Six alternative branches. These are subjective probabilities from the seed, not observed frequencies or investment recommendations.
Why this trend
Load-bearing assumption
Private credit portfolios continue to be valued at quarterly, manager-determined marks with roughly 70% covenant-lite issuance, so deterioration becomes visible with a lag of at least two quarters rather than in real time.
Preconditions
Not recorded.
The shadow rate converges down
Credit normalises without an event. The bad-PIK share drifts down from 6.4% as some borrowers refinance and others are restructured quietly; headline defaults stay in the 1-3% range; redemption pressure in evergreen vehicles fades as returns recover from the 1.6% year-to-date print through April 2026. Banks keep growing NDFI lending, but at a decelerating rate as the base gets larger. Nothing breaks and nothing is resolved.
Mechanism
The decelerating element is NDFI lending growth, which falls from 22.4% year over year toward the low teens as the denominator grows and supervisors ask more questions. Managers mark conservatively enough to avoid a discontinuity, borrowers with negative free cash flow either recapitalise or are handed to lenders, and the transition happens loan by loan rather than as a cohort event.
Preconditions, early indicators and assumptions
- Calibration Basis
base_rate
- Preconditions
- Policy rates move by no more than roughly 100bp in either direction
- No large sponsor default that forces simultaneous re-marking across managers
- Evergreen redemption requests stay near or below the 5% quarterly cap
- Early Indicators
- Indicator
Aggregate FDIC net interest margin stays within 3.20% and 3.45% for four consecutive quarters
- Source
S-07-01
- Threshold
3.20-3.45%
- Direction
within band
- Would Be Visible By
2027-09
- Resolves
Confirms the benign macro leg of base
- Indicator
Non-accruals stay below 3% of fair value in BDC 10-Q filings for the ten largest publicly filing BDCs
- Source
S-07-20
- Threshold
3% of fair value
- Direction
below
- Would Be Visible By
2027-11
- Resolves
Confirms base; above 3% at three or more managers moves mass to downside
- Indicator
H.8 NDFI loan growth decelerating to below 15% year over year without turning negative
- Source
S-07-02
- Threshold
between 0% and 15% y/y
- Direction
within band
- Would Be Visible By
2027-07
- Resolves
Confirms orderly deceleration rather than withdrawal
- Assumptions
- Text
The PIK-implied distress measure is a leading indicator of payment default that partially resolves benignly rather than a delayed recognition of losses already incurred
- Confidence
medium
- Load Bearing
true
- Basis
FSB 2026-05-06 and Lincoln International via CAIA 2026-04-20 record these as measuring different things, not competing estimates of the same thing
- If Wrong
Base and downside swap probability mass and the downside indicator has already fired
- Text
Quarterly manager marks do not require a discontinuous reset
- Confidence
medium
- Load Bearing
false
- Basis
No auditor-forced re-marking event is recorded in the corpus
- If Wrong
The transition to downside is abrupt rather than gradual
- Affected Industries
- 07
- 11
- 18
- Precedence Note
Distinguished from downside by whether the PIK-implied and headline default rates converge or diverge over three consecutive quarters of BDC filings, not by the level of either.
- Would Change Our Mind
The gap between the roughly 2% headline and roughly 6% PIK-implied rates widening for three consecutive quarters.
- What Businesses Should Do
Mid-market borrowers should assume refinancing is available but repriced, and should test covenants against a two-quarter reporting lag rather than against current performance.
Refinancing reopens and PIK unwinds
The constraint that releases is the cost of refinancing. If policy rates fall, the broadly syndicated market reopens to single-B credits and PIK toggles convert back to cash pay. The shadow default rate re-converges on the headline from above, evergreen redemption queues clear, and the asset class emerges from its first test with the argument that the marks were conservative all along.
Mechanism
Lower policy rates reduce the cash-interest burden on borrowers at 5-6x reported leverage. Sponsors refinance the 2021-2023 vintages into a reopened syndicated market; managers reverse PIK elections; redemption demand falls because returns improve. The actor that releases the constraint is the FOMC, and the channel is the cost of the marginal refinancing.
