Citigroup Inc. (NYSE: C) — Deep Dive

Reference price: $137.72 (close, 4 September 2026)

Latest reported period: Q2 2026 (quarter ended 30 June 2026)

This document is created from the knowledge cards of Citigroup Inc.

1. Metrics Defined

Every term below appears repeatedly in this document; each is defined once here.

The income statement, in one bridge

A bank's P&L has four lines that matter. Citigroup's FY2025 reconciles exactly:

The same bridge for H1 2026: revenue ~$49.5B, expenses ~$28.7B (58% efficiency ratio), PPNR ~$20.8B, provisions ~$5.1B, pre-tax ~$15.7B, net income ~$11.6B, to common ~$11.0B.

The metrics

Efficiency ratio = operating expenses ÷ revenue. Lower is better. Citi ran 64.7% in FY2025 and 57.4% in Q2 2026. A large-bank peer range is roughly 52% (JPMorgan) to 68% (Morgan Stanley). Every point of improvement on $85B of revenue is roughly $850M of pre-tax income.

TBVPS — tangible book value per share. Common shareholders' equity, minus goodwill and other intangible assets, divided by shares outstanding. Goodwill and intangibles are stripped out because they cannot absorb a loss — a regulator will not let a bank pay claims with goodwill. TBVPS is the closest thing a bank has to a liquidation floor and the standard denominator for bank valuation. Citi: $100.89 at 30 June 2026.

ROTCE — return on tangible common equity. Net income to common ÷ average tangible common equity. This is the profitability metric for a bank, and the number every part of this report turns on. Read it as: for every dollar of loss-absorbing equity the shareholder has in the business, how many cents come back per year.

Citi's FY2025 bridge from the more familiar ROE:

• Average common equity: ~$194B → ROE (common) = 13.2 ÷ 194 = 6.8%

• Less average goodwill ($19.8B) and intangibles ($3.6B) → average TCE = $170.6B

• ROTCE = 13.2 ÷ 170.6 = 7.7%

The ~90bp uplift is purely the smaller denominator. It is not extra earnings.

Three traps in ROTCE that this report labels explicitly every time:

1. Reported vs. adjusted. FY2025 reported ROTCE was 7.7%. Excluding Russia-related charges and the Banamex goodwill write-down, adjusted was 8.8%. Both are correct; they are never mixed in this document without a label.

2. Segment vs. consolidated. Segment ROTCE uses management-allocated tangible capital. Consolidated ROTCE uses GAAP average TCE. The two denominators are built differently and do not sum. This is why all five segments can show 11–31% while the firm shows 7.7% — see §5.

3. Quarterly vs. annual. A single quarter's ROTCE annualizes a seasonal result. Citi's Markets business is materially stronger in H1 than H2, which makes half-year ROTCE structurally flattering. See §6.

NCL rate — net credit loss rate. Loans charged off during the period, annualized, as a percentage of average loans. Citi firm-wide: 9.1B of charge-offs in FY2025. Card books run 4–6%; corporate books run 0.1%.

ACLL — allowance for credit losses on loans. The reserve held on the balance sheet against future losses, expressed as a percentage of loans ("coverage"). Citi: $20.0B, or 2.54% of loans, at Q2 2026. Rising coverage depresses current earnings and builds a cushion; falling coverage does the reverse. Under CECL accounting, the reserve is set against lifetime expected losses, so it moves with the economic forecast, not just with actual defaults.

CET1 ratio — common equity tier 1. The binding regulatory capital constraint. CET1 capital ÷ risk-weighted assets (RWA — assets scaled by their regulatory riskiness; a Treasury bond carries near-zero weight, an unsecured credit card balance carries a high one). Citi at 30 June 2026: $157.5B ÷ $1,232B = 12.78%.

Citi's required minimum is built from three pieces:

Headroom above the requirement is what funds buybacks. At Q2 2026 that headroom was 118bps, or roughly $14.5B of deployable capital.

P/TBV and why it is the right multiple. For a bank in steady state, the Gordon growth relationship gives:

P/TBV = (ROTCE − g) ÷ (CoE − g)

where g is long-run growth and CoE is cost of equity. The intuition is simple: a bank that earns exactly its cost of equity is worth exactly 1.0× tangible book. Everything above 1.0× is capitalised excess return. With CoE = 10.5% and g = 3.5%, each additional point of sustainable ROTCE is worth roughly 0.14× of tangible book — about $14 per Citi share. This single relationship does most of the work in §8.

Sanity check that the framework is calibrated: Bank of America earns ~14.2% ROTCE and trades around 1.4–1.8× TBV. Solving backward gives an implied CoE of ~10.2% — consistent with the 10.5% used here.

2. Business Model and Unit Economics

Citigroup operates in nearly 160 countries and jurisdictions across five reportable segments. Effective 1 January 2026, Retail Banking (~$90B of deposits) moved from U.S. Personal Banking into Wealth, and the remaining consumer business — Branded Cards plus Retail Services — was renamed U.S. Consumer Cards (USCC). Year-over-year comparisons spanning that boundary require recast prior-period data; this report labels every affected figure.

Segment master table — FY2025 (old structure)

The five segments earned $19.9B; the firm reported $14.3B. The $5.6B difference is the All Other / Corporate drag — see §5, where it is the single most important number in the report.

Multi-year segment trajectory

Two things stand out and are developed in §7:

• Services ROTCE has been flat at roughly 29% for three years. The Q2 2026 print of 30.9% is at the top of that range, not the start of a new trend.

• Markets swung from ~8% (FY2024) to 17–19% (2026). A 10-point swing in two years is cyclical, not structural.

