Infrastructure

The Shortage Windfall Left As A Buyback

Kioxia posted a 75 percent operating margin and authorised up to 800 billion yen of share repurchases on the same day. The buyback is the number that tells you how long memory stays expensive.

InfrastructureBy Servola Tech Desk2026-08-024 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

Key takeaways

Kioxia authorised a treasury share facility of up to 800 billion yen, roughly 4.4 billion euro, on the same day it guided to a larger quarter, so read the memory shortage as a payout to shareholders rather than a problem being solved.
Non-GAAP operating profit was 1,326.2 billion yen on revenue of 1,767.1 billion yen, a margin near 75 percent, and average selling prices rose about 70 percent. Margin at that level is scarcity being priced, not efficiency being found.
The quarter missed profit estimates and the shares still rose about 18 percent, which tells you the market is now paying for the distribution rather than the operating result.
Before signing a 2027 hardware specification, ask your supplier what it returned to shareholders against what it committed to capacity. Cash paid out in August cannot become wafer starts in 2028.

Three notices landed on the same afternoon

On 31 July, Kioxia Holdings published its consolidated results for the three months ended 30 June under IFRS. Alongside them it filed two further disclosures: a notice of a stock split, and a notice establishing a treasury share acquisition facility. Read separately they are routine corporate housekeeping. Read together, filed within hours of each other, they are a statement about where the money from a memory shortage is going.

The filing to read first is the buyback. The facility authorises the repurchase of up to 30 million shares, about 5.5 percent of shares outstanding excluding treasury stock, for a maximum of 800 billion yen. The split is three-for-one, with a record date of 30 September and an effective date of 1 October. Neither instrument adds a single wafer of capacity. Both widen and reward the shareholder base.

Seventy-five yen of profit in every hundred of revenue

Revenue for the quarter was 1,767.1 billion yen, up 76.2 percent on the prior quarter and 415.5 percent on the same quarter a year earlier. Non-GAAP operating profit was 1,326.2 billion yen. That is an operating margin close to 75 percent, on a product that has spent most of its history as a commodity sold at or below cost through the trough of every cycle.

The company was explicit about where it came from: average selling prices rose roughly 70 percent, with demand concentrated in data centre and enterprise customers building AI infrastructure. A margin like that is not a manufacturing achievement. It is the price of scarcity, collected. Kioxia also ended the quarter in a net cash position, which removes the usual excuse that a supplier cannot invest because the balance sheet will not carry it.

The market paid for the distribution, not the quarter

The operating result missed analyst estimates. The shares rose about 18 percent anyway. That combination is the most informative fact in the release, and it is the one the coverage treated as a curiosity rather than a signal.

When a miss is rewarded, the reward is not for the miss. Investors were pricing the 800 billion yen facility, the split and the net cash position, which together say that the proceeds of scarcity will be distributed rather than reinvested into the capacity that would end it. Guidance pointed the same way: Kioxia forecast second-quarter revenue of 2,390.0 billion yen, about 35 percent above the quarter just reported. A company expecting a bigger quarter and returning capital at the same time is not describing a spike it expects to pass.

What 800 billion yen does not build

Roughly 4.4 billion euro at current rates is real fab money. Whether or not Kioxia would have spent it on capacity, the authorisation places a ceiling on what that cash can now do, and the industry backdrop explains why shareholders find that attractive. TrendForce puts 2026 NAND demand growth at 20 to 22 percent against supply growth of 15 to 17 percent, a gap that does not close on its own.

The timing matters more than the gap. Meaningful new capacity is not expected to arrive before late 2027 or 2028, because fabs are commissioned years before they ship, and 2027 allocations are already being negotiated by the largest buyers. Every quarter that a supplier returns cash instead of committing it pushes the relief date further out. That is the loop: scarcity produces margin, margin is distributed, distribution is rewarded, and nothing in the sequence creates supply.

The question to put to your supplier before 2027

For a European buyer specifying storage now, the practical consequence is that enterprise SSD and memory pricing should be budgeted as a level, not a spike. Contract prices set on scarcity assumptions do not revert when a single quarter looks unusual, and the residual open market absorbs whatever the contracted volume does not.

Change what you read in a supplier's release. The useful line is no longer the margin or the capex language, both of which are written to be reassuring. It is the ratio of capital returned to capital committed. Ask your distributor or vendor what their supplier authorised in buybacks this year against what it committed to new capacity, and price your 2027 refresh on the answer rather than on a forecast of correction.

Frequently asked questions

Does a share buyback actually reduce future memory supply?

Not directly, and it would be wrong to claim Kioxia cancelled a fab. A repurchase authorisation is a ceiling on cash that may be returned, not a stated decision against investment. What it does is reveal priority. A company holding net cash, guiding to a larger quarter and authorising up to 800 billion yen of repurchases is allocating the proceeds of a shortage toward shareholders at the moment when the alternative use would be capacity.

Why did the shares rise if the quarter missed estimates?

Because the market was pricing the capital return rather than the operating line. The buyback facility, the three-for-one split and the net cash position all landed on the same day as the miss. When a distribution is large enough, investors will accept a weaker profit number, and that trade tells you what the equity is now valued on.

Should a buyer wait for prices to fall before purchasing storage?

The published supply and demand figures do not support waiting. TrendForce estimates 2026 demand growth of 20 to 22 percent against supply growth of 15 to 17 percent, and meaningful capacity is not expected before late 2027 or 2028. Waiting is a bet that the gap closes early, placed against suppliers who are currently distributing the proceeds of it not closing.

The number that should travel from this release is not the 75 percent margin, which every outlet led with, and not the 415 percent revenue growth. It is 800 billion yen, authorised for shares rather than silicon, in the same filing set as guidance for a bigger quarter. A shortage that pays its suppliers this well, and that they are rewarded for distributing rather than solving, is not a cycle waiting to turn. Budget the storage line at the level it is, and put the question about capital returned versus capital committed to every vendor you are about to sign with.

KioxiaNAND flashMemory pricesShare buybackEnterprise SSDStorage procurementCapital allocation

IBM's Worst Day Ever Is a Memory Price Warning

IBM lost roughly a quarter of its value in one session. Its CEO says clients moved quarterly budget into servers, storage and memory to beat price rises. That is a warning about your 2027 budget.

5 min read

Consumer Tech

AI Memory Costs Now Set Your Device Budget

DRAM and NAND could climb about 130 percent in 2026, lifting PC prices near 17 percent and phones near 13 percent. The AI data center is now in your hardware refresh line.

3 min read

Law & Regulation

The EA Buyout Now Hinges on Brussels

The largest take-private in history has cleared EA shareholders and US antitrust. The last real gate is the EU - and not classic merger review, but the Foreign Subsidies Regulation, facing its first blockbuster test.