In private credit, the marks haven't moved. The risk has.
I ran AI across 10 major public BDC quarterly filings last month: ARCC, BXSL, OBDC, FSK, HTGC, PSEC, OCSL, GSBD, MAIN, and TPVG. The finding: software positions are largely still marked near par. Consumer names are where the visible stress is. Enterprise software, where AI disruption risk is highest, has barely moved.
Nobody commissioned that analysis. There is no product that produces it. I built it myself. Loaded the filings into a single AI context, ran structured queries across all 10, and synthesized the output into a comparable table. The alternative was reading 2,000 pages of supplements by hand.
The classification problem alone tells you something. Bloomberg found more than $9 billion in software investments are misclassified across public BDC filings. Labeled "business services" by one manager, "software" by another lender on the same borrower. The reported exposure numbers are wrong before the analysis even starts.
This is not a niche issue. Across these 10 BDCs there is roughly $18 to $22 billion in software or software-adjacent exposure. Nobody has a clean systematic view of how it is marked, what the covenant structures look like, or where the overlapping credits sit across managers.
We are already in the early innings of the next credit cycle. The edge will go to whoever can synthesize across portfolios fastest. Right now there is no commercial tool built specifically for this across public and private credit.