How each segment actually makes money

Services (Treasury & Trade Solutions + Securities Services) — $21.3B revenue, 28.6% ROTCE. The highest-return business and the reason the franchise exists. TTS holds operating deposits for multinationals and financial institutions — balances that sit in the account because payroll, supplier payments, and FX settlement run through it. Those deposits are cheap and stable; management describes deposit betas as "quite stable." The NII engine is the spread earned on them.

Roughly 30% of Services revenue is fees rather than spread — cross-border transaction fees and custody fees — which gives the segment a floor that does not depend on the rate cycle.

The critical caveat: Services deposits grew 19% year-over-year in Q2 2026. Management's own normalized guidance is a return to the 2022–2025 CAGR of approximately 3%. Two-thirds of the recent deposit growth is therefore not expected to recur. Since Services NII scales directly with deposit balances, this is a cyclical tailwind being reported inside a structurally attractive business — the two need to be separated, and §7 does so.

Markets — $22.0B revenue, 11.6% ROTCE (FY2025), 17–19% (2026). The mix has shifted decisively: NII is now 46% of segment revenue, up from a historically much smaller share. Management is explicit that this is not a business-model change: Citi assesses Markets on total NII + NIR because trading book hedges create offsetting swings between the two lines, and the NII surge reflects a larger prime brokerage and securities financing book, which is funded on balance sheet.

Prime brokerage balances grew ~50% in FY2025 and a further ~60% year-over-year in Q2 2026 — roughly 2.4× in two years. This is real share gain from hedge fund clients with multi-year switching costs once operational infrastructure is embedded. It is also capital-intensive, balance-sheet-heavy, and the part of a markets franchise where counterparty risk concentrates fastest. This report treats fast prime-brokerage growth as a two-sided fact, not a one-sided positive.

Banking — $8.2B revenue, 11.3% ROTCE. From a $31M net loss in FY2023 to $2.3B of net income in FY2025. Note the composition: the $8.2B includes corporate lending, not just advisory. Pure investment banking fees are approximately $4.6B — roughly half of JPMorgan's $9.7B. Q2 2026 IB revenue was $1.5B (+44%), with ECM +92% and DCM +65%.

A structural note: when Services, Markets, or IB products are sold to Corporate Lending clients, a revenue credit flows to Banking. This is deliberate incentive design for the "One Citi" cross-sell model, but it means Banking's reported revenue is not a clean read on standalone banking economics.

Wealth — $8.6B revenue, 12.1% ROTCE. The clearest structural improvement in the firm: efficiency ratio from 92% (FY2023) to 76% (FY2025); ROTCE from ~6% to 14.4% in Q2 2026, against a 15–20% medium-term target. Q2 2026 revenue $3.18B (+13% on a recast basis), client balances $1.35T, AUM $727B. The January 2026 transfer of Retail Banking into Wealth adds ~$90B of deposits and a large retail cross-sell surface; referrals from retail banking into Citigold rose 23% in Q2 2026.

U.S. Consumer Cards (USPB → USCC) — $21.0B revenue, 13.2% ROTCE. The economics here are frequently misread. USPB's FY2025 revenue of $21.0B was almost entirely net interest income ($22.5B of NII — the spread on roughly $220B of average card loans), because gross interchange of $10.5B was more than offset by $12.1B of card rewards and co-brand partner payments, leaving net fee revenue at approximately −$1.6B.

This is not a Citi-specific weakness. It is how every large co-brand card issuer works: interchange is passed through to the partner and the cardholder, and the issuer's profit is the loan spread minus credit cost. The correct framing is that USCC is a secured-by-nothing consumer lending book wearing a payments coat — its P&L lever is the net credit loss rate, not fee income.

The 90+ delinquency rate is the leading indicator, and it turned down in Q2 2026. Management guides FY2026 USCC NCLs to 4.0–4.5%. Reserve coverage stands at 7.6% of funded loans, and approximately 86% of balances are extended to borrowers with FICO scores of 660 or higher.

Firm-wide revenue mix and funding

NII has risen 41% over four years while NIR fell 13%. The mix shift toward spread income reflects Services deposit growth, card portfolio expansion, the securities portfolio roll-up, and — importantly — the Markets financing book.

Funding. Deposits are the primary funding source, at approximately $1.35–1.38 trillion in average balances. Average deposit cost fell to 3.01% in FY2025 from 3.61% in FY2024; total cost of interest-bearing liabilities was 4.19%; net interest margin was 2.47%.

Services holds approximately 65% of total average deposits ($893B of $1.382T in Q3 2025), which makes TTS operating balances not merely the firm's most profitable business but its cheapest and most durable funding source. This is the single most important structural fact about Citigroup's balance sheet.

Rate sensitivity (as of 31 December 2025). The USD book is close to rate-neutral; the non-USD book, spread across 65+ currencies, is structurally asset-sensitive.

Note that these disclosures measure the banking book. They do not fully capture the sensitivity of Markets financing NII, which has become a meaningful contributor and would compress in a falling-rate environment.

3. Moat and Competitive Position

What the moat actually is

Citigroup's durable competitive advantage is Treasury and Trade Solutions: a proprietary cross-border payment and liquidity infrastructure operating in 90+ countries, which management describes as "a franchise that is very hard to replicate."

The evidence for durability, ranked by strength:

Strong — the geographic footprint is genuinely hard to replicate. Services generates 70% of its revenue outside North America ($14.8B of $21.3B in FY2025), spread across six international clusters:

For contrast, JPMorgan's entire firm-wide international revenue is approximately $42.8B — 23.5% of $182.4B. Citi's Services segment alone generates a higher proportion of international revenue than JPMorgan generates across its whole franchise. Building a licensed, staffed, locally-compliant clearing presence in 90+ countries is a decades-long, capital- and regulator-intensive undertaking that no U.S. peer is attempting. Per TABInsights' 2026 CIW ranking, transaction banking constitutes 60.4% of Citi's total CIW revenue — the highest proportion among all 100 ranked global institutions.

