Nanya Technology (2408.TW) — 2026Q3 FY2026 Earnings Call Analysis

60% Legacy Mix, AI Story Four Years Out

Cash flow confirms the beat. The revenue base is 60% on a product customers are designing out.

Thesis: Nanya is printing real cash on a DRAM shortage driven by DDR4 end-of-life allocation. That shortage is genuine in the near term but 60%+ of revenue sits on a product with structurally declining demand (12-14% of market and shrinking). The AI segment at 20% and targeting 40-50% by 2029 cannot replace the revenue base when DDR4 pricing normalizes. Capacity is capped through 2028. This is a cash-generative peak cycle on a legacy product, not a structural re-rating.

Verdict: HOLD — Conviction: MEDIUM

Catalyst: DDR4 pricing inflection in the 2027 calendar year when PC and consumer demand down-specs accelerate. Management themselves said demand 'will continue to get smaller' while capacity cannot grow substantially until 2H 2028 equipment moves in.

Key Risk: DDR4 pricing holds longer than modeled while DDR5 conversion accelerates faster than expected, expanding the addressable base. The Q4 capex step (~42B NT) colliding with committed dividend also risks a payout cut if cash tightens.

The Tell: Management volunteered: 'the key indicator, key index that we have to pay attention to is the long term, the AI investment, is it the return of those investment is sustainable?' This was unprompted in response to a supply-demand question. Management is naming the exact risk the market is underwriting: that hyperscaler ROI drives the whole shortage thesis.

Detected Patterns

AI Label Without Substance: AI revenue >20% includes DDR4, LPDDR4 for system management, storage, enterprise SSD. Management confirmed 'it's a mix of everything, including DDR4, DDR5, and low-power DDR.' Any DRAM byte near an AI box got counted.
Mix Shift Spin: ASP gains attributed to mix while >60% of product is legacy DDR4/LPDDR4 with stated declining demand trend. DDR5 sits at ~10%. The mix is legacy-heavy, not premium.
High Utilization as Ceiling: 100% LTA fulfillment presented as strength but management admitted non-LTA fulfillment 'is quite low.' Real book is thinner than the 50-70% headline. No organic volume growth through 2028.
Capacity Ceiling: Sold out but cannot expand. Equipment lead times block new supply until 2H 2028. Management stated output growth in 2027 will be 'tough.' Demand exceeds structural ability to supply by design.
Beat and Raise Machine: Operating income up NT$46.1B sequentially with 6.3pp margin expansion. December quarter ASPs guided to improve again. Consistent execution against guidance.
Pricing Power Signal: Double-digit ASP growth continuing into December quarter. LTA customers 'very very proactive' about extending terms with price floor discussions. Supply shortage has no near-term solution per management.
Capital Conviction: $10B Pingtung fab committed with equipment moving 2H 2028. NT$16B per 50K wpm fab. Funding from NT$270B net cash with no debt dilution. Backing words with capital.
Backlog Fortress: LTA coverage 50-70% of output, three-year contracts >20% of production, 100% fulfillment rate on signed contracts with both-side liability. Customers proactively extending terms.

Friction Level: HIGH_FRICTION — Both sides agree on the mix (DDR4/LPDDR4 >60%), the cash flow, and the capacity cap. The disagreement is whether DDR4 EOL pricing into a shortage is a durable pricing regime or a one-time allocation spike on a shrinking demand base. Structural shortage vs. cyclical peak on a legacy part.

Report not found

The report data is no longer available. Please return to the archive.

60% Legacy Mix, AI Story Four Years Out | Silicon Signal