After the market close on October 30, 2026, Kioxia Holdings reported revenue of ¥4,247.2 billion and operating profit of ¥3,264.7 billion for the April–September period. For October–December, however, it forecasts revenue of ¥2,500 billion, up just 0.8% from the previous quarter. Reports said the share price fell as much as about 7% in PTS (proprietary trading system) trading after the announcement.
President Hiroo Ota told Bloomberg in an interview on September 8 that "prices have already risen enough," according to a report published the next day. Taken together with the latest revenue outlook, this might suggest that NAND flash memory prices are nearing a ceiling and that Kioxia has begun holding back on price increases. But when the results are broken down into three factors (exchange rates, shipment timing and unit selling prices), they confirm that the momentum of price increases is slowing, without supporting the conclusion that prices have peaked.
A weaker yen impact and shipment timing shifts behind the 0.8% revenue forecast
Kioxia's October–December 2026 forecast calls for revenue of ¥2,500 billion (up 0.8% quarter on quarter), operating profit of ¥1,995 billion (up 0.0%) and profit attributable to owners of the parent of ¥1,360 billion (up 0.3%). Revenue, operating profit and net profit are all projected to be roughly flat from the prior quarter. According to the earnings presentation, the forecast assumes the yen strengthens by about ¥4, from ¥161 to ¥157 per dollar, which would reduce operating profit by about ¥55 billion. The Q&A summary of the earnings briefing explains that the exchange-rate conditions are fixed only for October, while November and December remain undetermined.
The effect of a stronger yen is easiest to see by converting dollar-denominated sales into yen. For example, $100 million in sales is ¥16.1 billion at ¥161 per dollar but ¥15.7 billion at ¥157. Even if dollar sales are unchanged, yen-denominated sales fall by about 2.5%. Because Kioxia discloses unit selling price trends in dollar terms, exchange-rate changes need to be taken into account when looking at yen-based results.
The company estimates that each ¥1 move in the dollar-yen rate affects quarterly revenue by about ¥15 billion and operating profit by about ¥14 billion. In other words, a ¥1 stronger yen reduces operating profit by about ¥14 billion. However, the company notes that this sensitivity is an estimate based on assumed sales volumes for October–December and cannot be applied as is to other quarters.
In fact, while the October–December revenue forecast is up 0.8% in yen terms, converting at each quarter's exchange rate puts the increase at about 3.4% in dollar terms. July–September revenue of ¥2,480 billion at ¥161 per dollar comes to roughly $15.4 billion, while the October–December forecast of ¥2,500 billion at ¥157 comes to roughly $15.9 billion. This is a rough estimate by our editorial staff. The October–December figure is a forecast, and some transactions may in practice be priced in yen. Also, the roughly ¥55 billion exchange-rate impact cited by the company applies to operating profit, so it covers a different base from this calculation, which converts total revenue into dollars.
Another factor holding back revenue growth is the timing of shipments. According to Kioxia, shipments in terms of memory capacity in October–December are expected to be at about the same level as in July–September. The company attributes this to the pulling forward of some product sales into July–September, as well as a rising share of data center and enterprise products, which have longer lead times from order to delivery, causing some shipments to slip into January–March 2027. The company expects shipments in January–March to rise sharply, partly as a rebound from this shift.
In other words, the 0.8% revenue growth rate for October–December is affected by two factors: the roughly ¥4 stronger yen and the shift in shipment timing. However, the disclosed materials do not confirm whether these two factors alone fully explain the small increase in revenue. At the very least, it would be premature to conclude from yen-based revenue growth alone that NAND flash prices have plateaued.