Moderate — operating deposits are sticky, but the evidence is management-narrated. The switching-cost argument is that TTS balances are embedded in client payroll, supplier payment, and FX workflows. That is plausible and consistent with a 3% normalized deposit CAGR and with the fee layer being 30%+ of segment revenue. But this report cannot verify it independently: client retention rates, revenue-per-client cohort data, and pricing/spread evidence are not disclosed. The moat claim rests on a structural argument plus management assertion, not on measured customer behaviour. Readers should weight it accordingly.

Weak — the recent numbers overstate it. Services deposits grew 19% in Q2 2026 against a 3% normalized pace. Whatever is driving that (global liquidity conditions, client cash build, mandate wins) is largely not the moat, and it is expected to reverse.

Not addressed in available sources — disruption risk. Stablecoins and tokenized deposits attack precisely the cross-border settlement float and FX spread that TTS monetizes. Citi Token Services (blockchain-based 24/7 cross-border settlement, deployed in a subset of markets) is Citi's response and is disclosed as a positive. No source in this analysis quantifies the threat or Citi's exposure to it. This is the largest unexamined risk to the moat thesis.

Peer comparison — FY2025

All figures on a FY2025 basis. BAC's efficiency ratio is derived (revenue $113.7B less implied expenses of ~$69.7B). Wells Fargo TBVPS is an estimate and is flagged in Appendix B.

The honest read of this table: Citigroup is the least profitable and least efficient of the six on every measure. The gap to JPMorgan (~20% ROTCE) and Bank of America (14.2%) is not a rounding difference — it is 6 to 12 points of ROTCE, worth roughly $10–20B of annual earnings on Citi's capital base. The bull case for Citi is not that it has a better franchise; it is that the gap is closing and that the market has not fully paid for the closure. §8 tests whether that is still true at $137.72.

Where Citi genuinely leads or is competitive

• Transaction banking scale. Q2 2026 Services revenue of $6.38B was the highest quarterly figure in the segment's history, on $1.017 trillion of average deposits — the first time that threshold has been crossed. Citi Payments Express is deployed in 22 markets; USD clearing processed 46.3 million transactions in Q2 2026.

• Prime brokerage momentum. Balances +60% year-over-year in Q2 2026, following ~50% in FY2025, indicating genuine mandate wins from hedge fund clients.

• Wealth improvement rate. ROTCE from ~6% to 14.4% in three years is the fastest improvement of any segment, though from a low base and against a peer set operating at a different scale entirely (BofA Global Wealth: $6.9B quarterly revenue, $4.9T client balances; JPM AWM: $7.1T).

• Investment banking share gains. Wallet share gains of approximately 30bp in FY2025 (following ~50bp in FY2024), with strength in Sponsors (+180bp), Leveraged Finance (+100bp), and M&A advisory (+90bp). Citi participated in 8 of the top 10 ECM transactions globally in Q2 2026.

Where the gap persists

• IB scale. $4.6B of pure IB fees against JPMorgan's $9.7B. This reflects JPM's dominant North American corporate relationship network — a coverage gap Citi has publicly acknowledged it is working to close, and one that takes a decade to close if it closes at all.

• Wealth scale. Citi Wealth's $727B of AUM against BofA's $2.3T and JPM's $4.8T. Citi operates at roughly a third of the smaller peer's scale.

• Structural efficiency. 64.7% versus JPMorgan's ~53%. Part of this is transformation spend that will roll off; part of it is the cost of running a 160-country franchise. Distinguishing the two is the central question in §7.

4. Management, Capital Allocation, and Culture

The transformation, and the credibility test

CEO Jane Fraser's tenure has been defined by a regulatory remediation program running against the October 2020 Federal Reserve and OCC consent orders. The single most-discussed leadership decision in the covered period was the January 2025 downward revision of the 2026 ROTCE target to 10–11%, disclosed on the Q4 2024 earnings call following 2024 civil money penalties (OCC $75M; Federal Reserve $61M):

"However, when we take the required investments into account, we now expect our 2026 RoTCE to be between 10% and 11%. The 2026 RoTCE is a waypoint. It is not a destination."

Fraser framed this as choosing investment durability over short-term target maintenance. The Board's Compensation Committee, in setting her FY2025 pay at $42M, described these as "bold choices," and noted her stock price performance led all industry peers over one- and three-year periods through 2025.

By H1 2026 the firm was running at 13.1% ROTCE — above the full-year target — and management explicitly declined to raise full-year guidance, directing the outperformance into front-office hiring, AI infrastructure, cards marketing, and banking headcount:

"It is 100% on the offense... we will lean in, bringing forward investments and other actions that will create value for our shareholders over the medium term."

How to read this. The pattern is consistent — invest through the cycle, guide conservatively, do not chase a waypoint. Two readings are available and the evidence does not decide between them: it is either disciplined long-term capital allocation, or it is a management team preserving optionality on a target it revised downward once already. What is missing from every source in this analysis is the base rate: no independent assessment of how Citigroup management teams have performed against publicly stated targets over the last two decades. For a firm with Citi's history, that absence matters.

Transformation status

The remaining open item is precisely stated: the 2020 FRB Consent Order and the underlying 2020 OCC Consent Order remain in effect. The December 2025 action terminated the July 2024 amendment to the OCC order — the penalty overlay — not the order itself. Full closure timing is at regulators' discretion:

"In terms of the timing of the removal of the consent orders, that is fully at the discretion of our regulators, in terms of reviewing the work that we have done, as we hand it over, and then going through their closure process, and that takes time. So completing the work is just the beginning of the end."

