Foreign-bank US branches (~$3T+) — the dollar-funding node, the yen carry, and the IHC gap

Built 2026-06-15. Structured data + edges: macro-foreign-bank-branches.json. Deep companion to macro-uncovered-risk-pools §5. (WebFetch unavailable; aggregate totals are primary Fed series (solid); per-bank-by-country and the line-item split are weak — the Fed publishes aggregates, not a league table. The "$2T+" premise is outdated — actual ~$3.0-3.4T.)

US branches and agencies of foreign banks are a ~$3.0-3.4T book — not FDIC-insured, funded through wholesale dollar markets and intra-firm "net due," ~87% concentrated in New York. The sharpest live channel is Japanese: Japanese banks hold the largest US-resident claims of any banking system and fund US assets (Treasuries, CLOs, leveraged loans) synthetically via yen/FX swaps — so BoJ normalization drove the Aug-5-2024 carry unwind and Norinchukin's ~$63B forced bond sale. And the Dodd-Frank IHC ring-fence EXCLUDES branches — leaving the ~$3T book outside the US consolidated capital/liquidity regime.

1. Size & funding

• ~$3.04T in US assets (Fed IBA, Sep 30 2024), NY ~87% (~$2.64T); the broader "foreign-related institutions" H.8 series ~$3.39T (Mar 2026). The node has grown well past the commonly-cited "$2T+."
• Funding fragility: branches are not FDIC-insured and have no retail base — they fund via large-time/brokered deposits, fed funds/repo, and "net due to related institutions" (intra-firm parent funding). Confidence-sensitive and dollar-denominated. (Exact cash/securities/loans/net-due split is weak — needs H.8 Table 10.)

2. By country

• Japanese banks dominate — the largest US-resident claims of any banking system (BIS): MUFG, SMBC, Mizuho (MUFG sold Union Bank's CA retail to U.S. Bancorp in 2022, shifting toward wholesale/branch). Canadian (TD, RBC, BMO, Scotiabank) and European (Barclays, Deutsche, BNP, Credit Agricole, Santander) operate more via IHC-held subsidiaries; Chinese (ICBC, BoC, CCB) have a smaller NY footprint. (Per-bank dollar sizes weak — Fed aggregates only.)

3. The yen-carry channel (the live risk)

• BoJ path: out of negative rates Mar 2024 (−0.1%→0) → 0.25% (Jul 31 2024) → 0.5% (mid-2025) → ~0.75% (Dec 19 2025, a 30-year high; 10y JGB past 2%). The Jul-2024 hike triggered the Aug-5-2024 unwind: Nikkei −12.4% (worst day since 1987), yen surged, yen-funded leveraged longs force-unwound.
• Norinchukin: announced (Jun 2024) it would sell ~¥10T (~$63B) of US Treasuries + EU sovereigns (~⅙ of its global portfolio) to stem unrealized losses (~¥2.19T at Mar 2024); posted a ~$12.6B FY2024 net loss and replaced its CEO. The concrete case of a Japanese institution force-selling Treasuries on the rate turn.
• The channel: Japanese banks run USD assets > USD liabilities, relying on synthetic dollar funding (FX swaps / the cross-currency basis) — off-balance-sheet forward dollar obligations not in standard debt stats (BIS/IMF flag this). Rising yen rates + a weaker basis raise their USD funding cost and pressure Treasury/CLO demand — a transmission line from BoJ policy to the US Treasury market.

4. Repo, Fed backstop & the IHC gap

• IOR arbitrage: FBOs account for the bulk of IORB arbitrage — not FDIC-insured, looser leverage rules, so they borrow cheaply (esp. from FHLBs) and park reserves at the Fed; marginal repo intermediaries.
• Fed backstop: dollar swap lines peaked ~$449B (May 2020; 82% BoJ + ECB); the FIMA repo facility (2020, standing since 2021) lets foreign central banks repo Treasuries for dollars — backstops for this same node.
• The IHC gap: Reg YY requires an FBO with ≥$50B in US non-branch assets to form a US Intermediate Holding Company (Basel III, CCAR, liquidity) — but the threshold counts non-branch assets only, so branches and agencies are EXCLUDED from the IHC and its consolidated regime. The ~$3T branch book sits outside the US ring-fence.

Synthesis

A ~$3T, NY-concentrated, FDIC-uninsured, wholesale-funded node outside the Dodd-Frank IHC ring-fence. The sharpest live risk is the yen-carry/synthetic-dollar channel: Japanese banks (the largest US claimants) fund Treasuries/CLOs via FX swaps, so BoJ normalization transmits straight to the US Treasury market (Aug-2024 unwind; Norinchukin's ~$63B sale). The offshore-dollar transmission line in the uncovered stack — and the reason the Fed's swap-line/FIMA backstops exist.

What is NOT asserted

• No claim of current FBO distress — the point is the size (~$3T+, not $2T), wholesale-funding fragility, the yen-carry transmission, and the IHC exclusion of branches.
• Per-bank by-country sizes and the line-item split are weak (Fed aggregates only).
• Overlay edges are excluded from the proofs.

Sources: Fed — Assets & Liabilities of US Branches & Agencies of Foreign Banks (IBA); FRED — foreign-related institutions assets; BIS CGFS — Japanese-bank US claims / FX-swap funding; CNBC — BoJ Dec-2025 hike; CNBC — Aug-5-2024 Asia markets; Bloomberg — Norinchukin $63B bond sale; NBER — 2020 swap-line usage; Fed — FIMA repo facility; Cornell LII — 12 CFR 252.153 (IHC).

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