KIOXIA reiterates Overweight: AI inference SSD product line expansion, with NAND supply-demand and cash flow providing support

Overview

This report is Morgan Stanley's follow-up call note after KIOXIA Holdings' June 2 investor day. It focuses on AI inference-driven NAND/SSD demand, KIOXIA's SSD portfolio, the BiCS FLASH technology roadmap, long-term agreements, capital expenditure, and capital allocation, while reiterating its Overweight rating.

Core views

The core view is that KIOXIA's growth drivers are shifting further away from traditional consumer storage toward AI inference, data center, and enterprise SSDs. The company plans to use the CM Series, GP Series, and LC Series to address needs ranging from high-bandwidth KV cache and ultra-high-IOPS direct GPU connectivity to large-capacity QLC storage servers. Meanwhile, tight NAND supply-demand, resilient ASP, higher LTA coverage, and improved free cash flow are expected to support medium-term earnings and valuation.

Analysis framework

The analysis is primarily based on key points from the company's post-investor-day call, the Morgan Stanley ModelWare framework, TechInsights' NAND EB CAGR forecasts, the company's product roadmap, capex plans, and rating/valuation framework. The focus is not on reviewing a single quarter's results, but on assessing medium-term investment implications from the angles of products, demand, supply, agreement coverage, and capital allocation.

Methodology notes

• Valuation methodsFCF yield

  Free cash flow yield valuation

  The report argues that, combined with a strong free cash flow outlook, shareholder return policy, and AI inference growth potential, an approximately 10% free cash flow yield in FY3/28e is sufficient to support the share price; based on the FY3/28 EPS forecast, the implied target-price P/E is about 11x.

• Industry supply-demandNAND supply-demand framework

  NAND supply-demand framework

  The report focuses on how factors such as growth in AI end-demand, supply constraints, ASP durability, and capacity expansion by Chinese manufacturers may affect the NAND cycle.

• Research modelMorgan Stanley ModelWare

  Morgan Stanley internal research model framework

  The report states that, unless otherwise noted, all metrics are based on the Morgan Stanley ModelWare framework; some metrics are Morgan Stanley Research estimates or GAAP-based approximations.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• KIOXIA Holdings (285A.T)

  Core covered company

  Strengths

  Expansion of AI inference SSD product lines; progress in the BiCS technology roadmap; rising share of data center and enterprise revenue; improved free cash flow and shareholder return policy.

  Weaknesses

  The NAND industry is highly cyclical; capital expenditure needs are high; product specifications may change as AI systems evolve.

  Comparison

  The report emphasizes that KIOXIA is increasing its exposure to data center/enterprise and higher value-added SSDs relative to consumer storage.

  Risks

  Deterioration in NAND supply-demand, weakening end-demand, capacity expansion by Chinese manufacturers, and yen appreciation.

• NAND flash industry chain

  Core industry exposure

  Strengths

  AI inference, data center, and enterprise SSD demand may drive EB growth and support ASP.

  Weaknesses

  Supply-demand balance is sensitive, and price forecasting is difficult.

  Comparison

  The report cites a forecast of 22% EB CAGR for 2025–2028 and believes actual demand could be stronger.

  Risks

  Weaker end-demand, overly rapid supply release, and intensified competition.

• AI inference storage

  Key growth application

  Strengths

  Rising demand for KV cache, high-throughput low-latency, direct GPU connectivity, and large-capacity server storage.

  Weaknesses

  Application demand is still evolving rapidly, and product specifications remain uncertain.

  Comparison

  The report notes that market focus is shifting from QLC high-capacity SSDs toward TLC high-bandwidth, low-latency KV cache applications.

  Risks

  Delays in customer qualification, changes in system architecture, and competition from alternative solutions.

Key data

• Report date2026-06-08The report cover page time is June 8, 2026 03:33 AM GMT.
• RatingOverweightThe chart annotation states “we reiterate OW”.
• Target price¥5,500The target price history chart shows the most recent target price record as 10/1/25:5500.
• Target share of data center and enterprise revenueIncrease from 30–40% to above 60% over the medium termThe company plans to increase the mix of higher value-added products and enterprise applications.
• NAND EB CAGR22% in 2025–2028KIOXIA cited TechInsights forecasts at its investor day, and these are broadly consistent with the company's own view.
• Average annual capex plan for FY3/27–FY3/29¥470bnAbove FY3/27 guidance of ¥450bn, with the three-year investment cycle expected to accelerate from 2H FY3/27.
• BiCS-8 output share targetAbout 80% of GB output by end-March 2027BiCS-8 is ramping mass production across multiple projects.
• BiCS-10 layer count332 layersSamples are expected to begin shipping around this summer, with mass production about one year later depending on market conditions.
• CY27 LTA coverage ratioMore than 50% of shipment volumeDiscussions are mainly with data center and enterprise customers.
• CY28 LTA coverage targetAbout 50% of shipment volumeThe company wants to lock in demand while retaining flexibility for changes in product specifications.
• GP Series performance targetMore than 100M IOPSTargeted at NVIDIA Storage-Next and direct GPU connectivity applications, with a second-generation product capable of 100M IOPS planned for launch in 2027.
• LC Series capacity245TBUsing BiCS-8 2Tb QLC, targeting large-scale data storage servers.

Impact & implications

If AI inference systems continue to drive demand for high-throughput, low-latency, and high-capacity SSDs, KIOXIA stands to benefit through product mix upgrades, enterprise-customer LTAs, progress along its technology roadmap, and expanded capital expenditure. For investors, the key implication is that the company is not only a play on the NAND cycle rebound, but is also positioning itself as a supplier of AI storage infrastructure. However, this thesis still depends on NAND supply-demand conditions, ASP durability, customer qualification progress, and capex returns.

Risks

• Improvement in NAND flash supply-demand may fall short of expectations, or worsening supply-demand due to weaker end-demand may last longer than expected.
• Capacity expansion by Chinese manufacturers may intensify NAND supply pressure.
• Yen appreciation would reduce operating profit; the report estimates that every 1-yen appreciation against the US dollar lowers annual OP by about ¥6bn.
• Changes in AI system demand may alter customer requirements for SSD products and specifications, affecting LTA coverage and product mix.
• A high capex cycle may bring investment return and cash flow execution risks.
• Morgan Stanley has investment banking business and shareholding-related disclosures involving KIOXIA Holdings; investors should treat this report as only one reference factor.

What to watch

• 2026 shipment progress for the CM Series targeting NVIDIA KV cache applications.
• 2026 shipment of the first-generation GP Series SLC product and progress toward the 2027 second-generation 100M IOPS product.
• Qualification and mass-production shipment timing for the LC Series at customers such as Dell Technologies PowerEdge R7725xd.
• The ramp-up of BiCS-8 toward about 80% of GB output by end-March 2027.
• BiCS-10 sample shipments, customer qualification, and the plan for mass production about one year later.
• CY27 and CY28 LTA coverage ratios, customer mix, and changes in terms.
• Whether NAND ASP remains resilient in 2H26 and 2027.
• The execution pace of the average ¥470bn capex plan for FY3/27–FY3/29 and its impact on free cash flow.