lender, in terms, which comparables it is pricing against and how much of the reference yield it would attribute to composition rather than credit.
The second is the lender's own book. Direct lending carries more than 20 percent exposure to software on KKR's figures, the highest of the three markets. A lender working through a concentrated sector problem tightens terms across its whole book, because that is how credit committees behave, and a borrower in an unrelated sector pays part of the cost. Knowing a prospective lender's sector mix is as useful as knowing its pricing grid.
The third is the queue. More than 100 billion dollars of loans and roughly a tenth of the bond market need refinancing within two years, and the largest borrowers are taking up investor attention as they decide between the markets. A middle-market refinancing does not compete with those deals for the same capital, but it does compete with them for the same underwriters' time and the same risk appetite. Starting a process early, while that queue is forming rather than after it has formed, is the one lever here that an owner controls.
What would change this picture?
The issuance mix, first. The 59 percent BB share of June supply and the single-B share below 9 percent are monthly figures and monthly figures revert. If single-B issuers return to the bond market in volume, the sort between the two markets is a window rather than a structure, and KKR's structural reading weakens on its own evidence.
The yield gap, second. A loan index yielding 100 basis points over a bond index is the observable symptom. If it closes because loan yields fall, the composition premium this article describes is being priced out. If it closes because bond yields rise, something else is happening and the explanation here no longer fits.
The software bid, third. KKR attributes the loan market's extra yield partly to a sector overhang, and the companion article on this site tracks that overhang in price terms. A narrowing of the software gap would remove the largest of the three stated causes.
Hold the frame where KKR puts it. These are market-level composition and yield figures as at dates between May and July 2026, attributed by KKR to Bloomberg, ICE BofA, FactSet, PitchBook LCD and Morningstar, in a note whose stated purpose is to make the case for an allocation to high yield. The durable finding for a borrower survives that purpose: the instrument that used to be the cheap one now yields more, because of who is left in it.
As of August 2026
Sources: KKR Credit, 'High Yield's Second Act: What AI Revealed About Credit Quality', published on kkr.com on 4 August 2026, with a companion opinion piece by Christopher Sheldon in the Financial Times dated 3 August 2026, for the US high yield index yield of roughly 7.3 percent and the US leveraged loan index yield of 8.1 percent, about 100 basis points apart, and for the attribution of that premium to software overhang, CLO technical pressure and documentation concerns rather than better fundamentals; for BB-rated credit at a record 57 percent of the US high yield market and 68 percent of the European market, and the loan market at approximately 60 percent single-B; for the US BB cohort growing from 1.19 trillion dollars in January 2026 to 1.25 trillion dollars in May and the CCC bucket shrinking from 213 billion dollars to 193 billion dollars over the same period; for BB-rated paper at 39 percent of US high yield issuance year to date, 59 percent of month-to-date June supply against 30 percent in May, single-B issuance below 9 percent at a fifteen-month low, year-to-date pro forma volume of 189 billion dollars through early July against 157 billion dollars at the same point in 2025, and a seven-month high in April 2026; for effective duration of roughly three years near a fifteen-year low against a long-run average of about four, spreads near the 96th percentile since the financial crisis and spread per unit of duration at the 74th percentile; for first lien secured bonds at 33 percent of the market, total secured exposure approaching 36 percent in the US and 38 percent in Europe, and data centre financing at roughly 3 percent of the US index; for software at roughly 3 percent of the US and European high yield markets against approximately 13 percent of US leveraged loans and over 20 percent of US direct lending, software loan bids at their lowest in four years and a gap of more than eight points to non-software loans at the end of the first quarter of 2026; for about 7 percent of the loan market, more than 100 billion dollars of par, maturing within two years and about 10 percent of the high yield market on the same horizon, the highest share since before the financial crisis; and for the description of a non-call bond as fixing cost of capital, removing repricing risk and reaching an investor base not subject to CLO-driven demand. KKR attributes its exhibits to Bloomberg, ICE BofA, FactSet, PitchBook LCD, Morningstar and KKR Credit Analysis as at 1 May, 30 June and 15 July 2026. The views in that note are those of its authors, are stated by KKR not to be research, and are published in support of an allocation case for high yield. The reading of the loan index as a benchmark carrying a composition premium, the suggestion that a lender's sector book is part of the borrower's price, the point about refinancing queues and underwriter attention, and the questions we suggest putting to a lender are entirely our own and carry no figure. A companion article on this site carries the software-versus-rest split of the loan market in secondary price terms, another covers CLO issuance as a signal of loan market conditions, and another covers the choice between private credit and the syndicated market, which sits one layer below the comparison made here.