Not so long ago, Kioxia was a corporate orphan. Carved out of a scandal-scarred and near-bankrupt Toshiba and sold to a private-equity consortium, the maker of NAND flash-memory chips limped onto the Tokyo Stock Exchange in December 2024, worth a modest ¥800bn. In the fiscal year of its listing, it eked out an operating profit of just ¥442bn, having lost money the year before. Eighteen months on, it is barely recognisable. The artificial-intelligence boom has left data centres desperate for storage, chip prices have soared, and Kioxia’s shares have climbed more than fiftyfold — briefly carrying its market value past Toyota’s — to make it, by some measures, the most valuable company in Japan. Analysts now reckon its operating profit could approach ¥7trn this fiscal year, some eight times last year’s and enough to crown it the country’s most profitable company. Not bad for a business nobody wanted.1
Kioxia’s giddy ascent is the most spectacular instance of a force that has remade the Japanese stock market. The AI investment theme has become its organising principle, reaching into corners of the economy that nobody thought to look — and turning castoffs into champions along the way.
This piece attempts three things. It traces the mark the AI investment theme has left on Japan’s market and its economy; it asks how the theme has evolved and where it might travel next; and, because no boom runs forever, it points to the signals that might warn of a pause. We should declare an interest: we are participants rather than bystanders, and the Jupiter Japan strategy holds a number of the companies named below. Along the way we will meet a battery firm reborn as a power-supply business, a maker of animal-health medicines turned semiconductor-materials star, and a leasing company building Japan’s first mega-casino that owes its recent form to an inspired bet on chips — for the most revealing thing about Japan’s AI boom is the company it keeps.
A market remade
Begin with the money. In the first half of 2026 foreign investors bought a net ¥9.7trn2 of Japanese shares — the largest half-yearly haul on record, and the culmination of a spree that has carried the market to a succession of all-time highs. Foreigners, long sceptical of a market synonymous with deflation and dormant governance, have returned with conviction. Most of that money, Nikkei reports, chased perceived AI winners. Alongside the obvious candidates, such as the chipmaking-equipment giant Tokyo Electron, sat less obvious ones: Fujikura, whose fibre-optic products for data centres are selling as fast as it can make them, and even Ajinomoto, a maker of seasonings that has turned its food chemistry into a commanding share of the insulating films used in advanced chips. “Japan possesses advanced technology in AI hardware, and the base of companies benefiting is broad,” Masashi Akutsu of BofA Securities told Nikkei.3 Breadth, as we shall see, is the operative word.
There is a deeper reason to suspect the enthusiasm is more than a passing fashion, and it concerns people — or the shortage of them. Japan’s chronic want of workers, the product of a shrinking and ageing population, has long acted as a brake on growth. The chart below plots the country’s employment-conditions index with its axis inverted, so that worsening scarcity registers as a climb; by this gauge labour is nearly as tight as it has ever been. In most rich economies, labour-replacing automation is a political headache and a threat to jobs. In Japan it is closer to a rescue: AI that performs the work of people the country simply does not have fills a vacuum rather than emptying offices. Automation has, in truth, been a Japanese speciality for decades, born of the same necessity; AI is merely its cleverer and hungrier successor. That is why the theme rests on firmer economic ground here than almost anywhere else.
Such logic complicates the fashionable habit of sorting the market into AI “winners” and “losers”. Enterprise-IT, software and business-services firms have been cast as casualties — and for the makers of simple chatbots, or the battalions that test and debug code, the label may fit. But the picture is muddier than the sell-off implies. NEC reports that AI adoption is creating demand rather than destroying it, and the notion that Japan’s corporates will abruptly bring software development in-house looks fanciful when, as NEC’s collaboration with Anthropic concedes, most organisations still face “a shortage of IT talent, insufficient accumulation of operational know-how [and] stringent security requirements”. Here the labour argument bites once more: information services and services-for-business rank among the most short-staffed industries in the entire economy. A company that cannot fill its own desks is not the most obvious candidate to be automated out of existence.4
The macroeconomic stakes are correspondingly large. If AI can lift the productivity of an economy whose workforce is shrinking by hundreds of thousands a year, it does more than flatter share prices — it addresses the single greatest constraint on Japanese growth. That is a prize no other big rich economy, with the possible exception of a similarly greying South Korea or Italy, can claim in quite the same terms.
