Musha called the cycle dead Thursday. 981 yen of the 1,802 yen rebound came from two stocks. By Friday's SOX -10%, the argument was on BOJ's desk.
On Thursday morning, June 12, Japanese strategist Ryoji Musha declared the cycle dead. "AI has ended semiconductor cyclicality. Nvidia's operating margin jumped from ~15% to 60–65%, and this pricing power creates a permanent hyper-profit zone." By the same day's close, the Nikkei 225 ended +1,802 yen (about +2.8%). The end-of-cycle thesis was printed on top of the cycle's most violent single-session rebound.
Two names sit on the receipt. Advantest (6857) and Tokyo Electron (8035) — two stocks alone delivered 981 yen of the 1,802-yen move, or 54.4%. If AI permanently killed the cycle, why does half the index sit in a tester and an etch-house?
Replay the tape. The Nikkei printed an all-time high of 68,786 on June 3. Over the next eight trading days it lost -9.4% to a 62,335 low on June 11, after a hot US payroll report pushed the SOX down more than 10% in a week and rolled back Fed cut bets. Musha's declaration was printed the day after that -9.4% — the rebound day. Then Friday, June 13, the Nikkei gave back -0.9% on the week as a fresh SOX -10% drag took out Japan's AI names again.
In one week: -9.4%, +2.8%, -0.9%, and an end-of-cyclicality call. If this is what a market without cycles looks like, the definition has been stretched.
One stat does sit on Musha's side. Kioxia (285A) overtook Toyota on June 12 to become Japan's largest listed company at ¥44 trillion market cap. On Kabutan's year-to-date excess-return list, Kioxia leads with a forecast +784% recurring profit growth, swamping its PER. What Japan is pricing here is not a quarterly EPS print; it's the persistence of a multi-cycle operating margin. Musha's thesis asks whether that persistence holds.
The support beam runs through the supply chain. Shin-Etsu Chemical (4063) is building a new rare-earth separation-and-refining facility in Fukui — eighteen years after Japan last did this domestically, the backward integration cuts China dependence for motors, HDDs, and packaging materials. Toppan is using AI to discover next-generation FC-BGA and glass substrate materials. The materials layer (Shin-Etsu, Toppan, TOK 4186) is the most direct lever on Musha's 60% margin claim — if AI compute pricing power is permanent, so is the toll the materials layer collects on top. The inverse: if Musha is wrong, the materials toll is the first thing to cyclicalize.
The conversion problem is unsolved. NHK's morning business segment on June 12 walked through three hurdles still blocking Japan's mass-production transition. Rapidus signed an MOU the same day with a UK semiconductor center to develop customers — while TSMC has Nvidia and Apple, and Samsung has Qualcomm, Rapidus is sourcing prospective customers through UK academia. Margin theses do not yet apply to a 2nm foundry without utilization.
Spot pricing partially supports Musha. DDR5 16Gb spot printed $44.83 on June 14, evidence the memory cycle is in its up-phase. Kioxia's +784% forecast is the meta-version of that spot signal. But spot prices are not the absence of cycle — they're the signal that the cycle is currently going up. If Musha is right, this price should not come down. Three decades of memory cycle averages teach a different lesson.
Today and tomorrow decide. The BOJ meets June 15–16, the FOMC June 16–17. A BOJ hike strengthens the yen, compresses Advantest and Tokyo Electron's dollar-revenue translation, and a Nikkei in which two stocks deliver half its moves will print the cycle again. A hawkish FOMC sends the SOX back toward its June 13 -10%. Kioxia's +784% forecast is FX-independent on its face, but the market accepting it as a normalized multi-cycle margin requires the cycle to not return — and the tape this week says the cycle is alive.
Bernstein's Tuesday call naming its top three Japanese semi picks is itself a quiet refutation of Musha. Buying "beta in bulk" is what you do when cyclicality is dead; naming three specific stocks is what you do when cyclicality remains the differentiator. The sell-side, by its act, answers Musha.
For positioning, Advantest (6857) sits at the intersection of both theses. If Musha is right, Advantest's tester-monopoly margin does not cyclicalize. If June 11's -9.4% and June 13's -0.9% are right, Advantest remains the single largest variance contributor in the Nikkei. The market has not chosen between them. It only printed the receipt. BOJ and FOMC settle the receipt this week.
Key Sources: - Musha: AI Price Revolution Births Hyper-Profitable Economic Sphere, Bull Just Starting (Kabutan, 2026-06-12) - Sugimura: AI/Semis Remain Market Leaders; Kioxia, Advantest, SoftBank Are Picks (Kabutan, 2026-06-14) - Nikkei jumps 1,802 yen as semis surge; Advantest and TEL alone add 981 yen (Google News, 2026-06-12) - Nikkei Slides 0.9% as SOX -10% Hits AI/Chip Names (Kabutan, 2026-06-13) - Kioxia tops Nikkei excess-return list; high PER but 784% profit growth forecast (Kabutan, 2026-06-12) - plus 18 more
If this analysis was helpful · ☕ Support Us · ✈️ Telegram