Preconditions, early indicators and assumptions
- Calibration Basis
analogue
- Preconditions
- Policy rates fall materially from 3.50-3.75%
- The broadly syndicated loan market reopens to single-B issuers
- No idiosyncratic fraud event resetting sponsor appetite
- Early Indicators
- Indicator
Bad-PIK share falling below 4% of private credit loans in two consecutive quarterly assessments
- Source
S-07-20
- Threshold
below 4%
- Direction
below
- Would Be Visible By
2028-02
- Resolves
Confirms upside
- Indicator
Evergreen fund year-to-date returns recovering above 6%
- Source
S-07-21
- Threshold
6% YTD
- Direction
above
- Would Be Visible By
2027-07
- Resolves
Confirms the redemption-pressure leg
- Assumptions
- Text
Borrower distress is a cost-of-capital problem rather than an earnings problem
- Confidence
low
- Load Bearing
true
- Basis
T-07-02: roughly 40% of borrowers have negative free cash flow, up from 25% in 2021, which cuts against this
- If Wrong
Upside collapses into base; nothing moves to downside
- Text
Rates fall rather than rise
- Confidence
low
- Load Bearing
false
- Basis
Macro brief: FOMC held at 3.50-3.75% with three dissents in favour of a hike; the 2026-09-16 decision was unresolved at authoring
- If Wrong
The whole branch is unavailable
- Affected Industries
- 07
- 11
- Precedence Note
Requires measured PIK reduction, not manager commentary. Commentary alone leaves the outcome in base.
- Would Change Our Mind
Two consecutive FOMC decisions holding or hiking.
- What Businesses Should Do
Do not position for this branch on the strength of manager commentary; require the disclosed PIK share to fall in filings before extending duration.
The shadow rate was right
The 6% PIK-implied distress figure turns out to be the better estimate. Evergreen vehicles gate broadly and retail holders, now roughly 13% of US BDC participation, are locked in. NDFI facility drawdowns rise exactly when fund NAVs are falling, so the bank exposure crystallises at the worst moment, against capital requirements that were reduced by roughly 2.4% CET1 for the largest banks in the same cycle. The thesis is not wrong: private credit still intermediates mid-market debt. It delivers materially less.
Mechanism
The deteriorating parameter is realised recovery on 2021-2023 vintage loans. It shows up first in BDC 10-Q non-accruals, then in tender-offer proration on Schedule TO-I filings, then in the H.8 NDFI line as banks pull undrawn lines. The transmission channel into the banking system is the undrawn commitment, which is drawn precisely when the borrower is stressed.
Preconditions, early indicators and assumptions
- Calibration Basis
base_rate
- Preconditions
- Refinancing conditions do not ease
- At least three of the five largest non-traded BDC managers face over-cap redemptions in the same quarter
- NDFI lending growth turns negative
- Early Indicators
- Indicator
At least five distinct non-traded BDCs file Schedule TO-I documents in 2027 disclosing proration from over-cap tenders
- Source
S-07-20
- Threshold
five distinct registrants
- Direction
occurs
- Would Be Visible By
2028-01
- Resolves
Confirms the liquidity leg of downside
- Indicator
H.8 NDFI loan growth turning negative year over year
- Source
S-07-02
- Threshold
0% y/y
- Direction
below
- Would Be Visible By
2027-12
- Resolves
Confirms banks withdrawing; combined with the first indicator this is the downside signature
- Assumptions
- Text
Bank NDFI exposure is genuinely larger than the roughly USD 220bn supervisors can identify, closer to the commercial estimates above USD 500bn
- Confidence
medium
- Load Bearing
true
- Basis
FSB 2026-05-06 records both figures and attributes the gap to scope rather than error
- If Wrong
The bank transmission channel is smaller than assumed and the branch resolves closer to base
- Text
Gating converts a credit problem into a retail liquidity problem rather than containing it
- Confidence
medium
- Load Bearing
false
- Basis
Q4 2025: five BDCs funded tenders above the 5% cap; Blue Owl subsequently eliminated quarterly tenders
- If Wrong
The retail political dimension is smaller and the branch is narrower
- Affected Industries
- 07
- 11
- 18
- 02
- Precedence Note
Stopper: capital cost. Requires both legs - over-cap tenders at multiple managers and NDFI growth turning negative. One leg alone stays base.
- Would Change Our Mind
Over-cap tenders occurring while NDFI lending keeps growing above 15%, which would mean retail redemption without bank withdrawal.
- What Businesses Should Do
Holders of evergreen private credit should model a gate as the base case for exit timing, not as a tail. Banks should map undrawn NDFI commitments against the same funds' redemption queues.
Banks reprice the fund-finance channel
A different mechanism wins: rather than private credit funds financing mid-market borrowers with bank warehouse support, banks reclaim the senior layer directly and price it themselves, using the capital released by the 2026 rules. Subscription lines at roughly 200bp and CLO investment at roughly 350bp start to look expensive relative to originating the loan. At least two large banks build direct mid-market origination inside the charter, and the fund becomes the junior tranche rather than the whole capital structure.
Mechanism
The cost delta that lets the substitute win is regulatory capital: a CET1 reduction of roughly 2.4% for Category I and II banks and 3.0% for Category III and IV changes the relative return on direct lending versus fund financing. The incumbent asset that becomes worth less is the independent direct lender's origination franchise and the management fee on perpetual capital attached to it.