The theme evolves
The first act of Japan’s AI story was written by the direct beneficiaries of the prodigious spending of the hyperscalers’ (Alphabet, Microsoft, Amazon, Meta and Oracle). The scale is hard to overstate, and Japan has been one of its principal foreign beneficiaries. The combined capital expenditure of these big American cloud operators has climbed towards a third of their sales, and Japan’s optical-cable makers, memory manufacturers such as Kioxia and chip-equipment firms such as Kokusai Electric have furnished many of the picks and shovels. Our strategy holds several of these names; doing so has been less a flourish of conviction than a condition of keeping pace with the market they have largely propelled.
Hyperscalers’ capex
The more adventurous money is now hunting the theme’s stranger beneficiaries — the sort that, as The Economist observed in May, are quietly minting money from AI. Take Panasonic, a fund holding up more than 120% this year to an all-time high: the battery lines it originally built to power Tesla’s electric cars are increasingly given over to the backup power units that AI data centres crave, and analysts reckon AI-related businesses could furnish 30% of group operating profit by March 2029. Or consider Nissan Chemical, another holding better known for its veterinary anti-parasite medicines, which now expects semiconductor materials to deliver 45% of operating profit by 2030. Stranger still is Orix, a leasing house presently building Japan’s first mega-casino, which has lately won admirers for an entirely unrelated reason: an inspired stake in Toshiba that hands it exposure to Kioxia, the memory-maker with which we began. The theme, in short, is broadening from the firms that plainly make AI to those that merely — and lucratively — enable it.5
Where next? Several paths beckon. The first is plumbing. As AI reaches the masses, the twin problems of power and cooling loom large: the International Energy Agency expects data-centre electricity demand to roughly double over the coming decade, with America’s data centres alone projected to swallow 9.3% of the country’s power by 2030, against 2.9% in Europe and 2.3% in China.6 That is a windfall for whoever manages the distribution of power and the dissipation of heat. The second path is “physical AI” — the application of the technology to industrial processes — widely tipped as the next great phase, and one for which Japan is well equipped. Our strategy holds Keyence, a sensor-maker (physical AI, like a person, needs good senses), and Minebea Mitsumi, in bearings and motors. Deloitte finds that 71% of Asia-Pacific firms already make at least minimal use of physical AI, ahead of the Americas and Europe, and expects that share to reach 90%8; Yaskawa Electric, for its part, says its new AI-equipped robot plant has doubled productivity.8 Realistically, the surface has barely been scratched. A third path is that the network operators carrying all this traffic — which PWC expects to roughly double between 2025 and 20319 — at last monetise their increasingly critical pipes; the funds own the fixed-line major NTT with that, plus a potential capability in quantum computing, in mind.
Data centre electricity demand forecast
A caveat threads through all of this. PWC’s map of the AI value chain suggests that the zone of highest value will gradually migrate from the infrastructure layer — the semiconductors, cloud and data centres in which Japan excels — towards the user-facing applications at the far end of the chain.10 That has never been Japan’s forte; a country that surrendered the smartphone to others may find the interface layer heavier going. Japan’s genius lies in the hardware and precision manufacturing that underpin the early and middle innings of the build-out. Whether it can compete in the consumer-facing endgame is, for now, an open question.
Reasons to watch the exits
Which brings us to the third act. After a run this exuberant, prudence demands a glance at what might interrupt it. Two risks stand out. The first is that the hyperscalers have simply overbuilt. On current trends their capital expenditure — swelling at roughly 70% a year — is set to overtake their operating cash flow, which is growing at some 23%, by the end of 2026. Capex financed from borrowings and investors’ goodwill is a more fragile creature than capex funded out of profits, and Japan’s suppliers are geared to its continuation. It is worth remembering that Japan’s AI winners are, for the most part, suppliers rather than developers — their fortunes rise and fall with someone else’s budget. Should the American cloud giants conclude that they have built enough, the picks-and-shovels trade would be the first to feel it.
The second risk wears a Chinese face. China’s frontier AI models are now within a whisker of their American rivals on raw capability — Jefferies puts the gap at a mere 8%, with Alibaba’s Qwen scoring 92% against the leading American model’s 100% — and they are startlingly cheap, priced at roughly a fifth of the American equivalent.11 Cheaper, nearly-as-clever models that run on markedly less hardware would undercut the very premise on which much of Japan’s AI trade rests: that the world’s appetite for expensive silicon is limitless. A market that needs fewer chips per unit of intelligence is not the one for which Japan’s exporters have been priced.
