Kioxia's Undervaluation Narrative Meets Earnings and Index Rebalancing

Analysis

The OP builds a strongly bullish narrative around Japanese memory manufacturer Kioxia (Tokyo ticker 285A): AI infrastructure development is still in its early stages, demand for NAND and enterprise SSDs will continue to grow, and expanding earnings combined with TOPIX passive fund demand could drive a valuation rerating. The OP cites a $172 billion market capitalization, full-year revenue “guidance” of $60 billion to $65 billion, and a forward P/E ratio near 3, calling it the world's most undervalued company. However, no earnings forecast model, valuation year, exchange rate, or price target is provided; the OP only says they will continue holding, without disclosing share count, cost basis, or options positions. The central catalyst is the second-quarter earnings release coinciding with the first phase of the TOPIX adjustment on October 30, 2026, alongside expectations for additional share buybacks. The official website does schedule the earnings release for 18:45 Japan time that day, but this does not establish that it was originally scheduled for November 5–6, nor can the release timing be interpreted as management signaling an earnings beat. (kioxia-holdings.com) Four images supplement the technology, supply-demand, scale, and profitability narratives. The first shows the world record achieved with Linus Media Group using Kioxia NVMe SSDs to calculate pi to 300 trillion digits. An official announcement supports this event, but it demonstrates storage performance in a specific computing task, not that AI has broadly solved mathematical and physical problems. (kioxia.com) The second media screenshot claims NAND capacity is sold out through 2027 and that the TOPIX free-float weighting will increase in two steps. The third chart, labeled Bloomberg, compares market capitalizations at the end of 2025 and on April 23, 2026, showing Kioxia rising from approximately 6 trillion to approximately 19 trillion yen and entering Japan's top ten. However, this historical screenshot is not evidence of its current market capitalization, much less a profitability ranking. The fourth image is a financial ranking without column headings that places Kioxia first and displays $23.13 billion and $8.04 billion. The OP uses it to reinforce the claim that Kioxia is “Japan's most profitable company,” but it is impossible to determine whether these columns represent trailing revenue, operating profit, or net profit. Official June-quarter revenue was 1.7671 trillion yen, operating profit was 1.2700 trillion yen, and net profit attributable to owners of the parent was 842.2 billion yen. These figures indicate a significant improvement in operations but are insufficient to confirm the screenshot's dollar-denominated figures or nationwide ranking. (ssl4.eir-parts.net) On the index side, JPX confirms that the free-float factor will be adjusted to 0.325 and 0.5 on October 30 and November 30, respectively, but does not provide the $18.9 billion inflow claimed in the title. The factor is also not the same as the stock's weight in TOPIX, and ongoing passive allocation cannot be treated as permanent one-way inflows. (jpx.co.jp) The most important gap is the unexplained valuation denominator: a $172 billion market capitalization at a P/E of 3 implies approximately $57.3 billion in annual net profit and a net margin of roughly 88% to 96% against the claimed revenue, which revenue scale alone cannot support. The company explicitly states that it does not provide a full-year operating plan, making the “full-year guidance” label problematic. (ssl4.eir-parts.net) The author's implicit assumptions are that AI demand, scarcity-driven pricing, and high profit margins will persist over the long term, that index demand has not already been priced in, and that additional buybacks will materialize. The post neither tests these assumptions nor develops a valuation scenario for a cyclical downturn.

Discussion Summary

There are 11 comments, but little substantive discussion and no community consensus yet. One commenter raises concerns about the direction of the sideways trading following the earlier steep rally and the risk of a drawdown, but offers no quantitative rebuttal. Another thread confuses Samsung's nationality; the OP clarifies that it is a South Korean company and attributes its earnings miss to overly high analyst estimates, without providing evidence. The remaining discussion mainly concerns Linus's visit video; nobody examines the P/E of 3, revenue guidance, or inflow calculations.

Fact Check

• questionable: Kioxia's full-year revenue guidance is $60 billion to $65 billion. — The official June-quarter earnings report explicitly states that, because of significant short-term fluctuations in the memory industry, the company does not provide a full-year operating plan or forecasts of progress against it; the disclosed outlook covers the September quarter and the first half. No basis was found for treating this dollar range as management's full-year guidance.
• questionable: The market capitalization is $172 billion, with a forward P/E ratio near 3. — No official market capitalization or full-year earnings forecast basis matching the timing of the post was obtained. Using the post's figures, a P/E of 3 requires approximately $57.3 billion in annual net profit, implying a net margin of roughly 88% to 96%. The post does not specify the forecast year, accounting basis, or stock-split adjustments, so the valuation cannot be reproduced.
• unclear: Kioxia has now become Japan's most profitable company. — The screenshot lacks column headings, a source, and a complete basis for comparison. Official June-quarter operating profit was 1.2700 trillion yen, and net profit attributable to owners of the parent was 842.2 billion yen. Actual earnings growth alone cannot establish that the company ranks first in profit among Japanese companies, nor can it confirm the meaning of the screenshot's $8.04 billion column.
• supported: The second-quarter earnings release and TOPIX adjustment both occur on October 30, 2026. — The company's official website schedules the second-quarter earnings release for October 30 at 18:45 Japan time. JPX confirms that the first-phase free-float factor adjustment takes effect on the last business day of October, which is October 30. This supports only that they occur on the same day, not that they occur simultaneously during trading hours.
• unclear: The company moved its earnings release forward from November 5–6 to October 30, signaling strong results. — The official website confirms the current date as October 30, but no official record was found showing that it was previously scheduled for November 5–6. The connection between an earlier release and an earnings beat is the author's speculation.
• questionable: The two-phase TOPIX adjustment will bring $18.9 billion in inflows and ongoing passive investment. — JPX supports the two-phase adjustment: a factor of 0.325 at the end of October and 0.5 at the end of November, but the announcement contains no $18.9 billion estimate. The post lacks assumptions about tracking assets, incremental index weight, exchange rates, and execution, so neither the amount of funds nor sustained net inflows can be confirmed.
• unclear: NAND capacity is sold out through 2027. — The attached media screenshot makes this claim, but the official earnings report obtained for this review confirms only AI data center demand and growth in average selling prices. It does not confirm the scope, contract duration, or binding nature of the claim that all capacity is sold out through 2027.
• unclear: More share buybacks are expected in the future. — The official website lists a July 31, 2026 buyback arrangement and an August 10 buyback completion announcement, supporting that the company previously conducted buybacks. No commitment to additional, as-yet-unannounced buybacks was found; future buybacks remain the author's expectation.
• supported: Kioxia and Linus Media Group set a world record for calculating pi using NVMe SSDs in 2025. — The official announcement dated May 19, 2025 confirms that pi was calculated to 300 trillion digits and certified by Guinness World Records, using Kioxia SSDs with a total capacity of 2.2 PB. This is not proof that AI has solved mathematical or physical problems.