GBank Financial Holdings Inc. GBFH

Key data

The business

Bank of George does two things most community banks its size do not. It is one of the more active Small Business Administration lenders for a bank with about $1.4B in assets: it originates 7(a) loans, keeps the unguaranteed sliver, and sells the government-guaranteed portion into a hungry secondary market for an immediate gain. That gain-on-sale income is why a bank this small throws off fee revenue a bank twice its size would envy, and why its earnings jump around with loan-sale timing. The second engine is deposits. GBank banks the gaming and gaming-payments world out of Las Vegas, prepaid and settlement accounts that sit cheaply and do not chase rate. The moat, if there is one, is that funding base: deposits the casino-payments plumbing leaves behind at a lower cost than the bank could buy them. The loan book and its credit ride on how durable that cheap money proves.

The numbers

Strip out the gain-on-sale noise and look at the engine. Net interest income, what the bank earns lending minus what it pays for deposits, has climbed every year and holds near $13M a quarter.

The quarters carry the real story of 2026. Net interest income barely moved, holding around $12.5M, but net income swung from $7.4M in the fourth quarter of 2025 to $1.3M in the first quarter of 2026 and back to $5.5M in the second. That first-quarter hole was not a revenue problem, it was a provision, the bank setting aside money against loans it expects to sour. The $0.38 in June still landed below what the two analysts covering it wanted, the fourth light print in five quarters, which is most of why the stock sits where it does.

Across the full years the pattern is a bank that grew its lending engine fast while per-share earnings sat still, then broke higher. Earnings per share held at $0.84 to $0.85 from 2021 through 2023 even as net interest income more than doubled, because the share count kept rising and provisions and loan-sale timing ate the difference, then jumped to $1.44 in 2025. Half of 2026 is in the books at $0.47, tracking below last year's pace.

So the compounding math is honest but unspectacular. Earnings per share grew about 14% a year from 2021 to 2025, but diluted shares grew 12% over the same stretch, so a good chunk of the growth was the company selling more of itself. Tangible book value is $11.12 a share and you pay 1.8 times that. Against a return on tangible equity of roughly 11%, 1.8 times book is a premium, not a bargain, a multiple that only works if the fee engine re-accelerates and credit behaves. A year ago the market paid almost four times book for exactly that. The re-rating from there to here is the whole move.

Which brings it to credit, the number that decides a bank. Loans on nonaccrual, borrowers who have stopped paying, climbed to $51.6M at June 30 from $32.1M two quarters earlier, while the cushion set against the whole loan book rose only from 1.03% to 1.19%. The bad loans are growing faster than the reserve behind them. The quarter that would test whether that gap closes has not been filed yet: the third quarter, due in November, is the first print where reserves either start catching the nonaccruals or the provision line does it for them again, the way it gutted the first quarter. That is the variant. The market is pricing a credit problem, and the bet against the market is that this is loan-book seasoning reserves absorb, not the front edge of a cycle. One print settles which.

Management

The record is a founder's. Edward Nigro, who built the bank, stepped back into the chief executive seat this past year after T. Ryan Sullivan left the president and CEO role, so the person running it now is the person whose name is on it. Insiders have voted with cash: fifteen open-market purchases against seven sales over twelve months, a net $1.8M bought, about 0.6% of the whole company, led by director Charles Griege's two buys of $833K and $696K. Set against that, the chief financial officer sold $748K in March, a discretionary sale, not a scheduled one. Nigro's $1.3M pay is a modest 6% of profit. The Chief SBA Officer's $4.2M package is not, and it measures how much of this bank's earnings ride on selling loans. The mark against them is the share count, 12% more of the company outstanding than four years ago.

How it fails or surprises you

The reserve never catches the bad loans. Nonaccruals at $51.6M against a 1.19% allowance is a gap, and the first quarter of 2026 already showed what closing it costs: a quarter of earnings gone to provision. If the back half of 2026 brings more of the same, this is a down year and 1.8 times book was too much. The tell shows first as the allowance ratio climbing while nonaccruals refuse to come down.

The SBA window narrows. A large share of profit is gain-on-sale on guaranteed loans, income that swings with rates, secondary-market appetite, and Washington's loan rules. The Chief SBA Officer's $4.2M pay measures the dependence. A quarter where loan-sale gains fall and net interest income has to carry the P&L alone would reveal how much of the earnings power was the funding base and how much was the sale desk.

The cheap deposits stay cheap and the engine re-rates (right tail). The bet the $42 buyers made, a gaming-payments deposit base funding a high-return SBA machine, is doubted, not dead. If cost of deposits holds while the noninterest-bearing share stays high and gain-on-sale recovers, return on tangible equity steps back toward the mid-teens and 1.8 times book is cheap for it. The market will not pay for that today because credit fear owns the tape. The tell is deposit cost flat or falling across the next two prints while nonaccruals stabilize.

Closing thoughts

This is not a fog you sit in, a specific number resolves it. The November filing either shows the allowance climbing to meet the $51.6M of nonaccrual loans, which turns today's fear into a reserved-for problem and makes 1.8 times tangible book look like the low it was, or it shows nonaccruals still running ahead behind another heavy provision, which turns $0.09 quarters from an accident into a pattern and 1.8 times book into a still-falling knife. An ambiguous print, nonaccruals flat and reserves flat, means you wait for the fourth quarter and keep the position small. The left tail is real and near, the credit cycle takes this down, not the multiple. Which tail is fatter is a judgment, and mine is that the loss you can name sits closer than the gain you are hoping for.

The bet, still, is that a small Las Vegas bank keeps making government-guaranteed small-business loans and funds them with cheap money from the casinos and gaming-payment firms it banks. What breaks it is the loans it keeps going bad faster than it reserves, and the one pair of numbers that tells you first is nonaccrual loans against the allowance ratio, side by side, every quarter. They stood at $51.6M against 1.19%. If the gap is wider in November, the market was right and the buyers at $20 were early or wrong. If the reserve has closed it, they were neither.