News Cutting Room

10月28日 朝の材料｜半導体は反発、資金の値段は別問題——Fed待ちの東京市場｜Chips Rebound. Capital Still Has a Price.

What changed At 15:30 JST on 28 October 2026 , after the Tokyo cash close, Advantest released its fiscal 2026 first-half results and revised full-year outlook. First

TokyoBrainie

— 7 min read
Share

English · 简体中文 · 日本語

What changed

At 15:30 JST on 28 October 2026, after the Tokyo cash close, Advantest released its fiscal 2026 first-half results and revised full-year outlook. First-half revenue was JPY 845.2 billion, up 60.5% year on year. Operating income was JPY 456.8 billion, up 96.5%, and net income was JPY 368.1 billion, up 116.8%.

The more important change for the next session is guidance. Advantest raised its full-year revenue forecast to JPY 1.863 trillion from JPY 1.714 trillion and operating-income guidance to JPY 967 billion from JPY 846 billion. Net-income guidance rose to JPY 741.5 billion from JPY 660 billion. The company said AI-related SoC production volume and complexity were increasing test demand, while data-centre investment supported DRAM and non-volatile-memory testers. It expects the semiconductor tester market to reach a new record in calendar 2026 as inference AI expands.

Advantest also authorised a market repurchase of up to 11 million shares, or 1.5% of shares outstanding excluding treasury stock, for as much as JPY 280 billion. The purchase period runs from 29 October 2026 to 24 March 2027. The ceiling is not proof that the full amount will be spent; actual monthly purchases must be checked.

The stock rose in after-hours PTS trading after the announcement. This is an initial price response, not the next Tokyo session's closing verdict.

Previous and updated judgment

Previous judgment at 12:24 JST: SK hynix had provided earnings and long-term-contract evidence for HBM and server-memory demand. Advantest still needed to confirm that this demand was reaching Japanese tester revenue, margins and guidance.

Updated judgment: that confirmation arrived. Advantest's results and raised forecast materially strengthen the earnings case for semiconductor testing and adjacent advanced-packaging demand. The buyback adds a company-specific demand source for the shares.

This does not validate a broad market re-rating. The Nikkei closed 685.82 points, or 1.08%, higher at 63,954.19, but the TOPIX fell 0.14%. On the Prime Market, 431 stocks rose, 1,092 fell and 25 were unchanged. Advantest alone added approximately 230.85 points to the Nikkei. The index gain therefore remained concentrated even as the company-specific earnings evidence improved.

Rates and FX

The BOJ's official 17:00 JST USD/JPY snapshot was 157.78–157.80, after a 157.76–158.03 intraday range. There was no confirmed intervention. The two-year JGB yield rose about 0.5 basis point to 1.880% and the 10-year yield about 2 basis points to 2.805% in Reuters' market update. These are modest moves ahead of the FOMC and BOJ meetings, not a new policy decision.

Tokyo transmission

The strongest next-session read-through is Advantest itself, followed by HBM testing, advanced packaging and suppliers whose customer mix is genuinely exposed to high-performance SoC and memory test demand. Tokyo Electron, Disco, SCREEN, Ibiden and Kioxia may benefit from the demand narrative, but their revenue bridges are not identical.

The broad market remains more fragile than the Nikkei headline suggests. If yields rise after the FOMC while market breadth stays weak, high-multiple semiconductor gains can coexist with losses in domestic, defensive and rate-sensitive sectors.

Counterevidence and next checks

  1. The 29 October Tokyo open and close: whether Advantest holds the PTS gain and whether related names outperform beyond the opening auction.
  2. The FOMC statement at 03:00 JST and press conference at 03:30 JST: whether US yields move decisively above 5%.
  3. Microsoft and Alphabet results after the US close: cloud growth, capex, depreciation and free cash flow.
  4. Advantest's order conversion, customer concentration, tester mix and ability to deliver the second-half revenue plan.
  5. Actual share-repurchase execution versus the JPY 280 billion ceiling.
  6. Whether TOPIX breadth improves. Another Nikkei rise with most Prime stocks falling would confirm index concentration, not broad risk appetite.
  7. USD/JPY near 158 and any new Ministry of Finance language or confirmed intervention.

The strengthened tester-demand thesis would be weakened if the next-session stock response fades, the raised forecast depends mainly on currency rather than volume and mix, or hyperscaler cash generation fails to keep pace with capital spending.

Sources

What changed

SK hynix released its preliminary third-quarter results on 28 October. The company did not state an exact release time on the English newsroom page; Reuters reported the results at approximately 08:50 JST. Revenue reached KRW 104.6155 trillion, up 328% year on year, and operating profit reached KRW 83.3338 trillion, up 632%, with an 80% operating margin. These extraordinary figures are preliminary and remain subject to the independent audit process, as the company itself notes.

The evidence is broader than one quarterly headline. SK hynix said HBM4 shipments have begun in earnest, expects HBM4 to exceed HBM3E in both volume and revenue in the fourth quarter, and has secured long-term agreements covering more than 50% of total 2027–2028 production capacity. It expects demand to exceed supply through 2030. Cash and cash equivalents stood at KRW 114.5 trillion, borrowings at KRW 15.8 trillion and net cash at KRW 98.7 trillion.

The company also announced approximately KRW 100 trillion of shareholder returns for 2026 and said most gains from its Kioxia investment would be shared with shareholders. This is SK hynix's capital-allocation decision; it is not a distribution to Kioxia shareholders.

