Five banks account for 96% of the Q2 2026 rise in uninsured deposits — and it isn't broad-based recovery

Direct answer

The FDIC's Q2 2026 Quarterly Banking Profile reported a $317.4 billion system-wide rise in estimated uninsured deposits. [C01] Within the 151 banks in OptimaYield's detailed Call Report layer with a verified uninsured-deposit ratio in both Q1 and Q2 2026, the matched increase was $291.75 billion — 91.9% of the FDIC's system-wide figure — and five banks (JPMorgan Chase, Citibank, Wells Fargo, Morgan Stanley Bank, and Bank of America) accounted for $280.77 billion of it: 88.5% of the entire system-wide increase. [C05, C07] JPMorgan Chase alone was responsible for a third of it, at a ratio that is now the highest point in its own decade of reported history. [C06]

Key findings

Why this population is not the FDIC's system-wide figure, and why that gap is informative

91.9% is a large majority, not the whole system. [C07] The 4,087 institutions outside this matched population collectively account for the remaining $25.65 billion of the FDIC's reported increase — a modest amount spread very thinly, consistent with the concentration finding above rather than contradicting it. This population is bounded by a real, disclosed requirement: an institution must report a verified (not abstained) uninsured-deposit ratio in both quarters. One large bank fails that requirement outright — see the limitations section.

Historical context: two very different post-2023 paths

A single-quarter comparison cannot show whether today's concentration is new. It isn't, for the two mid-size cohorts — and it's genuinely different for the largest banks.

The $10-50B and $50-250B cohorts both show a sharp decline in median uninsured-deposit concentration through the March 2023 SVB/Signature/First Republic episode, and neither has fully recovered to its earlier peak three years later. [C08] The $250B+ cohort shows no comparable pattern: it stayed inside a 44%-52% band across the entire 2016-2026 window and sits at 44.9% today, near the middle of its own range. [C09]

The banks driving this quarter's increase are not returning to a pre-2023 norm. For JPMorgan Chase and Citibank specifically, they are at or near the highest uninsured-deposit levels either bank has reported in a decade.

Named-institution trajectories

The five largest contributors, and the largest decliner, do not tell one uniform story:

• JPMorgan Chase Bank — ratio 62.53% (Q2 2026), the highest point in its full 2016-2026 series, above the prior high of 59.89% (Q4 2017). This is not a return to a prior level; it is a new one. [C06]
• Citibank — ratio 82.47%, the second-highest point in its own history, just below its all-time peak of 83.12% reached in Q3 2023, immediately after the SVB episode. Citibank's ratio rose through the 2023 stress window rather than falling — the opposite of the $10-250B cohorts above. [C13]
• Wells Fargo Bank — a real discontinuity, not gradual drift: the ratio fell from 53.0% to 42.0% in a single quarter (Q3→Q4 2023), and has only partially recovered to 50.6% by Q2 2026, still below its typical 2016-2023 range. [C11]
• Morgan Stanley Bank — nearly doubled within two quarters, from 19.9%-20.0% (Q3-Q4 2025) to 39.4% (Q1 2026) to 44.0% (Q2 2026). This is flagged as an unexplained discontinuity in this bank's own data, not assumed to be organic growth — no external cause was identified in this pass. [C12]
• Bank of America — ratio 45.4%, still below its 2016-2021 typical range (46%-50%), consistent with a slower, incomplete recovery.
• Bank of New York Mellon (largest decliner) — a $46.80 billion drop in dollar terms, but its ratio stayed near an all-time high (99.26%, versus a 99.41% peak the prior quarter). Its total deposits fell by almost the same dollar amount as its uninsured deposits, meaning this was almost entirely an uninsured-deposit outflow, not a shift toward insured funding. [C10]

Current-quarter cross-section: where the dollars sit

The population's aggregate ratio (47.59%) sits well above its median (39.45%) in Q2 2026, and the same gap existed in Q1 2026 (46.18% aggregate vs. 39.04% median). [C02, C03] That persistent gap is itself informative: an aggregate that runs meaningfully hotter than the median means the largest, most heavily-weighted institutions carry higher uninsured-deposit ratios than the typical bank in this population — exactly what the cohort history below confirms directly, rather than leaving it as an inference from one statistic.

Institutions with $1 trillion or more in assets — 4 of the 151 matched banks — contributed $257.0 billion of the $291.75 billion matched increase: 88.1% of the total from fewer than 3% of the institutions. [C14] This is consistent with, and explains, the five-bank concentration finding above: the increase is not evenly distributed across the $10B+ population; it sits almost entirely with the handful of banks large enough to be measured in trillions.