While the OCC order is in force, it requires prior OCC approval for any significant acquisition and preserves the OCC's right to impose business restrictions, limit Citibank dividend declarations, and require changes to directors and senior executives.

Operating model and efficiency

• Structure: Q1 2024 organizational simplification eliminated the ICG/PBWM layers and created five reportable segments, consolidating four regions to two. Q1 2026 moved Retail Banking to Wealth and merged Branded Cards + Retail Services into USCC.

• Headcount: reduced from ~240,000 (the 2022 post-pandemic peak within Fraser's tenure — not the firm's all-time high) to ~219,000 by Q2 2026, with ~$800M of severance incurred in H1 2026.

• Applications retired: 714 in FY2024 plus 548 in FY2025 — 9% of all applications.

• Efficiency ratio: 71.7% (FY2023) → 66.4% (FY2024) → 64.7% reported / ~63% adjusted (FY2025) → 58% (Q1 2026) → 57.4% (Q2 2026). Full-year 2026 target is "around 60%" — revised from prior "below 60%" language to allow investment flexibility. The two formulations are not equivalent and the revision was deliberate.

• AI: ~180,000 colleagues in 83 countries had access to proprietary AI tools in Q3 2025; 21 million tool interactions by Q4 2025 across 84 countries; adoption from ~70% (Q4 2025) to 90%+ (Q2 2026). Over 100 end-to-end processes mapped for automation. AI adoption is one of the few available proxies for organizational adaptability during a transformation, but it is an input metric, not an outcome.

Capital allocation

Management has articulated a consistent three-priority hierarchy on every call in the covered period: (1) fund organic business investment; (2) support accretive client-driven RWA deployment; (3) return excess capital via buybacks and dividends. No inorganic uses of capital are permitted under the current framework — a constraint reinforced by the OCC order's acquisition-approval requirement.

CET1 position and constraint stack (Q2 2026):

CET1 evolution:

FY2025 capital walk — and an honest gap. CET1 capital rose $1.7B. Against net income of $14.3B, the firm paid $1.11B of preferred dividends, $4.34B of common dividends, and repurchased $13.25B of stock — a net drain of $4.4B. Reconciling to +$1.7B therefore requires approximately $6.1B of other positive capital movement, of which the disclosed drivers are AOCI improvement (Citi is a full-AOCI-inclusion bank, so falling rates flow through to capital), declining DTA deductions, and share-based compensation. These are not individually quantified in available filings, and the walk cannot be fully closed from disclosed data. The CECL transitional provision (+$757M at December 2024) phased out during Q1 2025 and was a capital headwind, not a tailwind.

The H1 2026 walk has a similar gap: capital rose $448M against approximately $1.35B of net drain, with $570M of DTA improvement and $281M of own-credit DVA identified — leaving roughly $1B unexplained.

Capital return history:

FY2025 total payout was 133% of net income to common (dividends plus buybacks ÷ $13.2B) — a deliberate drawdown of CET1 from 13.6% toward the ~12.6% target, enabled by the SCB reduction and the Russia capital release.

The $30 billion program. Announced at the 7 May 2026 Investor Day. $4.0B executed in Q2 2026, all in June, at an average price of $138.58 per share, leaving $26.0B remaining at 30 June 2026. Q1 2026's $6.3B single-quarter repurchase was the largest in Citi's history. The common dividend was increased 12% to $0.67 per share per quarter effective Q3 2026.

The $138.58 average execution price is the anchor for §8. At a tangible book value of $100.89, Citi was repurchasing its own stock at approximately 1.37× tangible book value. This changes the sign of the buyback's effect on per-share tangible value, and it is developed quantitatively in §8.

Other capital items:

• DTA burn-down: disallowed deferred tax assets of $13.9B at end-2025; $13.4B at Q2 2026; management targets an $800M reduction in FY2026. Each dollar released adds directly to CET1.

• Banamex: approximately $5B of capital and $40B+ of RWA remain tied up. 25% was sold in April 2026; full deconsolidation is expected in early 2027, releasing that capital plus an additional ~$27B RWA reduction from deconsolidation of the Mexican financial statements. Roughly $9B of cumulative translation adjustment losses sitting in AOCI will be recycled to the P&L on deconsolidation — a non-cash, capital-neutral event that will nonetheless depress reported ROTCE for approximately one quarter.

• Russia: AO Citibank sold in Q2 2026, eliminating approximately $1.6B of CTA deconsolidation risk and releasing ~$4B of capital in Q1 2026, which funded the record Q1 buyback.

• Stress Capital Buffer: improved three consecutive years (4.3% → 4.1% → 3.6%). The 2026 DFAST implies 3.3%, but the Federal Reserve has confirmed the buffer stays at 3.6% until 1 October 2027 pending finalization of revised models.

• G-SIB surcharge: currently 3.5% under Method 2, expected to rise toward 4.0% for the 2028 requirement. Fraser has publicly called the surcharge "gold-plated relative to the Basel standard" and committed to advocating for reform.

Culture and incentives

Fraser's performance share units (50% of incentive compensation) are indexed to three-year weighted-average ROTCE and cumulative TBVPS growth — a direct alignment with the two metrics this report uses to value the firm. A Special Transformation Incentive Program ran from August 2021 to December 2024 for senior leaders, tied to transformation goal achievement, and concluded as the program neared completion. Firm-wide incentive compensation rose to $2.26B in FY2025 from $2.13B in FY2024.

