Direct Lending's 23% Software Exposure Is the Next Credit Cycle

4 hours ago

2 min read

What's New

Direct lending portfolios carry software concentration that no diversification rule would allow, and the valuation reset in that sector has already started. Bruce Richards, CEO of Marathon Asset Management, argues this in a podcast on Russell Investments' Without Boundaries. Software makes up 23% of direct lending. Leverage on those loans runs 6 to 10 times debt to EBITDA, averaging around 8 times. Enterprise values in software have been cut in half. Allocators should ask managers for software exposure by name, not inside a broad technology bucket.

Why It Matters

The argument puts direct lending managers on the other side, and BDC boards with them. The standard view holds that covenants and lender control make direct lending safer than the broadly syndicated market. On sector risk the ranking flips. Richards' own ceiling is 10 to 15% in any one sector. Marathon sells a multi-asset credit strategy spanning public and private markets, so a concentration argument supports what it sells.

By The Numbers

3% Software share of the high yield market, which Richards treats as the market that priced this correctly.
13% Software share of the broadly syndicated loan market. Senior secured software loans there carry 5.2 times leverage.
26% Software exposure at the top BDCs, the highest of any credit market he names.
0.5 times Average debt at a public software company, which is why he gives public names a fighting chance and private borrowers less of one.

Reality Check

Richards concedes that the core of his case is unresolved. He expects loss rates on 2020 to 2024 direct lending vintages to rise, then says the outcome is not yet known. The 1% industry loss rate he cites comes from managers who also set their own marks. His claim that Marathon's loss rate will be close to zero rests on no track record given in the conversation.

Memorable Quotes

“The top BDCs have 26% exposure to software.” Richards names the vehicles carrying the concentration, and retail investors own many of them.
“How you have more than 10 to 15% of your portfolio in any one sector is unexplainable to me. It's that simple.” This is the ceiling he runs Marathon to, and it is roughly half what direct lending carries in software.
“1% for direct lending isn't great, but it's not bad either.” He concedes current loss rates are survivable. By his own math a 1% loss rate costs about a point of yield.
“2026 is the year of the awakening where people have recognized the valuation reset for software.” He puts a date on the repricing, which makes the call testable against this year's marks.

The Wrap

Two forces have to meet for this call to land. Software borrowers from the 2020 to 2024 vintages reach their maturities with enterprise values still halved. Defaults then arrive in a sector where recovery value runs low once subscribers leave. Reinvention is the escape route, and it works only for borrowers whose free cash flow survives the move to AI-first products. Marks on the legacy vintages between 2027 and 2029 will show which path the sector took.