Top Japanese Chip Stocks to Watch Going Into Q3 Earnings, According to Bernstein

Bernstein’s Japanese semiconductor-stock calls ahead of September-quarter earnings are mixed: it expects Advantest to raise full-year operating profit guidance to ¥900bn and sees potential for Tokyo Electron to beat expectations of about ¥1tn. Kokusai may also raise guidance on strong memory orders, while Screen faces the clearest expected earnings miss; other calls hinge on guidance, pricing, capacity and AI-related demand. Bernstein cautions that high expectations could limit positive share-price reactions, including for companies with likely beats or raised outlooks.

Analysis

The better signal is dispersion in guidance quality, not a blanket AI-equipment call. Strong TSMC/ASML demand does not guarantee positive reactions: where investors already price in AI exposure, the incremental driver is guidance, margin conversion and evidence that capacity can be delivered. Advantest’s potential full-year operating-profit guidance of ¥900bn is a meaningful upside trigger but is not a firmly held consensus view; the key validation is whether CPU/CPO testing adds to growth beyond ASIC expectations. Kokusai’s orders across China and other memory customers support a separate memory-cycle read-through, though the earlier guidance reset may limit surprise.

Screen is the clearest downside setup in Bernstein’s data, but installation-based revenue recognition and limited regional data make timing a material risk; a delayed China DRAM contribution could defer, not eliminate, the miss. Tokyo Electron’s pricing upside appears less differentiated if already reflected in expectations, leaving supply-chain capacity and the 2027 outlook as the less-priced catalysts. Hoya offers a potentially cleaner, less equipment-dependent way to express AI infrastructure exposure through HDD substrates and mask blanks, but Bernstein itself expects only modest revisions and its revenue regression is based on a short, volatile sample.

Near term, earnings reactions may punish “good” results that merely meet elevated expectations. Over 1–3 months, pricing negotiations and guidance revisions should separate genuine margin upside from order momentum. Over 6–18 months, capacity execution and memory/HDD demand matter more than a single-quarter beat. The contrarian risk is treating Bernstein’s directional calls as independent evidence: most depend on noisy leading indicators, and the article provides no valuation or consensus-positioning data to establish how much is priced in.

Key Decisions for Investors

• Prefer a post-results relative-value expression: long Hoya versus short Screen, sized modestly and only after confirming Hoya’s sequential growth/stable-margin outlook and a Screen earnings or guidance shortfall. This pairs a potentially broader AI/HDD demand beneficiary against the article’s clearest equipment downside signal; reduce or exit if Screen’s reported weakness is only installation timing and its forward guidance remains intact.
• Treat Advantest as a catalyst watch, not an automatic pre-print long. Add only if full-year operating-profit guidance approaches or exceeds ¥900bn and management supports incremental CPU/CPO testing demand; falsify the thesis if the guidance raise is absent or commentary confirms ASIC is doing nearly all the work.
• Kokusai is a conditional memory-cycle long if orders from both China and non-China customers translate into a guidance increase and credible capacity/pricing commentary. Do not extrapolate order strength into earnings without that conversion; a flat outlook or weaker pricing would negate the setup.
• Avoid chasing Lasertec solely on anticipated order guidance and be selective with Tokyo Electron: high expectations and already-communicated pricing upside create event risk. Reassess after the prints using supply-chain delivery capacity and the 2027 equipment outlook; missing evidence on consensus positioning and valuation is a reason to keep exposure limited.