Previous and updated judgment

Previous judgment: the semiconductor rebound had physical-deployment and financing catalysts, but it still required evidence that end demand could produce earnings and cash.

Updated judgment: HBM and server-memory demand now have a new, company-reported earnings and contracted-capacity confirmation. This strengthens the sector-specific demand case for memory, testing and advanced packaging. It does not prove that every AI infrastructure project earns an adequate return or that Japanese equipment companies receive equal benefits.

Tokyo price confirmation—and its limit

At 11:00 JST, the Nikkei was 848.33 points, or 1.34%, higher at 64,116.70. Advantest, Tokyo Electron, SoftBank Group, Kioxia and Ibiden together contributed approximately 816 points, or about 96% of the index rise, based on the published contribution figures. This was a real semiconductor-led advance, but not proof of a broad market re-rating.

Kioxia was approximately 7.9% higher in a delayed public quote. The move can reflect the memory cycle, index-related demand and positioning; price action alone cannot identify one cause. Advantest's results are due after the Tokyo close, so its pre-earnings rise is not confirmation of the company's own guidance.

USD/JPY was 157.81 in Reuters' morning update and the US 10-year yield was 4.9967%. Earnings improved; capital did not become cheap.

Tokyo transmission

The strongest channel is HBM testing, advanced packaging and memory manufacturing. Advantest, Tokyo Electron, Disco, SCREEN, Kioxia, Ibiden and semiconductor-material suppliers are research targets, but actual customer mix, order conversion and margins determine the earnings impact.

The memory evidence is less direct for optical cable, power equipment and general data-centre construction. Those names still require project-level orders and profitability.

Counterevidence and next checks

  1. Whether the semiconductor leaders hold their gains through the close, not only 11:00 JST.
  2. Advantest's post-close orders, guidance, tester mix and customer concentration.
  3. Whether Kioxia strength persists after separating HBM optimism, NAND fundamentals and index flows.
  4. The FOMC statement at 03:00 JST on 29 October and the 03:30 press conference.
  5. Microsoft and Alphabet results, especially cloud growth, capex, depreciation and free cash flow.
  6. Whether long-term supply agreements are take-or-pay, price-indexed or cancellable; the company has not fully disclosed those terms.

The strengthened demand judgment would be weakened if Advantest guidance fails to validate tester demand, Japanese semiconductor leaders surrender the morning move, or hyperscaler cash generation fails to keep pace with investment.

Sources

A semiconductor rebound is not cheaper capital

This is a Tokyo-session framework for 28 October, based on the completed US session of 27 October. The shared data table below distinguishes closing prices from newsletter snapshots. Premarket futures are not closing returns; after-hours moves are not regular-session returns.

Reuters’ Fed preview expects a pause, not an assured end to tightening. Its equipment report records stronger core shipments. Its bond report highlights AI risk and borrowing pressure; Mike Dolan’s commentary examines volatility transmission rather than declaring AI obsolete. These are different questions, not one universal “AI bubble” verdict. See sources 1–5.

My interpretation: demand can be real while the price of financing remains uncomfortable. A supplier books revenue when equipment is delivered under its contract; a cloud operator must subsequently turn installed capacity into paid usage and cash. More investment can support manufacturers while reducing the purchaser’s free cash flow. There is no contradiction.

Core capital goods are a broad business-investment proxy, not an AI-only ledger. Nominal orders do not establish unit-volume growth, margins or investment returns. A stronger shipment reading and an order reading below economists’ expectations can coexist. “Strong” and “a positive surprise” are different tests.

Tokyo transmission

Semiconductor equipment and testing are the first sentiment channel: Tokyo Electron and Advantest are research examples, not interchangeable exposures. Examine customer investment plans, product mix and order conversion. A strong chip index does not guarantee an equal rise in every Japanese supplier.

Fujikura, Furukawa Electric and Sumitomo Electric require an additional bridge: actual optical, cable and power-network orders, production capacity and margins. A server headline alone does not quantify their incremental earnings.

Lower oil can ease prospective fuel costs for airlines and other users, while weighing on upstream earnings expectations such as INPEX’s. Exchange rates, hedging, purchase timing and contract pricing determine how much reaches profit. One lower futures quotation does not erase delivery or insurance costs.

Banks do not simply benefit from every increase in yields. Lending margins must be assessed alongside funding costs, securities valuation and credit losses. For exporters, a broadly stronger dollar does not by itself establish a weaker yen: check USD/JPY directly.

What would change this assessment?

Constructive: semiconductor leadership persists beyond the opening gap, more stocks participate, and corporate guidance supports conversion from investment into cash.

Fragile: an opening rally fades while yields or credit spreads rise; financing pressure can overwhelm a sound industrial demand story.

Adverse: higher oil and higher yields coincide with weaker earnings guidance. That would pressure both costs and valuation.

These are conditional scenarios, not probabilities or a prediction of today’s closing index. Compare trading volume with the same time of day and observe whether opening gains hold. Price action alone cannot prove institutional buying or short covering.

The FOMC meets on 27–28 October US time. Microsoft and Alphabet earnings are scheduled after the US close on 28 October, corresponding to the morning of 29 October in Japan; Amazon follows after the US close on 29 October, Japan’s 30 October morning. These are upcoming events for this edition, not known results. Check cloud growth, capex, depreciation, cash flow and funding terms together.

Technology’s usefulness, the return on its infrastructure and the price of its securities deserve separate judgments. Today’s rebound answers only the last of those, for one session.

Sources and related research are shared below.

Original Notion archive

Read more