Beyond AI: Where Japan's Next Leg of Gains Leads Into Year-End

by Kenichi Ogura - Former Chief Editor of PRESIDENT (Japan’s No.1 business magazine)

For overseas investors, the bigger change in Tokyo is not weaker AI demand. It is a Japanese market learning to live with expensive energy, a less predictable yen and bond yields near 3%.

The easiest mistake a foreign investor can make in Japan this autumn is to treat Tokyo as a cheaper version of the U.S. market. The same AI boom is visible in both places, but it reaches corporate profits in a different way. Japan does not have its own Microsoft or Meta. What it has are companies that sell the equipment, testers, cables and materials used to build the AI infrastructure boom. Tokyo Electron Limited <8035>, ADVANTEST CORPORATION<6857> and Fujikura Ltd. <5803> have all benefited from that spending.

The business case has not disappeared. Data-center investment is still growing. Yet the stock-market argument is getting harder because something else has changed in Japan: money is no longer almost free.

Japan's 10-year government bond yield touched 2.945% on August 18, the highest level since 1996. To an American investor, a yield near 3% looks ordinary. In Japan it is not ordinary at all. For most of the careers of today's senior executives and portfolio managers, Japan was the country of zero rates, negative rates and a central bank that bought enormous amounts of government debt.

A generation of Japanese valuations was built inside that world.

Now the Bank of Japan's policy rate is 1%, and the bank has kept open the option of raising it again. Long-term yields have already moved ahead of the BOJ. The practical result is simple: investors can demand more from a stock before accepting a high valuation. That matters most where expectations are highest. A semiconductor supplier can report excellent orders and still fall if investors decide that those orders were already in the price. This is where the Japanese version of the AI trade becomes more interesting than the usual "AI up or AI down" debate.

Japan sells the picks and shovels

The first phase of the AI boom was unusually easy to read. U.S. technology companies raised capital-expenditure plans; chipmakers needed more capacity; Japanese equipment and component suppliers benefited. The connection was visible and, for a while, almost mechanical.

It is not quite so mechanical anymore. Reuters, using LSEG consensus estimates, found that capital spending at Microsoft, Alphabet, Amazon, Meta and Oracle is expected to rise much faster than operating cash flow through 2027. The exact totals are less useful than the message behind them. Investors have stopped treating every extra dollar of AI spending as proof of future value.

For Japanese suppliers, that changes the market test. Orders still matter, but so does the quality of the customer spending behind those orders. If the hyperscalers turn AI into faster cloud growth, higher prices and stickier customers, their suppliers have a long runway. If the spending starts to look defensive - money spent because no technology company can afford to fall behind - the same order book can receive a lower valuation.

There is no need to predict an AI crash to see the risk. The market can simply become less generous.

This point is especially important for overseas investors because Japan's technology winners often sit farther upstream than familiar U.S. names. Tokyo Electron makes semiconductor production equipment. ADVANTEST is a major testing-equipment supplier. Fujikura has gained from the fiber and wiring needs of data centers. Their customers' customers may ultimately decide how much growth is left. The companies are different, and their exposures should not be bundled into one "Japan AI" trade. That shortcut worked better when capital was cheap and spending was accelerating everywhere.

Three percent means something different here

Interest rates are not the only part of Japan that looks different once an overseas investor gets below the headline numbers.

Consider oil. Brent crude closed above $94 a barrel on August 21. In the United States, expensive oil transfers income toward a large domestic energy industry as well as hurting consumers. Japan has almost no such cushion. Its energy self-sufficiency rate was only 15.3% in fiscal 2023, the lowest among the G7, and the country still relies heavily on imported fossil fuels.

An oil shock therefore arrives in Japan largely as an import bill. It reaches company margins, household purchasing power and the trade balance. It can also keep inflation high enough to make the BOJ more cautious about stopping rate increases. The same barrel of oil can hit a Japanese stock through its costs and, later, through the valuation multiple investors are prepared to pay.

The yen complicates things further, although not in the simple way it is often described overseas.

