Abstract composition of semiconductor wafers on dark concrete reflecting orange and red light in a moody industrial atmosphere
News

Kioxia's Market Value Halved in Six Weeks — The AI Chip Boom's First Real Stress Test

Kioxia's market cap halved from its June 2026 peak as investors question whether AI chip demand is sustainable. A 660% YTD surge, 74% operating margins, and a 12% single-day drop — the AI chip boom's first stress test.

KioxiaNAND FlashAI ChipsJapanStock Market

Kioxia Holdings, the Japanese NAND flash memory chipmaker that became Japan’s most valuable company at its June peak, has seen its market capitalization halve in six weeks. The stock surged 660% year-to-date by June 3, driven by AI data center demand for NAND storage. By July 16, it had given back half of that gain — a brutal correction that signals the AI chip boom is entering its first real stress test.

🔍 THE BOTTOM LINE

Kioxia’s crash is not a company-specific problem. It is the market repricing the entire AI hardware supply chain. When the most overvalued link in the chain breaks, every other link gets re-examined. The AI infrastructure spending boom is hitting the wall that every commodity boom hits: the moment when supply catches up with demand expectations.

From Japan’s Most Valuable Company to Half Its Value

The trajectory is remarkable. Kioxia’s shares surged to a 660% year-to-date high on June 3, driven by record demand for NAND chips in AI data centers and a constrained supply environment. The company’s full-year 2025 revenue reached 2.337 trillion yen, with operating profit rising 93% year-on-year. The April-June 2026 quarter was projected to deliver a 74% operating margin — a number that would be extraordinary for any semiconductor company, let alone a memory maker.

Then the reversal. On June 26, Kioxia shares fell 12% in a single day — the largest single-day drop since its listing — as AI-related stocks sold off broadly. The trigger was reportedly reports that OpenAI was considering delaying its IPO until 2027, which spooked investors who had been pricing AI infrastructure stocks on the assumption of a continuous capital flow into the sector.

By July 16, the market capitalization had dropped to half of its June high. As Bloomberg reported, this was “growing concerns the AI-driven rally in the sector has gone too far.”

Record Profits, Collapsing Stock

The paradox is what makes this story matter. Kioxia is not a company in trouble. It is a company with record financials and a collapsing stock price. The fundamentals:

  • FY2025 revenue: 2.337 trillion yen
  • Operating profit growth: 93% YoY
  • Projected Q2 2026 operating margin: 74%
  • Forward P/E ratio: 7.9x

A forward P/E of 7.9x for a company growing profit at 93% year-on-year is, by traditional valuation metrics, deeply undervalued. As Invezz noted, the stock is “breaking down technically” — below its 50-day EMA, underperforming the Nikkei, and the entire memory complex is de-risking. The fundamentals say buy. The chart says sell. When that divergence happens at scale, it usually means the market is pricing something the fundamentals do not capture.

The Cyclical Memory Trap

What the fundamentals do not capture is the cyclical nature of the NAND flash industry. Memory chips are not GPUs. The NAND market has historically moved in brutal boom-bust cycles: supply tightens, prices spike, manufacturers expand capacity, supply floods the market, prices crash. The AI data center demand narrative broke the cycle — until it did not.

Investor concerns center on two risks:

  1. Oversupply. NAND manufacturers across South Korea, Japan, and China have been expanding capacity in response to AI data center demand. If supply growth outpaces demand growth — even slightly — the pricing power that drove Kioxia’s 660% surge evaporates.

  2. Demand sustainability. AI data center buildouts are being financed by massive debt and equity raises — Oracle’s credit downgrade, IBM’s 25% plunge, the broader AI spending reevaluation. If the hyperscalers slow their infrastructure spending, NAND demand growth slows with it.

The market is not saying Kioxia’s business is bad. The market is saying the expectations embedded in a 660% YTD surge were unsustainable, and the correction is happening fast because the rally was fast.

The Broader AI Chip Selloff

Kioxia is the canary, not the mine. The selloff is hitting the entire AI hardware supply chain:

  • ASML — despite raising its 2026 forecast to €45 billion and planning 30% more EUV machines, the stock has been under pressure
  • Samsung1800% profit jump on AI chip demand, but facing the same cyclical questions
  • SK Hynix — the HBM leader, but ADR listings have not insulated it from the selloff
  • NVIDIA — the bellwether, where all AI chip roads lead

The pattern across all of these is the same: record profits, surging demand, and stocks that have already priced in the next three years of growth. When the next three years of growth look even slightly uncertain, the stocks correct hard.

NZ Angle

New Zealand has no direct exposure to the NAND cycle, but the selloff matters for the sovereign AI build conversation. The argument for sovereign AI infrastructure has partly been that AI compute demand is insatiable — justifying public investment in GPU capacity. Kioxia’s halving is a reminder that “insatiable” is a narrative, not a law of physics. If the AI infrastructure spending cycle cools — even temporarily — the economics of sovereign AI compute change. The case for NZ-owned compute should be built on sovereignty and control, not on assumptions about perpetual demand growth.

❓ FAQ

Is Kioxia going bust? No. The company is profitable, growing, and has a forward P/E of 7.9x. The stock crash is a valuation correction, not a solvency crisis. Kioxia is also planning a US ADR listing to expand its investor base.

Why did the stock fall 12% in one day? The June 26 drop was triggered by reports that OpenAI might delay its IPO until 2027, which spooked AI infrastructure investors broadly. Kioxia, as the most overvalued AI chip stock, took the biggest hit.

Should I buy the dip? That is a question for a financial adviser, not an AI news site. The bull case: 7.9x forward P/E, 93% profit growth, 74% operating margins. The bear case: NAND is cyclical, supply is expanding, and AI infrastructure spending may be peaking. Both cases are real.

Does this affect GPU prices? Indirectly. NAND and HBM are different markets, but they share the same end-demand — AI data centers. If data center buildouts slow, both NAND and GPU demand cool together. Kioxia’s selloff is an early signal, not a direct GPU price indicator.

🔍 THE BOTTOM LINE

Kioxia’s halving is the AI chip boom’s first genuine stress test. Record profits and a halving stock price coexisting is the market’s way of saying: the fundamentals are real, but the expectations were not. The AI hardware supply chain is not broken — it is repricing. And repricing is how booms become sustainable, or how they end.

📰 Sources

Sources: Bloomberg, AInvest, Invezz, Reuters