US bank hierarchical exposure — the vulnerable subset, HTM vs AFS, and the stablecoin squeeze

Built 2026-06-07 from FDIC per-institution call reports (2026Q1) via models/graph/bank_exposure.py → data/bank_exposure.json (top 200 banks). Aggregate series from macro-fdic.json. Stablecoin/policy via GENIUS Act + ICBA.

1. The HTM-vs-AFS divergence (your thesis — confirmed)

Two facts from the FDIC aggregate series, both as you suspected:

• HTM losses are narrowing slower than AFS — and both worsened QoQ into 2026Q1 (long rates backed up on the AI-debt issuance wave + fiscal supply; see macro-history-dereg-manipulation.json).
• Why it matters: AFS losses flow through AOCI and (for large banks) into regulatory capital; HTM losses do NOT — they're the un-marked hole. A bank can look well-capitalized while its HTM book is deeply underwater. This is the exact SVB blind spot. HTM is also stickier (selling any taints the whole portfolio's held-to-maturity intent), so it heals only as securities roll to par — slowly.

2. The hierarchy (tiered by the data, not by reputation)

From data/bank_exposure.json (2026Q1). The vulnerability is not uniform — it concentrates by tier:

• Tier 1 — G-SIB/mega (JPM, BofA, Citi, Wells, GS; ~$11.2T): mostly resilient, with one glaring exception. BofA HTM −$81B = −34% of equity; Wells HTM −$33B = −19%, Wells investor-CRE/T1 52%. JPM is the fortress (HTM −5%, CRE 20%) — but uninsured 65%. In a stress, deposits flee toward these names (2023 SVB→JPM), which is the structural "squeeze": the mega-banks gain funding precisely when mid-tier banks lose it.
• Tier 2 — super-regionals (~$7.7T): mixed. US Bank flagged (HTM −14%, uninsured 51%); custody banks BNY Mellon/State Street carry 90%+ uninsured (institutional cash) + HTM −10/−15%.
• Tier 3 — large regionals (~$1.3T): the CRE-concentration cohort. Western Alliance 175%, Webster 175%, East West 168%, Old National 178%, Zions 140% investor-CRE/Tier-1. These are the banks most exposed to the CRE + housing unwind in major cities (macro-cre-privatecredit.md: office CMBS delinquency record 11.76%).
• Tier 5 — community (CRE-heaviest by ratio): S&T 266%, Tri-Counties 243%, EagleBank 238%, Amerant 215% — small in dollars, but these are where the >300% supervisory flag bites first.

The biggest "hidden" HTM holes (un-marked, as % of equity)

USAA −50% · Charles Schwab Bank −43% · Bank of Hawaii −35% · BofA −34% (−$81B) · Morgan Stanley PBNA −23% · Wells −19%. (Schwab and USAA are deposit-franchise-specific; BofA is the systemically large one.)

Flagged vulnerable subset (≥2 of CRE>300% / secs-loss>15%eq / uninsured>50%)

Frost Bank (TX), First Hawaiian, City NB of Florida, Banco Popular (PR), Citizens Business Bank (CA), US Bank, Washington Trust, Farmers & Merchants (CA) — the securities-loss + run-risk overlap. The CRE-heavy regionals (Western Alliance/Zions/Webster) are a second watch-list: high CRE concentration but (so far) smaller securities holes.

3. The squeeze: mega vs mid, and the stablecoin/tokenized-deposit shift

• Mega-banks can squeeze/absorb mid-tier banks structurally: in any run, uninsured deposits migrate to the perceived-safe G-SIBs; mid-tier banks with CRE concentration + HTM holes + high uninsured ratios are the fragile node (the SVB/First Republic 2023 pattern, now mapped per-bank above).
• GENIUS Act (Jul 2025) — the new pressure: payment stablecoins must be 100% reserve-backed (Treasuries/cash/insured deposits) and cannot pay yield; tokenized deposits (bank-issued, yield-bearing) are preserved. ICBA lobbied specifically against community-bank disintermediation. Mechanism: full-reserve stablecoins + Treasuries pull deposits out of fractional-reserve banks → shrinks the cheap-deposit base that funds CRE/housing → tightens funding on exactly the CRE-exposed mid/small banks while they're already absorbing CRE and housing unwinds. Big banks (and the stablecoin issuers, often partnered with them) can ride the shift; thin-capital regionals/neobanks cannot.
• Neobanks/fintech banks (e.g. SoFi in the data: CRE 0%, HTM 0%, uninsured 8%) have the opposite risk profile — little CRE/securities risk but funding-model and 1:1-backing-margin sensitivity. A "tech bank" reliant on deposit spread is squeezed by the no-yield-stablecoin / tokenized-deposit shift compressing its leverage economics.

4. Community-bank politics (user-supplied items, graded)

• ICBA (Independent Community Bankers of America) is the documented small-bank lobby that shaped GENIUS Act guardrails — verified.
• The user references a specific community-bank-org leader with a family history in small banking whose own bank is currently vulnerable, and a neobank ("Bank on Roger") being pulled into this network. I could not verify either specific identity from public sources and will not fabricate names; the structural dynamic they describe (a vulnerable mid/community bank's leadership opposing the stablecoin shift while larger banks benefit) is consistent with the data and the ICBA disintermediation fight — graded plausible-structural, specific-identity UNVERIFIED.

5. Limits (honesty)

True bilateral interbank exposures (who owes whom) are not publicly granular — only aggregates (Fed funds, FHLB advances, the H.8 NDFI line ~$1.97T, brokered deposits). This hierarchy is built from each bank's own balance-sheet risk (CRE, HTM/AFS, uninsured), which is the right proxy for run/CRE-unwind vulnerability, but it is not a contagion-netting map. The TLA+ cascade (models/tla) models contagion at the system-tier level instead.

← Research index · structured data: macro-bank-hierarchy.json · macro-bank-hierarchy.md