Risk culture is designated the first pillar of the Enterprise Risk Management Framework. The Chief Risk Officer (second line) reports to both the Board Risk Management Committee and the CEO; Internal Audit (third line) reports to the Audit Committee Chair. The Board RMC oversees "risk culture" as an explicit board-level accountability.

A note on evidence quality for this entire section: every assessment above is drawn from management-narrated sources — eight earnings call transcripts, the FY2025 10-K, the Q2 2026 8-K supplement, and a compensation 8-K. No independent third-party assessment of leadership quality, no regulatory examination material, and no critical external analysis was available. For a firm whose central issue is a six-year regulatory remediation, this is a material limitation on how much weight §4 can bear.

5. Financial Quality and Earnings Sustainability

The ROTCE trajectory

Two caveats attach to the endpoints of this series:

FY2021's 13.4% is not a clean baseline. It benefited from approximately $9B of allowance releases as COVID-era reserves were unwound. Normalizing for reserve releases, the 2021 run-rate ROTCE was materially lower. This figure should not be treated as "what Citi used to earn."

H1 2026's 13.1% is not a clean run-rate either. The reasons are the subject of §7.

Revenue quality

The most important structural claim management makes is that NII growth is volume-driven, not rate-driven:

"The vast majority of the growth we have baked in for the year is driven by client engagement and momentum reflected in deposit and loan growth." — CFO Gonzalo Luchetti, Q2 2026

The rate sensitivity disclosures support this for the banking book: a ±100bp parallel shock moves NII by only ±$1.3–1.4B against a $59.8B base — roughly 2%. NII is genuinely more volume- than rate-sensitive.

The qualification is that "volume" itself is cyclical. Services deposits at +19% and prime brokerage balances at +60% are volumes that expand and contract with global liquidity conditions. Volume-driven is not the same as durable.

Fee revenue durability by segment:

• Services: Q2 2026 fee revenue $1.8B (+22%); cross-border transactions +13%. Backed by transaction volume, which is recurring. Most durable.

• Wealth: fee revenue $793M in Q2 2026; net new investment assets $30.4B in H1 2026 (+64%). Recurring, scaling. Durable.

• Banking: IB revenue $1.55B in Q2 2026 (+44%); DCM +65%, ECM +92%. Market-dependent and cyclical. Least durable.

• USCC: net fee revenue is structurally negative; revenue durability is entirely NII-dependent.

Credit quality

The central credit fact: consumer NCLs were essentially flat between 2024 (2.24%) and 2025 (2.22%), which suggests the post-CECL normalization cycle has plateaued rather than accelerated. The 90+ delinquency leading indicator turned down in Q2 2026 (1.39% → 1.21% for Branded Cards). H1 2026 total NCLs of $4.6B were 2% lower than H1 2025.

Concentration. Branded Cards and Retail Services represented approximately 94% of USPB's total NCLs. Institutional credit is a rounding error: in Q3 2025, Services ($61M), Markets ($32M), and Banking ($157M) together contributed $250M, or roughly 11% of firm-wide quarterly NCLs, against a card book generating $1.8B in the same quarter. 78–79% of corporate exposure is investment grade.

Two flags:

1. Retail Services (private label cards) carries a 4.79% NCL rate against 10.8% ACLL coverage — the highest reserve ratio in the consumer book, reflecting structurally elevated intrinsic risk in the co-brand retail partner portfolios.

2. Mexico consumer NCLs rose three consecutive years: 4.01% (2023) → 4.52% (2024) → 5.56% (2025), with 90+ delinquency at 1.79% in Q2 2026. This deterioration overlaps with EM geopolitical exposure and is exiting the perimeter with the Banamex deconsolidation.

Other credit observations: non-accrual loans rose 35% in FY2025 to $3.62B, with corporate non-accruals rising from $1.38B to $2.00B on idiosyncratic downgrades; ACLL-to-non-accrual coverage declined from 691% (2024) to 532% (2025). Commercial real estate exposure grew from $65B (year-end 2024) to $78.4B (year-end 2025, +20.6%) to $107.5B (30 June 2026), driven by increased Markets client activity in real estate; reserve coverage on funded CRE declined from 1.6% to 1.1% as balances grew. California wildfire residential mortgages contributed to the FY2025 increase in consumer non-accruals.

The three efficiency levers

CFO Luchetti has identified three distinct components of the expense improvement:

1. Stranded costs — overhead left behind by exited consumer franchises. Declining from a $1.3B annual run-rate toward $800M as the overhead is absorbed. Finite, mechanical.

2. Transformation costs — peaked at approximately $3.3–3.4B total (half sitting in Corporate/Other as temporary charges) and declining as programs reach completion. Finite, but the timing is regulator-controlled.

3. AI and technology automation — 100+ end-to-end processes mapped; currently front-loaded with severance (~$800M in H1 2026, of which $500M in Q1 alone). Structural, but the run-rate benefit is not separately disclosed.

Expense-versus-revenue trajectory: FY2025 revenues +6% against flat-to-down expenses on an adjusted basis (+266bp operating leverage); H1 2026 revenues +14% against expenses +5–6% (~+500bp operating leverage).

The structural gap between segment and firm ROTCE

This is the single most important analytical fact in the report.

Every operating segment earned between 11% and 29% in FY2025. The firm earned 7.7%. The five segments produced $19.9B of net income; the firm reported $14.3B. The $5.6B difference — the All Other / Corporate drag — is the entire story of Citigroup's ROTCE gap.

It has three components:

1. Transformation and severance costs held in Corporate/Other rather than allocated to segments (~$3.4B of transformation spend in FY2025, roughly half in Corporate/Other, plus ~$800M of severance).

