JPMorgan Raises Kioxia Earnings Forecast but Cuts Target Price to ¥130,000
Miles Bennett
JPMorgan kept its Overweight rating on Kioxia and raised three-year profit forecasts, yet cut the target price from ¥155,000 to ¥130,000 — better earnings, lower valuation, because the market has not seen long-term agreements actually stabilize results.
Earnings raised — where is the upside coming from?
The main driver is enterprise SSD (eSSD) pricing beating expectations — the price recovery in Q2 2026 was stronger than the market had assumed.
Three-year operating profit raised across the board: FY2027 ¥800.2 bn (+7.9%), FY2028 ¥1,124.1 bn (+6.4%), FY2029 ¥1,456.2 bn (+4.6%).
This means → JPMorgan is betting that rising eSSD penetration can offset weakness in consumer NAND — the flash memory in phones and USB drives. It projects the eSSD addressable market will exceed 820 EB next year, growing roughly 55% year-on-year.
Better earnings — so why cut the target price?
The core reason is a deliberate compression of the valuation multiple: the target P/E drops from roughly 11× to roughly 9×, back to the fifteen-year global memory-sector average.
The previous premium assumed that long-term agreements (LTAs — multi-year supply contracts with customers) would smooth pricing and stabilize profits. JPMorgan now says the market "needs to see stability show up in actual results before it will pay for it."
In plain terms = Kioxia promised LTAs would lock in steadier margins, but the proof is not yet on the table — the market will not pay a premium for a promise. The earliest checkpoint arrives in 2027.
What exactly do LTAs need to prove?
Kioxia plans to introduce multi-year LTAs for the first time starting in 2027, targeting 50% coverage by 2028.
JPMorgan flags two dimensions of credibility: enforceability — whether penalty clauses actually bind customers — and pricing smoothing — whether LTAs genuinely reduce price volatility. Past cases where LTAs failed to deliver keep the market cautious.
This reflects a perennial memory-industry problem: cycles are violent, and customers may refuse to honor contracts when spot prices plunge — so the market defaults to "show me first."
An ¥800 billion buyback — what signal does it send?
Kioxia announced a buyback of up to ¥800 bn, with an execution window of just three months (Aug 3 – Oct 30). JPMorgan calls it "the first large-scale buyback in the memory ecosystem."
Free-cash-flow yield — cash earned minus required capital spending — is projected at 13%, 27%, and 37% for FY2026–FY2028 respectively.
This means → if actual shareholder returns exceed JPMorgan's current assumptions (FCF payout ratios of 10% / 30% / 50%), there is room for additional valuation re-rating. For reference, Samsung and SK Hynix both target 50% of cumulative FCF as shareholder returns.
The GP product line — why can a niche product move the whole supply picture?
Kioxia's GP-series eSSD is built on SLC architecture — single-level cell, storing just 1 bit per cell, which delivers speed but consumes far more capacity — and targets 100 million IOPS (input/output operations per second).
The supply impact is the key: SLC consumes three to four times the capacity of TLC/QLC (multi-level architectures). Even if GP shipments are modest in bit terms, their drain on total NAND supply is disproportionately large.
In plain terms = it is like baking a premium loaf that uses three to four times the flour — you do not sell many loaves, but the flour inventory drops fast, tightening the whole market. JPMorgan lists GP-series qualification progress as a key medium-term catalyst.
Content is for reference only, not financial advice.