Below-threshold & out-of-scope risk pools — what the FDIC bank lens misses

Built 2026-06-14. Sources: NCUA Quarterly Data Summaries + press (2025) & Ceto (CU failures); Fed FEDS note / NY Fed / Morgan Stanley / With Intelligence (private credit & BDCs); CRS R46489 + ICBA (ILCs); Farm Credit Funding / FCS results; Federal Reserve IBA release (foreign-bank branches); Wolf Street (FDIC problem-bank suppression). Full URLs at the bottom.

The FDIC bank aggregate is one slice of US credit. This maps the comparable HTM / CRE / duration / credit risk that sits outside or below that lens — a different regulator, a smaller size cut, or no public disclosure at all — so "banks look clean" isn't mistaken for "the system is clean." Companion to macro-fdic §5 and macro-bank-htm-marks.

1. Credit unions (NCUA — a ~$2.43T parallel system, different regulator)

• Size & capital: federally-insured CUs held ~$2.43T at Q4 2025 (+$126B / +5.4% YoY); aggregate net-worth ratio ~11.26% — well-capitalized in aggregate.
• Rising stress: delinquency 91bp (Q2) → 95bp (Q3) → 103bp (Q4), up ~5bp YoY; CRE/non-commercial-RE delinquency ~88bp (+10bp YoY).
• Failure cluster: 6 credit-union failures Apr–Aug 2025 — the largest cluster in ~7 years, ending a near-two-year no-liquidation stretch; all small (largest ~$58.5M), weak profitability + credit + governance.
• Why it's a blind spot: CUs file NCUA Form 5300 (not FDIC), carry their own HTM/AFS unrealized losses and CRE/member-business-lending concentration, CECL-covered — none of it in the FDIC aggregate the bubble thesis usually cites.

2. Private credit / NDFIs (the biggest uncovered reservoir)

The risk that migrated off bank balance sheets — and where AI-datacenter financing now lives.

• Size: US private credit ~$1.7T (some scopes $2.3–3T), projected ~$5T by 2029 (Morgan Stanley) — ~5× since 2009.
• Vehicles: BDCs ~$500B; private-wealth vehicles (BDC/interval/tender) >$400B; evergreen funds ~$644B (Jun 2025, +28% H1) — manager-set NAVs, the third self-marked number (macro-private-credit-marks).
• Bank linkage: banks lend to NDFIs ~$1.97T (Fed) — bank risk didn't vanish, it moved one step out and turned opaque (macro-cre-privatecredit).

3. Industrial loan companies (FDIC-insured, Fed-unsupervised)

• Size: total ILC assets ~$247.7B (2025), up from ~$25.1B (1997); ~23 active charters, 6 over $10B, largest ~$116.3B.
• The loophole: ILCs are FDIC-insured yet their commercial parents escape Fed consolidated (BHC) supervision — the basis of the ICBA "close the loophole" campaign and of fintech (Square et al.) and retail/auto interest in the charter.

4. Farm Credit System (a federal GSE off the FDIC map)

• A GSE lending complex (funded via the Federal Farm Credit Banks Funding Corp) holding roughly ~$188B farm real-estate + ~$81B non-RE farm loans at YE2024 — a ~$400B+ asset GSE.
• Exposure: concentrated in farmland values and ag commodity cycles; a farmland correction or trade-war ag shock is its tail — distinct from the bank/CRE map.

5. Foreign-bank US branches (a ~$3.0-3.4T dollar-funding node)

• US branches/agencies of foreign banks hold ~$3.0-3.4T in US assets (Fed IBA release), heavily in wholesale dollar funding and repo — a transmission node for offshore-dollar stress, outside the FDIC domestic-bank set.

6. The suppressed problem-bank total (hidden even within FDIC scope)

• Since Q4 2024 — the first time since 1990 — the FDIC publishes only the problem-bank count (~54), not the asset total, citing the risk of a "disorderly run." The distribution is deliberately below disclosure; the per-bank reconstruction in macro-bank-htm-marks partly recovers it.

The total uncovered stack — ~$33–37T outside the FDIC bank lens

Each pool below now has its own deep-dig block. Most share the SAME underlying collateral — Treasuries, agency MBS, private-credit NAVs — so the load-bearing point is correlation, not a clean sum.

Honest tally: the cleanly-sized, low-overlap pools alone (MMFs, agency/GSE complex, munis, credit unions, private credit, foreign-bank branches, family offices, ILCs, Farm Credit) sum to ~$33–37T outside the FDIC domestic-bank deposit lens — itself comparable to the ~$24T FDIC-insured bank asset base. This is an order-of-magnitude floor, not a precise sum: the pools overlap heavily (MMFs/hedge funds/pensions all hold the same Treasuries; mREITs/banks/Fed hold the same agency MBS; pensions hold the private credit), so the real systemic point is the shared collateral, not the dollar total.

Synthesis — the FDIC bank number is a floor, not the system

The same ~91% common factor (macro-cross-sectional-analysis) that prices the bank HTM hole also prices credit-union securities, private-credit NAVs, pension alts, mortgage-REIT books, MMF/hedge-fund Treasuries, and farmland — so "the banks are fine" understates a correlated pool that mostly shares the same Treasuries, agency MBS and self-marked NAVs. Credit unions and private credit show rising stress now; the rest is latent until a rate or credit shock prices them together.

What is NOT asserted

• No claim any pool is insolvent system-wide — CUs and the FCS are well-capitalized in aggregate; the point is they're uncovered by the FDIC lens and carry correlated risk.
• Private-credit size is scope-dependent ($1.7T–$3T) — a range, not false precision.
• The ILC "loophole" is a documented policy debate, not an allegation against any ILC.
• FBO/FCS exact totals live in their primary releases; figures here are as-reported approximations.
• Overlay edges are excluded from the SCC / Z3 / TLA+ proofs.

Sources: NCUA — Q4 2025 system performance, Ceto — credit-union failures hit a 7-year high in 2025, Fed FEDS — bank lending to private credit, Morgan Stanley — private credit to ~$5T by 2029, CRS R46489 — Industrial Loan Companies, Farm Credit Funding — earnings/investor materials, Federal Reserve — Assets & Liabilities of US Branches & Agencies of Foreign Banks (IBA), Wolf Street — FDIC ends disclosing problem-bank assets.

← Research index · structured data: macro-uncovered-risk-pools.json · macro-uncovered-risk-pools.md