2. Divestiture-related and legacy items — Russia, Banamex, stranded costs from exited consumer franchises.

3. Denominator and definitional differences — preferred dividends deducted in the consolidated numerator but not in segment ROTCE, and management-allocated segment TCE not summing to GAAP average TCE.

Therefore the primary lever for consolidated ROTCE expansion is the elimination of the drag, not a shift in segment mix. Mix shift toward Wealth and USCC does not mechanically lift consolidated ROTCE unless capital is reallocated; the drag does, dollar for dollar. Roughly $1.7B of after-tax drag reduction is worth 1 point of firm ROTCE, which under the Gordon framework is worth about $14 per share.

One-off items to strip out

Non-recurring, FY2025:

• Russia AO Citibank held-for-sale accounting (Q4 2025): NIR loss of $1.173B; net income impact −$1.123B. Includes ~$1.6B of CTA reclassified from AOCI — capital-neutral. Entity sold February 2026, releasing ~$4B of capital.

• Banamex goodwill impairment (Q3 2025): ~$726M, non-cash, capital-neutral. Adjusted FY2025 net income $16.1B (+27%); adjusted ROTCE 8.8%.

Explicitly transitional:

• Transformation costs ~$3.4B in FY2025, declining from a $3.3B peak run-rate.

• Severance ~$800M in FY2025; ~$800M in H1 2026 alone.

• Stranded costs from $1.3B toward an $800M annual run-rate.

Known future one-off:

• Banamex CTA deconsolidation (~early 2027): approximately $9B of CTA losses in AOCI to be recycled to the P&L, temporarily depressing reported ROTCE for roughly one quarter. Capital-neutral (already in common equity via AOCI) and non-economic.

AOCI note: Citi is a full-AOCI-inclusion bank and did not elect the opt-out. Available-for-sale unrealized losses and cumulative translation adjustments flow through both GAAP common equity and CET1. Cash flow hedge AOCI and own-credit DVA are excluded from CET1. This differs from peers who elected AOCI opt-out during the CECL transition, and it makes Citi's capital ratio more rate-sensitive than theirs.

Liquidity: consolidated available liquidity of $1.1 trillion at 30 June 2026 ($648B of HQLA in the LCR calculation, $252B of non-transferable unencumbered HQLA, plus unused FHLB/Fed capacity); average LCR 114%. At year-end 2025: HQLA $607.2B, LCR 115%.

6. Risks and Bear Case

Four risk clusters, ordered by how much of the investment case each can destroy.

Risk 1 — Consent order removal is entirely at regulators' discretion

The exposure. The October 2020 FRB and OCC Consent Orders remain in force as of Q2 2026. In July 2024 regulators escalated rather than relaxed, imposing civil money penalties (FRB $61M; OCC $75M) and formally finding that Citibank "had failed to make sufficient and sustainable progress toward achieving compliance with its 2020 Consent Order." The December 2025 partial milestone was the termination of the July 2024 amendment, not of the underlying order.

The FY2025 10-K's Item 1A risk factor is the hardest-worded formal disclosure the firm makes:

"There can be no assurance that efforts by Citi to address the deficiencies and resolve the OCC and FRB Consent Orders will occur in a manner satisfactory, in both timing and sufficiency, to the FRB and OCC."

Why it matters beyond cost. While the OCC order stands, the OCC must approve any significant acquisition and retains the right to impose business restrictions, limit Citibank dividend declarations, and require changes to directors and senior executives. Transformation expenses were ~$3.3B in FY2025 (+14% year-over-year).

Bear case. Regulators conclude data governance in regulatory reporting is not remediated to their standard. New civil money penalties are imposed. Transformation expense continues beyond 2026. Buybacks or dividends are curtailed. The "around 60%" efficiency target and the medium-term ROTCE path are delayed by one to two years.

Risk 2 — The H1-to-H2 arithmetic implies a sharp step-down

This is the most quantifiable near-term risk in the report, and it comes from the company's own numbers.

H1 2026 ROTCE was 13.1%. Management explicitly reaffirmed full-year 2026 guidance of 10–11% at the Q2 2026 earnings call. The arithmetic is unavoidable:

If H1 = 13.1% and the full year = 10–11%, then H2 must run at approximately 7–9% — a sequential step-down of 4 to 6 percentage points.

Two drivers explain most of it:

Markets seasonality. CFO Luchetti guided that Markets revenues historically decline approximately 20% from H1 to H2, and warned that given H1 2026 performance "the magnitude of that decline could be greater this year." He added: "from one day to the next, trading revenue can swing in a relatively short period of time to a negative position for the industry as a whole." Markets Q2 2026 revenues were +17% year-over-year with Equities +45% — creating an elevated comparison base.

U.S. Cards negative operating leverage. USCC revenue grew only 1% in Q2 2026 while expenses rose 10%, driven by severance, marketing, and customer engagement costs related to the Barclays AA portfolio acquisition (~$6.6B of loans and 2 million accounts acquired in April 2026). Year-to-date: revenue +3%, expenses +6%. Management explicitly guided that expenses will outpace revenue in cards "for the next few quarters."

Bear case. Markets H2 decline exceeds 20% on adverse macro or geopolitical conditions. Cards investment spending persists longer than guided. FY2026 ROTCE prints below the 10% lower bound. The 14–15% medium-term target is deferred beyond 2028.

How to read this risk. Whether the H2 step-down is a risk or a known quantity depends on the reader's prior. It is fully disclosed, management-guided, and arithmetically certain given the guidance. What is not certain is whether the 13.1% H1 or the 10–11% full year is the better guide to normalized earning power. §7 addresses this directly.

