Mizuho Financial Group, Inc. MFG

Key data

The business

Mizuho Financial Group is one of Japan's three megabanks, alongside Mitsubishi UFJ and Sumitomo Mitsui. It takes deposits from households and companies across Japan, lends to the country's largest corporations, and runs a trust bank, a securities dealer and an asset manager on top. The engine underneath everything is a deposit base of ¥61.6 trillion (as of June 2026) that, through two decades of zero and negative rates, earned the bank almost nothing. Depositors left their money because moving it bought them nothing either. Now the Bank of Japan is lifting rates for the first time in a generation, and that mountain of cheap, patient money finally earns a spread. That is the moat in one line: scale plus depositor inertia at one of three names the country trusts with its savings. Set against it, about a third of revenue comes from markets and fees that swing with the cycle, plus a large global corporate and investment bank that competes on price far from home. The saver getting paid on a deposit for the first time in twenty years is the whole story.

The numbers

For years the interest Mizuho kept after funding costs barely moved. Then it jumped.

Net interest income crept near ¥1.2 trillion for years, then rose 34% in the year to March 2026 as the loan book repriced upward faster than deposits. That single inflection is the entire re-rate. The live proof sits in the June quarter.

The loan-deposit spread widened to 1.27% from 1.04%. What Mizuho earns on loans rose 38 basis points, what it pays on deposits rose only 15. The spread is expanding because savers have barely begun to ask for more. That lag is the case for owning it and the clock at once, because deposit costs catch up eventually.

Credit is quiet. The reserve line is not.

Bad loans fell to 0.70% of the book, yet the money set aside for losses nearly doubled to ¥188 billion. Reserves are being built ahead of any visible stress, prudent against a bigger, repricing book, and worth watching if it turns out to be early recognition instead. Reported profit reached about ¥1.16 trillion, on a vendor basis that runs above the company's own profit-attributable figure. What this memo doubts that the tape does not: the spread is closer to its peak than its start, because loan yields have already moved and deposit costs have not. The print that settles it is next quarter's two deltas, side by side.

Management

No insider bought or sold a share in the open market over the past year, which is normal for a Japanese issuer and carries no signal, so read the capital account instead. Buybacks were the loud move: about ¥404 billion repurchased in the year to March 2026, four times the prior year, with repurchased stock cancelled outright, and the monthly filings show the program still running. That is real capital return, but it was done with the stock at roughly 1.9 times tangible book, the richest it has been in over a decade, a strange price at which to be the most aggressive buyer. Much of the firepower comes from unwinding Japan's old cross-shareholdings, selling stakes held for relationship reasons and handing the proceeds back. Equity still compounded to ¥10.9 trillion, and the ADR earnings have come in above the consensus estimate four quarters running. The judgment to weigh: aggressive return, bought dear.

How it fails or surprises you

The margin plateaus (downside). The whole 1.9x book rests on the spread widening. If the next loan-deposit print lands flat near 1.27% while cost of deposits keeps climbing 15 basis points a quarter and loan yields stall, the spread has peaked and a top-of-range multiple has nothing left to discount. Watch the two deltas: the quarter loan yields stop outrunning deposit costs is the quarter the story changes.

The reserve is a warning, not prudence (downside). Credit costs nearly doubled to ¥188 billion while bad loans fell. If that is early recognition of stress in a fast-repricing loan book or in the global corporate arm rather than conservatism, charge-offs follow the provision. Watch the NPL ratio turning back above 0.80% and the provision run-rate climbing a second year.

Savers stay asleep (right tail). Deposit costs rose only 15 basis points against 38 on loans, so the deposit beta is still tiny. If Japanese households stay inert through the next two hikes, the spread widens further than a plateau assumes and cross-shareholding sales keep fuelling buybacks. The market prices a peak. A second year of 20-basis-point spread gains is not in the 1.9x. Watch cost of deposits holding under about 0.5%.

Closing thoughts

This is a setup one print resolves, and the print is the quarterly pair of loan yield against deposit cost. If deposits keep lagging, the re-rate to 1.9 times book is earned and extends. If deposit costs catch up, 1.9x becomes a peak multiple sitting on peak margins, the worst pairing there is. The left tail, a Japanese downturn hitting the book mid-reprice, is real but survivable, with equity near ¥11 trillion and bad loans at 0.70%. The fatter tail is the milder one, plain multiple compression, because you are paying the top of a twelve-year range for a margin that arithmetic puts closer to peak than trough. A slide back toward the old 0.7 to 1.0 times book would roughly halve the equity with earnings flat.

The bet is still that Japan stays off zero and Mizuho's loans keep out-earning its deposits by a widening gap. What breaks it is the day cost of deposits rises as fast as loan yields, and the pair that tells you first is return on loans against cost of deposits, quarter over quarter. If that spread prints flat or lower next quarter while you are paying 1.9 times book, the market handed you the peak and called it a compounder.