None of this amounts to a prophecy of doom. The labour shortage that makes AI so welcome in Japan is not about to reverse; the sheer breadth of the country’s beneficiaries — from memory-makers and chip-equipment champions to battery firms and casino-owning leasing houses — lends the theme a resilience that a narrower boom would lack; and the phases still to come, in power, physical AI and the network layer, play to Japan’s manufacturing strengths. Yet a theme that has propelled the market to record highs is, by definition, one in which a great deal of good news is already in the price. For now, the balance of probability still favours the optimists, but the margin is slimmer than the drumbeat of record highs would suggest. We remain invested and, frankly, fascinated. We are also, as ever, keeping a weather eye on the exits.
Footnotes
1Kioxia Holdings, results for the fiscal year ended March 2026 and first-quarter FY2026 outlook, 15 May 2026; ‘Kioxia overtakes Toyota as Japan’s most valuable listed company’, Nikkei Asia, June 2026.
2Foreign investors scoop up half-year record $60bn in Japan stocks’ Nikkei Asia, July 2026.
3Foreign investors scoop up half-year record $60bn in Japan stocks’ Nikkei Asia, July 2026.
4NEC, ‘NEC Announces Strategic Collaboration with Anthropic Focused on Enterprise AI’, press release, 2026; employment-by-industry data from CLSA and the Bank of Japan, cited in Benthos, ‘AI: No country for young men’, 26 June 2026.
5 ‘The strange Japanese companies minting money from AI’, The Economist, May 2026; company disclosures — Orix participated in the Japan Industrial Partners-led consortium that took Toshiba private in 2023, and Toshiba retains a substantial stake in Kioxia.
6 International Energy Agency data-centre electricity projections, cited in Mizuho Securities Equity Research, June 2026.
7 Deloitte, ‘State of AI’ report, January 2026.
8 ‘Japan eyes AI-powered comeback in factory robot race with China, Europe’, Nikkei Asia, July 2026.
9 PwC, ‘AI Adoption in Japan’ report, April 2026.
10PwC, ‘AI Adoption in Japan’ report, April 2026.
11Jefferies, research note, June 2026.
Strategy risks:
- Currency (FX) Risk - The Strategy can be exposed to different currencies and movements in foreign exchange rates can cause the value of investments to fall as well as rise.
- Pricing Risk - Price movements in financial assets mean the value of assets can fall as well as rise, with this risk typically amplified in more volatile market conditions.
- Market Concentration Risk (Geographical Region/Country) - Investing in a particular country or geographic region can cause the value of this investment to rise or fall more relative to investments whose focus is spread more globally in nature.
- Derivative risk - the Strategy may use derivatives to reduce costs and/or the overall risk of the Strategy (this is also known as Efficient Portfolio Management or "EPM"). Derivatives involve a level of risk, however, for EPM they should not increase the overall riskiness of the Strategy.
- Counterparty Default Risk - The risk of losses due to the default of a counterparty on a derivatives contract or a custodian that is safeguarding the strategy's assets.
- Charges from capital - Some or all of the Strategy’s charges are taken from capital. Should there not be sufficient capital growth in the Strategy this may cause capital erosion.
- Liquidity Risk (general) - During difficult market conditions there may not be enough investors to buy and sell certain investments. This may have an impact on the value of the Strategy.
- ESG - Investments are selected or excluded on both financial and non-financial criteria. The Fund's performance may differ from the broader market or other strategies that do not utilize ESG criteria when selecting investments.
- ESG and Sustainability - Investments are selected or excluded on both financial and non-financial criteria. The Strategy's performance may differ from the broader market or other strategies that do not utilise ESG/Sustainability criteria when selecting investments.
For a more detailed explanation of risk factors, please refer to the "Risk Factors" section of the Prospectus or of the Scheme Particulars.
Important information
The funds have not been registered under the United States Investment Company Act of 1940, as amended, nor the United States Securities Act of 1933, as amended. None of the shares may be offered or sold, directly or indirectly in the United States or to any US Person, unless the securities are registered under the Act, or an exemption from the registration requirements of the Act is available. A US Person is defined as (a) any individual who is a citizen or resident of the United States for federal income tax purposes; (b) a corporation, partnership or other entity created or organized under the laws of or existing in the United States; (c) an estate or trust the income of which is subject to United States federal income tax regardless of whether such income is effectively connected with a United States trade or business.
Issued by Jupiter Asset Management Limited, registered address: The Zig Zag Building, 70 Victoria Street, London, SW1E 6SQ is authorised and regulated by the Financial Conduct Authority. Issued in the EU by Jupiter Asset Management International S.A. (JAMI), registered address: 5, Rue Heienhaff, Senningerberg L-1736, Luxembourg which is authorised and regulated by the Commission de Surveillance du Secteur Financier.
No part of this presentation may be reproduced in any manner without the prior permission of JAM and JAMI.