A weak yen has helped many large exporters because overseas earnings translate into more yen. A stronger yen reverses some of that benefit. Japan, however, is also a huge importer of fuel and raw materials, so a stronger currency lowers costs elsewhere TOYOTA MOTOR CORPORATION <7203> and an airline do not experience the same yen.

Currency moves in Japan can therefore produce quite violent sector rotations without saying much about the underlying strength of the economy. The move worth watching is not necessarily the jump after a BOJ meeting. It is the one that forces companies to change the exchange-rate assumptions buried in their earnings guidance.

There is also a calendar issue that is easy to miss outside Japan. Many large Japanese companies close their fiscal year in March rather than December. Their October-November results are half-year reports, and management teams often use them to reconsider full-year forecasts for the year ending the following March.

This autumn's reporting season should tell investors whether the summer's higher oil, higher yields and currency volatility have actually reached profits.

So far, the earnings picture is better than the mood in the bond market might suggest. An SMBC Nikko Securities tally reported by Reuters projected a 27.9% increase in net profit for TOPIX companies with March fiscal year-ends in the year ending March 2027. The calculation covered only about half of the reporting universe and used analyst estimates for companies that had not yet reported, so 27.9% is not a finished result.

I would pay more attention to another part of the same tally: 95 companies had raised their profit forecasts and only 7 had cut them.

That is not an AI statistic. Banks, transport equipment and electrical machinery all helped. If the gap between upgrades and downgrades remains wide in autumn, the Japanese market has something more durable than a handful of expensive technology stocks.

Why the Nikkei can mislead

This is also where overseas investors should be careful with the Nikkei 225.

The Nikkei is the best-known Japanese index outside Japan, but it is price-weighted. A company with a high share price can exert a large influence even when its market value is much smaller than another company's. Semiconductor-related names can therefore make the headline index look stronger, or weaker, than the market underneath. TOPIX is closer to the market-capitalization-weighted benchmarks familiar to international investors and covers a much wider section of the Japanese market. A Nikkei rally driven by a few AI names tells us that those names are doing well. A rising TOPIX alongside earnings upgrades in banks, industrial companies and domestic businesses tells us something more useful about corporate Japan.

For year-end, I would rather see the second version.

There is another reason. Corporate cash is being treated differently from a decade ago.

Japanese companies historically tolerated large cash balances, cross-shareholdings with business partners and low returns on equity for longer than U.S. investors would usually accept. Cross-shareholdings are particularly Japanese: a manufacturer might own stock in a supplier or customer partly to preserve the business relationship, even when holding those shares was not the best use of capital.

That culture has not vanished. But the Tokyo Stock Exchange has been pushing Prime and Standard Market companies since 2023 to examine their cost of capital, profitability and market valuation. In April 2026 it sharpened that request, with more emphasis on how management allocates resources.

This is not an order to buy back stock. The TSE says explicitly that one-off dividends and repurchases are not the goal. What has changed is the conversation inside Japanese boardrooms. Cash, underperforming businesses and strategic shareholdings now need more explanation.

For investors, that opens a route to returns that has little to do with AI or even rapid economic growth. A Japanese company does not need explosive sales growth to become a better stock if it sells unnecessary assets, unwinds cross-shareholdings, raises dividends or stops putting capital into businesses that earn poor returns.

The autumn results will show how much of the current earnings strength survives higher costs and a less forgiving currency. By then, investors should also have a better idea of whether 3% Japanese bond yields were a brief shock or something companies must learn to live with.

That distinction is more interesting than trying to guess another round number for the Nikkei.

The market I would want to own into year-end is not one in which every AI stock keeps rising. I would rather see earnings upgrades survive outside AI, companies continue to improve capital allocation, and businesses protect margins without needing a permanently weak yen.

If that breadth appears, Japan has an investment case that is increasingly its own. If it does not, the nearly 3% available on a 10-year Japanese government bond will stop looking like an odd historical milestone.

It will start looking like competition.