Why this matters

Deposit-mix and deposit-substitution are live themes in current bank-policy debate — Deloitte's 2026 banking outlook has modeled over $1 trillion at risk of migrating from bank deposits into stablecoins as the GENIUS Act's ecosystem matures, and BPI and ABA have both made public arguments about deposit substitution risk (see PUBLICATION-TOPIC-BACKLOG.md). This quarter's data does not test that claim directly — it measures uninsured deposits, not deposit levels overall — but it does show that whatever is happening to system-wide deposit funding right now, it is concentrated at a handful of the largest, most diversified institutions, not spread across the regional and community banks a deposit-substitution story would typically implicate first.

Methodology

Reuses OptimaYield's existing, already-validated uninsured_deposits_ratio metric contract (RC-O RCON5597 ÷ RC RCON2200) via the shared deriveValidatedMetrics function, unmodified. Current-quarter population: every institution in the detailed ($10B+) Call Report layer with a verified (non-abstained) ratio for the period. Matched-bank figures use only institutions verified in both Q1 and Q2 2026. Historical cohort figures reuse the already-computed percentile history in semantic.peer_statistics (42 quarters, 2016-03-31 through 2026-06-30) rather than recomputing distributions by hand. Named-institution trajectories query each bank's full 2016-2026 history directly. Live queries: scripts/analysis-uninsured-deposit-population.ts and scripts/analysis-uninsured-deposit-history.ts.

What changed and why

This draft was rewritten twice before this long-format version. The first pass (102 banks, all under $100B) wrongly blamed a supposed gap in the shared metric contract; that was checked directly against JPMorgan Chase's real data and found to be wrong — the bug was in this analysis's own eligibility filter. The second pass (152 banks, $320.02B) fixed that, but this long-format rewrite's own historical build surfaced a third issue: that version computed the ratio by naive division instead of reusing deriveValidatedMetrics, which let State Street Bank & Trust's out-of-domain (>100%) ratio into the population. The contract itself correctly abstains on any ratio outside [0,1]; naive division doesn't apply that check. Fixing it dropped the matched population from 152 to 151 and the matched total from $320.02B to $291.75B — a real, disclosed change, not a rounding difference.

Limitations and counter-evidence

• State Street Bank & Trust is excluded, not estimated around. [C15] Its own reported ratio exceeds 100% in several recent quarters, which the validated contract correctly treats as outside its reviewed domain. This is disclosed as a real population gap, not silently patched with an assumed value.
• 91.9%, not 100%. [C07] This population is a large majority of the FDIC's system-wide figure, not a full reconstruction of it — the remaining 8.1% is real and unmeasured here.
• Morgan Stanley Bank's discontinuity is unexplained. [C12] No external cause was identified in this pass; a future update should investigate rather than assume either an organic or a technical explanation.
• No cause is claimed for any bank's ratio. Every named figure above is an arithmetic fact about one balance-sheet line for one quarter, not a statement about that bank's funding stability, risk profile, or any supervisory concern.
• Not a frozen snapshot. Figures were pulled live from the production database rather than a versioned, hashed catalog artifact.

Glossary

• Schedule RC-O — the Call Report schedule covering deposit-insurance assessment base items, including the estimated-uninsured-deposits line.
• RCON5597 — the FFIEC field code for a bank's own estimate of its uninsured deposits.
• RCON2200 — the FFIEC field code for total deposits, reported on Schedule RC.
• Verified vs. abstained — OptimaYield's metric contracts report a ratio as "verified" only when every required input is present and the result falls within a reviewed domain (here, 0%-100%); otherwise they abstain rather than report a possibly-misleading number.
• Matched population — institutions with a verified value in both quarters being compared, so a dollar change can be attributed to an actual reporting institution rather than a population-composition shift.

Related on OptimaYield

• Q1 2026 covered-bank growth divergence — the prior quarter's companion distributional piece.
• Q2 2026 CRE concentration screen — this quarter's companion piece on commercial real estate concentration.
• The bank-profile pages for the named institutions above (not yet linked pending publication approval).

Official sources

• Federal Deposit Insurance Corporation, Quarterly Banking Profile — Second Quarter 2026, released September 1, 2026.
• FFIEC, Consolidated Reports of Condition and Income instructions, Schedule RC-O and Schedule RC.