Unit price growth narrows from more than double to the mid-20% range, with a modest rise expected for October–December
The key indicator for gauging price trends is the average selling price (ASP) that Kioxia discloses each quarter. The main metric the company publishes is the "ASP including product mix," which reflects changes in the mix of products sold and is presented as a quarter-on-quarter increase in dollar terms. This measure more than doubled in January–March 2026, then rose about 70% in April–June and in the mid-20% range in July–September, a narrowing pace of increase. For October–December, the company expects a modest rise.
| Period | ASP including product mix (USD, QoQ) | Shipments by memory capacity (QoQ) |
|---|---|---|
| Jan–Mar 2026 | More than doubled | Down about 10% |
| Apr–Jun 2026 | Up about 70% | Up low single digits |
| Jul–Sep 2026 | Up mid-20% range | Up low double digits |
| Oct–Dec 2026 (forecast) | Modest increase | Roughly flat |
These figures confirm that the pace of increase in unit selling prices has narrowed each quarter. The phase of sharp surges, in which unit prices more than doubled in a single quarter, has passed, and the momentum of price increases has clearly slowed. However, because the company discloses only approximate figures such as "more than double" and "mid-20% range," the exact rates of increase are unknown. The decline in shipments in January–March resulted from planned equipment maintenance and sales from inventory, which are factors separate from unit price movements.
Another point to note is that the ASP including product mix and the like-for-like selling price for the same products show different movements. Kioxia also discloses the "like-for-like ASP," which excludes the effect of product mix. In July–September, the ASP including product mix rose in the mid-20% range, while the like-for-like figure rose in the low 30% range. In other words, price increases on a same-product basis were larger than the mix-inclusive average.
This is because changes in the mix of products sold can push down the overall ASP. For example, if the share of products with a lower unit price per unit of memory capacity rises, the growth in the overall ASP will be smaller even if prices of individual products are rising. Therefore, a narrowing increase in the ASP including product mix does not necessarily mean that price increases on the same products have also narrowed.
For October–December as well, Kioxia explains that higher unit selling prices are the main driver of revenue growth. At the same time, it cited a relatively higher share of products for smart devices as the reason the mix-inclusive price increase will be modest. As some shipments of data center and enterprise products shift into January–March, the share of products used in smartphones and similar devices rises. However, the company has not disclosed a like-for-like ASP outlook for October–December.
Even compared with industry-wide forecasts, uncertainty about price trends remains. Research firm TrendForce forecast on September 30 that NAND flash contract prices in October–December 2026 would rise 15–20% from the previous quarter. This exceeds its July forecast for July–September (up 10–15%) and points to accelerating price increases across the market.
By product, TrendForce says price increases for enterprise SSDs are widening from the previous quarter, while PC SSDs show a different trend. Because PC makers are reducing the SSD capacity installed in their main products, memory makers are setting prices flexibly, including by holding back on price increases, according to TrendForce. However, the 15–20% figure is a forecast for industry-wide contract prices and is defined differently from Kioxia's mix-inclusive ASP. Whether Kioxia is holding back price increases, or whether changes in product mix simply make average unit price growth look smaller, cannot be determined at this stage, since the like-for-like outlook has not been disclosed.
Did Kioxia hold back on price increases and miss revenue opportunities?
In his September 8 interview with Bloomberg, President Ota said of NAND flash prices that "prices have already risen enough," and reportedly told the sales department not to seek large price increases from data center operators. The background is said to be concern that soaring prices could dampen appetite for investment in AI. Ota expressed the view that another price jump of about 70%, like that of the previous quarter, is unlikely, saying "if we raise prices too much, we will damage our own market and growth" and "even hyperscalers (large cloud operators) have limited budgets." At the same time, he did not rule out future price increases and indicated that the near-term priority is to maintain the current high price level.
These remarks raise the question of whether Kioxia is holding back on price increases even when demand is strong, and thereby missing out on revenue it could otherwise have earned. In particular, if long-term contracts are signed at low prices, it may not be able to fully reflect later rises in market prices in its profits. To consider this point, the first thing to check is the actual results for July–September.
Kioxia explained that in July–September, both unit selling prices and shipments exceeded the company's outlook as of July. Revenue for the quarter was ¥2,480 billion, up 40.3% from the previous quarter. That exceeds, in a single quarter, the full-year revenue of ¥2,337.6 billion for fiscal 2025 (the year ended March 2026). At least in the July–September results, there is no clear sign of sluggish revenue growth resulting from restraint on price increases.
