Citigroup Inc. C

Key data

The business

Citigroup is a global bank rebuilt around five lines: Services (its crown jewel, moving cash and securities for multinationals and governments), Markets (buying and selling securities for clients), Banking (deal advice and corporate lending), Wealth, and US Personal Banking, mostly branded credit cards like Costco and American Airlines. For a decade it was the opposite of a compounder, too sprawling and under-earning, selling off consumer banks from Mexico to Asia and swallowing charges to do it. What is left is the plumbing of global commerce, the part rivals cannot easily copy, plus a large, profitable US card book. The moat is Services, a switching-cost network where a corporate treasurer who clears dollars through Citi in ninety countries does not move for a few basis points. The tension: the good part sits inside a company the market spent years distrusting, and that distrust discount is what is now closing.

The numbers

The direction is unmistakable in the last five quarters. After a soft fourth quarter, earnings stepped up hard.

The fourth quarter of 2025 was the air pocket, EPS of $1.19, down 11% and well short of the $1.80 analysts carried, dragged by seasonally weak markets and year-end charges. Then the two 2026 quarters landed 56% and 61% above the year before, each clearing forecasts by more than 40 cents. The swing is not revenue heroics. Second-quarter net income rose 45% on a 14% revenue gain, and the extra sixteen points of earnings growth came from a shrinking share count.

Read the per-share line, not the totals. Net income in 2025 sat a third below 2021, but the per-share picture is better, because Citi is aggressively retiring stock: 2025 buybacks hit $13.3B, more than five times the 2024 pace, done while the shares changed hands near or below tangible book. Buying back stock under book is the one lever that mechanically builds tangible book value per share for everyone who stays, and it is the heart of the case. First-half 2026 EPS of $6.21 already nearly matches all of 2024. The pairing the lens demands is return on tangible common equity against price to tangible book. Reported returns are marching from high single digits toward low teens on the 2026 run-rate, and at about 1.3 times tangible book the stock already prices low-teens returns as durable. That is the variant in one line: the market has stopped pricing Citi as broken but has not yet paid for it as good, and the print that settles which is right is whether quarterly RoTCE holds in the low teens once markets revenue cools and provisions normalize.

Credit, the thing that sinks banks, is quietly improving.

Charge-offs peaked at 1.33% of loans in early 2025 and have fallen every quarter since, to 1.02% by the first quarter of 2026, while loans going bad have held flat near 0.75%. The credit line to watch has held: card net losses eased to 4.01% of balances from 4.20% a year earlier, and reserves are drifting down because the portfolio is getting better, not because the bank is starving the cushion. This is a benign, improving picture, which is exactly the setup that flatters a bank right before the cycle turns, so the reserve at 2.4% of loans is the number to keep honest.

Management

Jane Fraser inherited a conglomerate and has spent four years making it smaller on purpose, exiting more than a dozen international consumer markets and taking the charges to do it. Capital allocation is the record that matters, and it is strong: $13.3B of buybacks in 2025 at roughly book value, a dividend just raised 12% to $0.67 a quarter, and CET1 capital at 12.8%, about 120 basis points above the requirement. The guidance record is three beats and one clear miss, the weak fourth quarter Fraser did not sugarcoat. Insiders sold about $23M over the year and bought nothing, led by the banking head and two other officers, sales that are routine for a megabank at these prices, plan status not disclosed, no signal either way. Pay is tied to the tangible-return target the whole case rests on, which is the right thing to reward.

How it fails or surprises you

Normalized returns disappoint. The stock prices low-teens returns as permanent. Strip out unusually strong 2026 markets revenue and an unusually low provision, and the through-cycle return is closer to 10%. If RoTCE settles there, 1.3 times tangible book is full, not cheap, and the re-rating stalls. The first sign: a quarter where provisions climb back toward $3B with no revenue to match.

The re-rate to peers (right tail). If low-teens RoTCE proves durable through 2027, there is little reason Citi stays at 1.3 times tangible book while peers fetch near 2.0. Closing even half that gap is a large move on top of double-digit tangible-book growth from buybacks. The market will not pay for it because it has been burned before. Two more clean quarters of returns and buybacks change the story.

Credit is too good. Charge-offs at 1.02% and reserves easing to 2.4% is the memo's own soft spot, the picture that looks best just before it turns. Card losses at 4.01% are benign now, but cards are where a consumer slowdown shows first. If early 30-89 day delinquencies, already up to 1.17%, keep rising while reserves fall, the earnings surge reverses fast.

Closing thoughts

Whether the stock re-rates to peers comes down to quarterly returns: the next few quarters will show if low-teens RoTCE survives normal provisions and quieter markets. The evidence says Citi is a genuinely better bank than the one the market punished for a decade, and the buyback under book value builds per-share value regardless of the multiple. What decides the re-rate is whether RoTCE holds low-teens through normal provision levels and quieter markets quarters. The named print is quarterly RoTCE over the next few quarters: hold low teens and the peer discount is unjustified and closes, slip back to 10% and today's 1.3 times tangible book is simply fair. The left tail is a credit turn hitting the card book while reserves are lean. The fatter tail, given improving credit and relentless buybacks below book, points up.

The bet is still that Citigroup is finally a simpler, better bank and stays one, its five businesses earning more while it retires cheap stock. What breaks it is credit turning on the card book before returns prove durable, and the one pair of numbers that tells you first is quarterly RoTCE against the card net-loss rate. Low teens with losses near 4% says it is working. High-single-digit returns with losses pushing past 5% says the discount was deserved.