Nvidia Turmoil Ripples Through Japanese AI Stocks as Valuations Shift from Premium to Discount

nashnova research
今天发布阅读约 10 分钟

Jensen Huang's departure triggered a global AI selloff; Japan's Kioxia and Advantest have each dropped over 9%, as investors shift from paying a premium for AI exposure to demanding a discount — a systemic re-rating is under way.

01

Why are Japan's AI stocks falling harder than the broader market?

Kioxia Holdings and Advantest have each dropped more than 9% since the announcement, versus a 2.9% decline in the Topix.
This means → the selloff in AI-linked names is running at triple the pace of the broad market — capital is exiting AI exposure specifically.
Nomura analyst Yoshitaka Suda frames it as a "triple hit": deteriorating earnings outlook + post-midterm regulatory tightening fears + Nvidia's own leadership upheaval.
02

How did "AI premium" become "AI discount"?

Mitsubishi UFJ fund manager Hiroaki Tomori puts it bluntly: stocks used to get a valuation boost for being AI-linked — now AI linkage itself triggers a markdown.
In plain terms = investors once paid *more* for anything touching AI; now they demand to pay *less*, because AI exposure now signals greater uncertainty, not guaranteed upside.
Tomori adds that large-cap AI names still trading at elevated premiums face further selling risk, while stocks that have already corrected substantially may be nearing a floor.
03

Where do valuations stand now?

The forward P/E ratios — share price divided by expected earnings over the next 12 months — for the TSE Non-Ferrous Metals Index and the MSCI Japan Semiconductor Index have both fallen below their late-March levels, the point when the AI rally accelerated.
This means → both indices now trade at valuation multiples below the Topix overall; the AI sector has gone from "more expensive than the market" to "cheaper than the market."
This reflects a market already pricing in a cooling AI narrative, not just a short-term sentiment shock.
04

Kioxia at under 2× P/E — is it actually cheap?

Kioxia's forward P/E sits at less than 2×. Fund managers broadly view that as "cheap."
But Tomori warns: that valuation rests on still-optimistic earnings forecasts. If AI infrastructure spending slows and profit estimates are cut, today's "cheap" may prove illusory.
In plain terms = P/E has a denominator — expected profit. If profit expectations are slashed, the same share price produces a higher P/E, and the apparent bargain disappears.
05

Have analysts revised their earnings forecasts?

So far, analysts have largely held their estimates. Bloomberg data shows the six major U.S. tech companies are expected to spend $885 billion in capex this year, jumping 64% to roughly $1.45 trillion next year.
Nvidia's earnings this week included a quarterly revenue outlook above Wall Street expectations, signaling that demand remains firm.
This means → the critical test is whether current earnings forecasts for Japan's chip stocks already reflect a scenario where AI capex slows — if they don't, the valuation adjustment may not be over.
06

How much could data-center spending fall?

Tomori says a 20% to 30% decline in data-center investment "cannot be ruled out."
This means → if that scenario materializes, Japanese chip stocks reliant on AI infrastructure orders — especially memory (Kioxia) and test equipment (Advantest) — face significant earnings downgrades.
This reflects the market's core unresolved gap: share prices have already fallen, but earnings forecasts haven't caught up — the distance between the two is the remaining downside risk.

市场有风险,内容仅供研究参考,不构成投资建议。