Compared with industry forecasts as well, Kioxia's unit selling prices rose substantially. The ASP including product mix for July–September rose in the mid-20% range in dollar terms, exceeding the 10–15% rise in NAND contract prices that TrendForce had forecast as of July 3. However, TrendForce's figure was a forecast made at the start of the quarter rather than an actual result, and Kioxia's figure is an internal average that includes product mix, so the two cannot be compared directly. All this comparison shows is that Kioxia's price increase exceeded the industry forecast made at the start of the period; it does not prove that the company was selling at higher prices than the market average.
Moreover, Ota's remarks were reported on September 9, partway through the July–September quarter. Because the company does not separately disclose prices of transactions concluded before and after his remarks, it is unclear how much the president's policy affected actual selling prices. The July–September results show that unit selling prices rose more than the company had assumed, but they provide no basis for verifying the effect of the restraint policy on earnings.
In the Q&A summary of the October 30 earnings briefing as well, the company did not indicate a policy of halting price increases altogether. Kioxia expects the NAND flash supply shortage to continue throughout 2027 and outlined a view that unit selling prices are "basically on an upward trend." However, it said it will watch market developments to see how far prices rise. Combined with the president's September remarks, this can be read as a policy of avoiding sharp price spikes while maintaining high levels and continuing to raise prices as needed.
In judging whether the company is missing revenue opportunities, the price terms of long-term supply agreements (LTAs) are also important. According to Kioxia, its 2027 production capacity is almost fully allocated, and for 2028 it has secured, through multi-year contracts, well over 50% of shipments by memory capacity. It has already begun discussions with customers on supply for 2029–2030, but does not disclose the price terms of these contracts. SK hynix, a peer, has described a mechanism for reflecting market trends within agreed price ranges in its long-term contracts, but it is unclear whether Kioxia adopts similar terms.
What can be confirmed at this point is that unit selling prices in July–September exceeded the company's outlook and rose more than industry forecasts made in July of the same year. The earnings materials offer no clear evidence that holding back on price increases caused the company to miss out on revenue. However, because the price terms of the long-term contracts are undisclosed, it is impossible to judge whether it could have sold at higher prices, or whether it will be able to fully reflect future price rises in its profits.
Tax burden surges while ¥330 billion in capital expenditure is planned for the second half
With the sharp rise in profit in the first half, Kioxia's tax burden has also increased substantially. Income tax expense for April–September 2026 reached ¥1,005.5 billion, about 31.4% of the ¥3,203.4 billion in interim profit before tax (our calculation). However, income taxes actually paid in the same period were only ¥110.7 billion. Income taxes payable at the end of September were ¥974.7 billion, up ¥870.2 billion from the end of March. The company explained that it has secured ample cash on hand to prepare for large tax payments ahead, and because it expects the tax burden to grow in the second half as profit accumulates.
Income tax expense of ¥1,005.5 billion for April–September is about 7.4 times the ¥135.3 billion spent on acquiring property, plant and equipment in the same period. However, income tax expense includes deferred taxes, and the amount actually paid was the ¥110.7 billion mentioned above. Also, the outlay for property, plant and equipment excludes investment in intangible assets and receipt of government subsidies, so it is not appropriate to simply compare the scale of investment and tax burden using these two figures alone. Kioxia expects gross capital expenditure of ¥470 billion for fiscal 2026, of which ¥330 billion is planned for the second half. Therefore, the fact that first-half investment was well below the tax expense cannot be taken to mean that capital spending is insufficient.
On shareholder returns, the company has already carried out large-scale measures. In August, Kioxia completed a ¥800 billion share buyback, and it also resolved an interim dividend with a record date of September 30. The dividend is ¥90 per share, which is ¥30 per share after the 1-for-3 stock split carried out on October 1, for a total of ¥47.937 billion. The effective date is December 8, and it is the company's first dividend since listing.
Regarding additional share buybacks, the company explained that it has put them "on hold" rather than ruling them out, and said it will decide after assessing the second-half investment situation. It also made clear its thinking that, over the medium to long term, depending on the status of strategic investment, it may return 100% of surplus funds to shareholders.
