cd /news/artificial-intelligence/kioxia-s-31-fold-profit-surge-still-… · home topics artificial-intelligence article
[ARTICLE · art-81517] src=startupfortune.com ↗ pub= topic=artificial-intelligence verified=true sentiment=· neutral

Kioxia's 31-fold profit surge still wasn't enough to satisfy Wall Street

Kioxia Holdings reported a 46-fold surge in profit attributable to owners of parent to ¥842.2 billion for the June quarter, with revenue rising to ¥1.767 trillion, but shares fell 18% on July 28 amid concerns about Chinese memory maker ChangXin Memory Technologies' market debut and potential supply increases. The company announced a 3-for-1 stock split and a ¥800 billion share buyback to address investor concerns.

read4 min views1 publishedJul 31, 2026
Kioxia's 31-fold profit surge still wasn't enough to satisfy Wall Street
Image: Startupfortune (auto-discovered)

Kioxia's June-quarter profit was enormous, but you shouldn't read it as a clean victory lap. The stock is being judged against Chinese capacity, not only today's income statement.

A profit jump from ¥18.3 billion to ¥842.2 billion in one year should be enough to quiet almost any room. It didn't. Kioxia Holdings reported on July 31 that revenue for the three months ended June 30 rose to ¥1.767 trillion, operating profit reached ¥1.270 trillion, and profit attributable to owners of parent rose roughly 46-fold from the same quarter last year. Those are official company figures. They are also why the market reaction around Kioxia feels so sharp: investors are no longer asking whether AI servers are buying NAND. They are asking how long this pricing lasts.

The filing gives you the real story without much drama. SSD and storage revenue reached ¥1.175 trillion, up ¥957.3 billion from a year earlier, while smart devices brought in ¥525.7 billion. Kioxia said the jump came mainly from higher average selling prices, stronger bit shipments and a weaker yen, with demand from data-center customers focused on generative AI doing the heavy lifting. That's not a weak quarter. It's an extraordinary one.

But markets don't pay you twice for a number they already expected. In May, Reuters reported that Kioxia had forecast April-June net profit of ¥869 billion, backed by demand from AI investment by U.S. technology companies. The actual profit attributable to owners of parent came in at ¥842.2 billion. Close, but lower. When a stock has already been treated as one of the cleanest ways to buy the AI storage boom, close can still feel like a miss.

The China scare is real #

The bigger issue arrived before Kioxia's results. ChangXin Memory Technologies, China's leading DRAM maker, began trading on Shanghai's STAR Market on July 27. The Associated Press reported that CXMT's shares surged 466% on debut and gave the company a market value of more than $487 billion, making it the most valuable company listed on a mainland Chinese exchange. South China Morning Post put the valuation at 3.31 trillion yuan, or about $489 billion.

Look carefully at that. CXMT is a DRAM company, not a direct NAND twin of Kioxia. Still, investors rarely wait for perfect overlap when fear enters the trade. Memory is cyclical. Chinese capacity is politically backed. And once the market starts imagining cheaper domestic supply arriving faster than expected, every company that benefited from tight supply gets re-priced.

That is what happened on July 28. Kabutan reported that Kioxia briefly hit its daily limit-down level at ¥44,550, down ¥10,000 on the day, and said the stock had fallen about 60% from its June 22 intraday high of ¥112,700. Business Insider also reported a broad Asia chip selloff, with Kioxia down 18% and Samsung and SK Hynix sliding alongside it. SNDK fell sharply in the same wave. The selling wasn't only about Kioxia. Kioxia was simply standing in the wrong place when the memory trade turned.

The draft version of this story leaned too hard on a simple idea: Kioxia missed analyst expectations and therefore the AI memory cycle is losing altitude. That's too neat. The cleaner reading is harsher and more useful. Kioxia can post record numbers and still face a lower multiple if investors think Chinese supply will break pricing discipline.

The buyback buys time #

Kioxia's board clearly understood the need to answer shareholders. On July 31, the company approved a 3-for-1 stock split with an October 1 effective date. It also set up a treasury share acquisition facility of up to ¥800 billion, covering as many as 30 million shares before the split, with purchases scheduled from August 3 through October 30.

Those moves are not empty. The split should lower the investment unit and make the stock easier for individual investors to buy. The buyback is more forceful: ¥800 billion is a serious sum, and Kioxia said the program was designed to improve capital efficiency and shareholder returns while taking account of market conditions and future investment capacity.

But here's the thing: a buyback can support a stock. It can't set NAND prices. Kioxia's own filing says the flash memory industry remains highly volatile over short periods, which is the most useful sentence in the whole release. The company is making exceptional money because AI data-center demand has tightened supply and pushed prices up. If that balance holds, today's selloff will look overdone. If Chinese memory output changes the supply curve, today's profit may be remembered as the top of the cycle.

You don't need to pretend Kioxia is in trouble to see why investors are nervous. A company earning ¥1.270 trillion of operating profit in three months is not fragile. That's not the question. But a memory company trading on perfect AI demand, tight supply and disciplined pricing is exposed when any one of those assumptions starts to wobble.

For now, Kioxia has the numbers. CXMT has the momentum story. The next six months will tell you which one the market decides to pay for. Also read: China's homegrown DUV lithography machine goes into production and ASML loses 8% in a dayRevolut is turning ChatGPT Go into a bank account perk for 75 million customersA federal judge says the Trump administration still hasn't proven its case for restricting Anthropic

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @kioxia holdings 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/kioxia-s-31-fold-pro…] indexed:0 read:4min 2026-07-31 ·