Preconditions, early indicators and assumptions
- Calibration Basis
analogue
- Preconditions
- The 2026-03-19 capital proposals are finalised broadly as proposed
- At least two large banks announce or disclose direct mid-market origination platforms
- Spreads on fund finance compress toward direct lending spreads
- Early Indicators
- Indicator
A joint final rule implementing the March 2026 capital proposals published in the Federal Register
- Source
S-07-06
- Threshold
final rule published
- Direction
occurs
- Would Be Visible By
2027-12
- Resolves
Precondition for disruption; without it this branch is unavailable
- Indicator
Two or more US G-SIBs disclosing a direct middle-market lending business as a named segment or product line
- Source
S-07-04
- Threshold
two firms
- Direction
occurs
- Would Be Visible By
2028-06
- Resolves
One company is a project; two is a sector
- Assumptions
- Text
Released capital is deployed into direct origination rather than into buybacks
- Confidence
low
- Load Bearing
true
- Basis
Q2 2026 already shows an 87% dividend payout ratio and Tier 1 ratios down 17bp, which points to distribution rather than deployment
- If Wrong
This branch stays a project and its mass returns to base
- Text
Banks can rebuild mid-market origination capability they largely exited after 2010
- Confidence
low
- Load Bearing
false
- Basis
No evidence in the corpus either way
- If Wrong
The substitution is slower and partial
- Affected Industries
- 07
- 11
- 02
- Precedence Note
Requires two named banks with disclosed direct origination. A single bank's announcement resolves to base.
- Would Change Our Mind
The capital package failing to be finalised by mid-2027, which removes the cost delta that makes this branch work.
- What Businesses Should Do
Private credit managers should treat bank capital rulemaking as a competitive event, not only as a systemic one.
- Calibration Note
Analogue: banks exited middle-market direct origination after the post-2010 capital build and marketplace lenders filled the gap; the 2013-16 marketplace-lending episode resolved with banks financing the disintermediators and then buying them. The comparability argument is that the driver in both cases is relative regulatory capital cost, not credit appetite.
A capital charge lands on the NDFI channel
Supervisors stop describing bank exposure to non-bank lenders and start pricing it. A specific capital treatment, disclosure requirement or reporting line is attached to fund finance, and the fastest-growing asset class in US banking stops growing at 22% because it becomes expensive to hold. The direction is restrictive, but a liberalising version also sits in this branch: agencies could instead confirm favourable treatment and accelerate the flow.
Mechanism
The instrument is a joint agency rule or a call-report schedule change. The FSB published a dedicated private-credit vulnerabilities report on 2026-05-06 and the Bank of England FPC flagged the interconnection in its July 2026 Financial Stability Report; in this sector supervisory reports have historically preceded reporting requirements by one to three years. The issuing bodies are the Federal Reserve, OCC and FDIC jointly, and the observable stage change is from proposed to final.
Preconditions, early indicators and assumptions
- Calibration Basis
base_rate
- Preconditions
- No credit event forcing emergency action, which would move this to failure
- Agency leadership continuity
- A proposal is issued rather than only a speech
- Early Indicators
- Indicator
A joint agency proposal attaching a specific capital treatment or reporting schedule to bank lending to non-depository financial institutions
- Source
S-07-06
- Threshold
NPRM published
- Direction
occurs
- Would Be Visible By
2028-06
- Resolves
Confirms regulatory
- Indicator
FSB or Bank of England FPC publishing a follow-up report recommending a specific supervisory requirement rather than monitoring
- Source
S-07-09
- Threshold
a named recommendation
- Direction
occurs
- Would Be Visible By
2027-12
- Resolves
Leading indicator for the US rulemaking
- Assumptions
- Text
The current deregulatory posture does not prevent the agencies from adding a targeted requirement in one category
- Confidence
medium
- Load Bearing
true
- Basis
The March 2026 package reduces capital overall while re-weighting specific exposures, so targeted increases are not inconsistent with the posture
- If Wrong
Mass moves to base with the exposure growing unmeasured
- Text
Supervisors can measure the exposure well enough to price it
- Confidence
low
- Load Bearing
false
- Basis
FSB can identify roughly USD 220bn against commercial estimates above USD 500bn
- If Wrong
Any rule is written against the wrong denominator
- Affected Industries
- 07
- 11
- Precedence Note
Regulatory outranks base when a proposal issues, even if lending volumes still look base-like.
- Would Change Our Mind
Two consecutive quarters with no supervisory publication on the topic from the FSB, the FPC or the US agencies.
- What Businesses Should Do
Model fund-finance pricing with a capital-charge scenario. Track the call-report schedule as closely as the rule text; a reporting line usually arrives before a capital charge.
Correlated software-credit repricing
The thesis breaks rather than slows. Roughly 25-26% of direct lending sits in software, underwritten against recurring revenue rather than earnings, at the moment AI pressure on seat-based pricing is being questioned by the same allocators. A correlated repricing of ARR-backed collateral hits many managers at once, transmits through bank facilities financing those lenders, and arrives while capital buffers are at their lowest point since the post-2008 reforms began. Roughly 70% covenant-lite issuance means the signal arrives at payment default, simultaneously.