On the financial side, cash on hand has grown significantly on the back of rising profit. Cash and cash equivalents at the end of September were ¥1,397.3 billion, up ¥926.6 billion from the end of March. Meanwhile, bonds and borrowings fell from ¥1,047.6 billion at the end of March to ¥603.9 billion, leaving net cash, as defined by the company, at ¥812.7 billion.
Trade receivables have also increased with the rapid expansion in sales, but their ratio to revenue has not changed significantly. Trade receivables at the end of September were ¥1,648.4 billion, equivalent to 66.5% of July–September revenue. At the end of March, trade receivables were ¥660.6 billion and January–March revenue was ¥1,002.9 billion, for a ratio of 65.9%. Trade receivables have grown at roughly the same rate as sales, and on this ratio alone, receivables do not appear to have risen disproportionately to revenue. However, trade receivables include "other receivables," and changes in customer composition and collection terms have not been disclosed.
Long-term contracts that will shape 2027 prices, and a capacity expansion plan targeting fiscal 2029
Looking ahead at the NAND market, Kioxia continues to expect demand to grow. It forecasts that market-wide growth in memory capacity terms (bit growth) will be in the high teens in percent in 2026 and the low 20s in 2027. It expects its own 2027 shipments to grow in line with the market, assuming that data center and enterprise demand offsets weak demand for smartphones and PCs.
Joint venture partner Sandisk has also reported growth in its data center business. Its data center revenue for the quarter corresponding to July–September was $4.802 billion, up 61% from the previous quarter. Storage demand for AI infrastructure and cloud services has become an important factor supporting growth in the overall NAND market.
On the other hand, expanding supply capacity takes time. Kioxia raised its fiscal 2026 gross capital expenditure outlook from ¥450 billion as of July to ¥470 billion, planning to invest ¥330 billion in the second half. This fiscal year's spending stays within the three-year investment plan, with part of the investment scheduled for next fiscal year being brought forward, and the company says it does not intend to make major changes to the plan as a whole.
As for new production sites, it has begun site preparation for the third fab (K3) at the Kitakami plant, with operations planned to start in fiscal 2029. It is also pursuing, with Sandisk, a plan to invest about ¥5 trillion in Japan by 2032, with production capacity shared between Kioxia and Sandisk at a 60-to-40 ratio. However, both the construction of K3 and the roughly ¥5 trillion investment plan are premised on government support.
There is also debate over whether this scale of investment is sufficient given demand growth. Makoto Kubo, a former Toshiba CFO, commented on President Ota's stated policy of continuing annual investment of ¥400–500 billion that "the investment could stand to be a digit larger" (Bunshun Online, August 9). Kubo's remark was directed not at the setting of selling prices but at the scale of investment needed to address a growing market.
As this demand growth and supply capacity constraint continue, the price terms of long-term contracts will shape Kioxia's earnings. If it can secure well over half of 2028 shipments in memory capacity terms through multi-year contracts, it can expect a certain sales volume even if market conditions fluctuate. On the other hand, future profits could differ greatly depending on whether contract prices are fixed or vary in line with market prices. Because Kioxia does not disclose the price terms of the contracts, it is unclear how much of the current supply-demand tightness will be reflected in future profits.
There are three main things to watch. First, how much the like-for-like ASP rose in the October–December results. Second, whether the sharp increase in shipments that the company expects for January–March 2027 actually occurs. And third, how much will become clear about how prices are set in long-term contracts.
If like-for-like unit selling prices continue to rise and shipments are confirmed to increase in January–March, that would be evidence that President Ota's policy of maintaining high price levels while avoiding sharp increases and the company's outlook that prices will keep rising in 2027 are compatible in terms of results.
The October–December revenue forecast is roughly flat in yen terms, but converted at the assumed exchange rates it represents growth of about 3.4% in dollar terms. It is certain that the pace of unit price increases is narrowing, but that alone does not mean NAND flash prices have reached a ceiling. To judge how Kioxia's price restraint policy will affect actual earnings, we will need to check future like-for-like selling prices and the price terms of long-term contracts.