Risk 3 — Credit is concentrated in U.S. cards and deteriorating in Mexico

Total firm-wide NCLs rose from $3.8B (2022) to $9.1B (2025) — a 140% increase in three years, before stabilizing in FY2025. The U.S. card book is overwhelmingly the driver.

Current position is better than a year ago: USCC NCL rate 4.19% in Q2 2026 versus 4.43% in Q2 2025; 90+ delinquency improved to 1.31% from 1.45% at year-end 2025. Total ACLL at Q2 2026 is $20.0B (a $0.8B build in H1 2026); total reserves including unfunded commitments are $22.2B.

Bear case. U.S. unemployment rises toward the 6.9% downside scenario embedded in Citi's own ACL models, triggering approximately $4B of additional allowance build. Mexico consumer deterioration accelerates from its already-elevated 5.56% NCL rate. CRE stress materializes, particularly in the commercial sub-book ($49.3B funded at Q2 2026), where reserve coverage has thinned from 1.6% to 1.1%. Total annual provisions exceed $10–12B, compressing ROTCE by 4 to 5 percentage points.

Risk 4 — Capital requirements constrain the pace of return

CET1 at Q2 2026 is 12.78%, approximately 118bp above the 11.6% requirement. Management targets ~12.6% under current rules, with an Investor Day target of ~13.1% by 2028 to absorb the expected G-SIB surcharge increase from 3.5% to 4.0%.

The G-SIB surcharge is the primary source of Citi's requirement uplift versus the Basel standard. Fraser has publicly called it "gold-plated relative to the Basel standard" and raised three advocacy positions: material double-counting of operational and market risk between the surcharge and the SCB; the SCB not reflecting post-divestiture risk profile; and the standard itself. Management expects a "moderately net positive" outcome from Basel III finalization overall.

Bear case. An adverse DFAST result reverses the SCB improvement trend (e.g., back toward 4.1%), raising the requirement from 11.6% to 12.1%+. The G-SIB surcharge is not reformed and rises with balance sheet growth. Buybacks are curtailed or suspended. The 13.1% CET1 target for 2028 forces RWA de-risking at the expense of business investment.

Risk 5 — Macro, emerging markets, and FX translation

• Rates: the USD book is near-neutral (−$33M on +100bp). The adversarial scenario is cuts: −100bp parallel shock costs approximately $1,338M of NII. The non-USD book is structurally asset-sensitive (+$1,402M on +100bp) and diversified across 65+ currencies. NII ex-Markets growth for 2026 is guided at 5–6%, driven by volume.

• Emerging markets: EM revenues represent approximately 25% of FY2025 revenues, or roughly $5.5B of exposure to macro volatility, with $51.7B of TCE deployed. As of Q2 2026 management cited Middle East conflict, inflation, and slower global growth as active headwinds.

• FX translation: a 5% instantaneous appreciation of the U.S. dollar against all currencies would reduce TCE by approximately $1.7B (1.0%), driven primarily by EUR, MXN, and SGD exposures. This is a structural feature of Citi's geographic footprint that U.S.-centric peers do not carry.

Composite bear-case framework

Risks not covered by available sources

The following are material to the investment case but could not be assessed from the sources in this analysis, and their absence should be treated as a gap rather than as evidence of low risk:

• Stablecoin and tokenized-deposit disruption to cross-border settlement — the direct attack surface on the TTS moat.

• Prime brokerage counterparty concentration — balances up ~2.4× in two years, with no disclosed concentration or leverage detail.

• Management's long-run track record against stated targets — no independent base rate available.

• Independent regulatory examination material — all consent-order progress reporting is management-narrated.

7. Normalized Earnings Power

The single most important question for valuation is not "what did Citi earn in H1 2026?" but "what will Citi earn in an average year?"

Starting point and adjustments

Estimate: 12–13% normalized ROTCE, midpoint 12.5%.

How to read this estimate

It sits below the H1 2026 run-rate of 13.1%, above management's FY2026 waypoint of 10–11%, inside the 11–13% near-term (2027–2028) target, and below the 14–15% medium-term (2029–2031) target.

The three downward adjustments (−2.2 points) are the ones a value investor should care about most, because each is a cyclical tailwind currently being reported as if it were structural. The three upward adjustments (+1.85 points) are largely mechanical and well-disclosed, though the "further All Other drag reduction" line is the most judgmental item in the build and is worth stress-testing: setting it to zero moves the estimate to ~11.9%; doubling it moves it to ~13.5%.

What would push normalized ROTCE to 14–15% (management's medium-term target): full consent order closure eliminating the remaining transformation drag; Wealth reaching its 15–20% segment target from 14.4%; the All Other drag falling to near zero; and Banamex fully out of the perimeter. All four are plausible. None is assured, and the fourth depends on a 2027 event whose timing has already moved.

What would push it to 9–10%: Markets reverting to the FY2024 level rather than mid-cycle; credit normalizing above the guided 4.0–4.5% card NCL range; and transformation expense persisting beyond 2026 on regulator timing.

8. Valuation

Starting position

What the current price already assumes

Using the Gordon relationship from §1 (P/TBV = (ROTCE − g) ÷ (CoE − g)) with g = 3.5% and CoE = 10.5%:

At $137.72, the market is capitalizing a sustainable ROTCE of approximately 13.0%.

That is at the top of management's 11–13% near-term (2027–2028) target, above the FY2026 guidance of 10–11%, and above this report's normalized estimate of ~12.5%. Against the normalized estimate, fair value is approximately $130, and the current price sits ~6% above it.

Cost-of-equity sensitivity — the single largest source of uncertainty in the framework:

Fair-value range on normalized earnings: $121–140. The current price is at the upper end.