Mechanism
The constraint that breaks the thesis is the assumption that software revenue is durable collateral. Outstanding SaaS loans grew from roughly USD 8bn in 2015 to over USD 500bn by end-2025, about 19% of total direct loans. If AI compresses seat counts or pricing, loan-to-value deteriorates across managers at the same time because they underwrote the same assumption. Who is left holding the asset: direct lending funds, their retail feeder vehicles, and the banks that financed them. The recovery period for a correlated ARR impairment is unknown because there is no prior episode.
Preconditions, early indicators and assumptions
- Calibration Basis
analogue
- Preconditions
- Observable compression in software seat counts or pricing at scale
- Non-accruals rising specifically in software-sector BDC holdings
- H.8 NDFI growth decelerating at the same time
- Early Indicators
- Indicator
Software-sector non-accruals rising in BDC 10-Q filings while the H.8 NDFI line decelerates in the same quarter
- Source
S-07-20
- Threshold
both legs in one quarter
- Direction
occurs
- Would Be Visible By
2028-06
- Resolves
This divergence is the failure signature - deterioration in the asset and withdrawal by the financier at once
- Indicator
A named large direct lender disclosing a write-down concentrated in software borrowers
- Source
S-07-04
- Threshold
disclosed write-down
- Direction
occurs
- Would Be Visible By
2028-06
- Resolves
Corroborates; a single disclosure alone stays in downside
- Assumptions
- Text
AI pressure on software pricing is large enough and fast enough to impair ARR-based collateral within the window
- Confidence
low
- Load Bearing
true
- Basis
T-07-14 records this causal link as asserted rather than demonstrated, with concentration figures from a single analytical source
- If Wrong
The branch is near-remote and its mass returns to base and downside
- Text
The software concentration figures of 25-26% and over USD 500bn are approximately right
- Confidence
low
- Load Bearing
false
- Basis
Single-source estimate, CAIA 2026-04-20, needing independent verification
- If Wrong
Measurement revision, and the entire branch is mis-sized
- Affected Industries
- 07
- 02
- 11
- 01
- Precedence Note
Failure outranks all other branches on occurrence. Requires the divergence - software non-accruals up while NDFI lending decelerates - not either leg alone.
- Would Change Our Mind
Independent verification that software is materially below 20% of direct lending portfolios.
- What Businesses Should Do
Anyone with exposure to both AI equity valuations and private credit should test whether they are the same bet held twice; the corpus records that nobody is currently joining those two analyses.
Additional scenario notes
- Review Required
true
- Review Reason
Financial-stability adjacent. The set concerns retail investor liquidity in gated vehicles and bank exposure to non-bank lenders, both flagged by the FSB and the Bank of England FPC.
- Probabilities Sum
1
- What Must Be True To Grow
- Capability: the asset class intermediates mid-market debt at scale - satisfied; FSB sizes it at USD 1.5-2tn at end-2024 with the US near USD 1tn
- Economics: realised returns clear the fee load at a 3.50-3.75% policy rate - contested; evergreen year-to-date returns fell to 1.6% through April 2026 from 7.4% in 2025
- Supply: continued borrower demand at 5-6x reported leverage - satisfied but with roughly 40% of borrowers at negative free cash flow
- Demand: allocator conviction - satisfied on fundraising, contested on redemptions
- Permission: no binding rule in any major jurisdiction as of 2026-09-15 - satisfied today, and the least durable of the six gates
- Capital: bank financing available at roughly 200-350bp - satisfied, and growing at 22.4% year over year
- What Could Stop It
- Capital cost: refinancing conditions failing to ease, converting PIK toggles into payment defaults
- Regulatory reversal: a capital charge or reporting requirement attached specifically to fund finance
- Measurement revision: bank exposure ranging from USD 220bn to over USD 500bn depending on whose data is used, which means the systemic size is not known to within a factor of two
- Demand deflation: retail redemption demand persisting through a year of weak returns and forcing broad gating
- Substitution: banks reclaiming senior origination with capital released by the 2026 rules
- Uncertain Assumptions
- That the PIK-implied distress rate leads payment default rather than recording losses already incurred - load-bearing for the base case, confidence medium
- That bank exposure is closer to the commercial estimates above USD 500bn than to the supervisory USD 220bn - confidence medium, and the gap is a measure of what supervisors cannot see
- That software is 25-26% of direct lending portfolios - single-source, confidence low, and the whole failure branch is sized off it
- That the rate path after the unresolved 2026-09-16 FOMC decision does not move by more than roughly 100bp - assumption inherited from the macro brief and dated
- Authored
2026-09-15