A note on the cost of equity: 10.0–10.5% is calibrated from where peers trade (Bank of America at ~14.2% ROTCE and ~1.4–1.8× TBV implies ~10.2%). A case exists for assigning Citigroup a higher CoE than its peers — open consent orders, execution risk on a multi-year target, 25% EM revenue exposure, and FX translation risk that U.S.-centric peers do not carry. At an 11.5% CoE, fair value on 12.5% normalized ROTCE falls to ~$113.

The buyback:

At 1.365× tangible book, repurchases reduce tangible book value per share. The arithmetic:

Starting from TCE of $169.2B and 1,677.4M shares (TBVPS $100.89), assume one year at 12% ROTCE ($20.3B of earnings), $4.5B of dividends, and $10B of buybacks:

Buying back stock at 1.365× tangible book costs approximately 1.2 percentage points of annual TBVPS growth versus simply retaining the capital. At the $92 price, the same buyback added ~1.3 points. The price correction flips a 1.3-point tailwind into a 1.2-point headwind — a 2.5-point swing in the rate of per-share compounding.

This does not make the $30B program a mistake. Above 1.0× book, buybacks still increase earnings per share and are still preferable to leaving capital idle at a 12%+ ROTCE. But they are no longer the value-creation engine they were at 0.55× book in 2023, and the report should not describe them as though nothing has changed.

What the re-rating has already delivered

Decomposing the total price return: approximately 84% came from multiple expansion and 16% from tangible book value growth.

This is the most important context for anyone underwriting the position today. The discount-to-book that defined the Citigroup thesis from 2021 to 2024 has closed. What remains is a bet on operating improvement being delivered and sustained — a different, and considerably harder, bet than the one that worked.

Three-scenario valuation to year-end 2028

Horizon: 2.3 years from 4 September 2026. TBVPS is projected explicitly, incorporating dilution or accretion from buybacks at each scenario's own price level.

Bull — full roadmap execution. ROTCE reaches 14–15% by 2028–2029 (the Investor Day medium-term target). All five segments at or above target: Services 30%+, Markets 17–18%, Banking 18%+, Wealth 15%+, USCC 20%+. Consent orders lifted; efficiency ratio below 58%; All Other drag largely eliminated. Buybacks ~$12B/year, executed at a rising average price.

Base — near-term targets achieved and held. ROTCE reaches and sustains 12–13% through 2027–2028, consistent with the normalized estimate in §7 and the upper half of official near-term guidance. Revenue CAGR 5–7%; expenses grow 4–5%; DTA burn-down continues at ~$800M/year; buybacks ~$10B/year.

Bear — execution risk or macro deterioration. ROTCE reverts to 8–9%. Unemployment rises toward 7%; USCC NCL rate spikes to 5.5–6.0%; total credit costs $12–14B/year. Markets revenues decline 20%+ in H2 2026 and remain subdued. Consent order closure is delayed. Buybacks cut to ~$3B/year to preserve capital.

Cumulative dividends over the 2.3-year horizon are approximately $6.50 per share (~4.7% of the current price) and are included in total return. Returns are annualized over 2.3 years.

Note on the bear-case TBVPS. At 8–9% ROTCE the firm is still solidly profitable, so tangible book per share continues to grow — roughly 5–6% per year. A bear case in which TBVPS declines requires actual losses, which none of the disclosed stress scenarios produce. This is why TBVPS functions as a floor: the bear case is a multiple compression story, not a book destruction story.

Probability-weighted arithmetic

At a 25% / 50% / 25% weighting across bull, base, and bear:

Expected total return ≈ +11% over 2.3 years ≈ ~5% IRR

Against an estimated cost of equity of 10.0–10.5%, the probability-weighted expected return sits below the required return at $137.72. This is a statement about the model's output at these inputs, not a recommendation; readers assigning different scenario probabilities or a different cost of equity will get different answers, and the sensitivity is high — shifting the weighting to 40/45/15 raises the expected IRR to approximately 8%.

Entry-price sensitivity

Holding the base case terminal value constant at $151 plus $6.50 of cumulative dividends ($157.50 total):

On base-case assumptions, an entry price of approximately $125 clears a 10.5% cost of equity. On the bull case, the current price clears it comfortably; on the bear case, no entry price above ~$98 does.

EPS cross-check

An independent triangulation, to confirm the P/TBV framework is not producing an artifact:

EPS derived from ROTCE applied to projected average TCE, divided by projected diluted shares (declining from 1,677M toward ~1,540M on continued buybacks).

The EPS approach produces $99–192 against the P/TBV approach's $98–186. The two methods agree, which is expected — for a bank, P/E and P/TBV are algebraically linked through ROTCE — but the consistency confirms no arithmetic error.

Summary of the valuation position

The valuation question reduces to a single variable: whether normalized ROTCE is 12.5% or 14.5%. At 12.5% the current price is fair to modestly full. At 14.5% there is meaningful upside. The evidence for 14.5% is management's medium-term target and the segment-level trajectory; the evidence for 12.5% is the cyclical composition of the current run-rate and the persistence of the All Other drag. Nothing in the available disclosure resolves it, and the reader should form that judgment directly.

Appendix A — Canonical Data Set

Every figure in this report traces to this table. Where the source draft contained conflicting values, the value retained here is the one corroborated by the greatest number of independent sources; Appendix C records each choice.

Firm-level, five-year

FY2025 supporting detail

Prepared 9 September 2026. Reference price $137.72 (4 September 2026). Latest reported financial period: Q2 2026. This document is an analytical framework output for internal research use. It presents valuation ranges and the assumptions behind them; it does not constitute investment advice or a recommendation to buy or sell any security.