Market Commentary and Fund Performance

Performance data quoted represents past performance; past performance does not guarantee future results. The investment return and principal value of an investment will fluctuate so that an investor’s shares, when redeemed, may be worth more or less than their original cost. Current performance of the fund may be lower or higher than the performance quoted. Performance data current to the most recent month end, and standardized performance can be obtained by viewing the fact sheet or by clicking here.

Market Highlights

In October 2025, the TOPIX, a representative index of the Japanese stock market, rose 1.81% compared to the end of the previous month.

The first half of the month started off soft amid concerns over a potential U.S. government shutdown. Market sentiment shifted sharply when Sanae Takaichi became leader of the ruling Liberal Democratic Party (LDP), as expectations for proactive fiscal policy and growth initiatives triggered a rapid rally in stocks and a weaker yen—in what came to be known as the “Takaichi Trade.” Around mid-month, reports that the Komeito party would end its 26-year coalition alliance with the LDP fueled political uncertainty. Around the same time, the announcement of additional U.S. tariffs on China and China’s retaliatory measures intensified risk-off sentiment, temporarily driving the Nikkei Stock Average lower. Political uncertainty eased following reports of coalition talks between the LDP and the Japan Innovation Party (JIP), while gains in the U.S. SOX Index (Philadelphia Semiconductor Index) further supported a market rebound.

In the latter half of the month, global macro factors ranging from renewed U.S.-China trade tensions and credit concerns surrounding U.S. regional banks weighed on market sentiment. While short-term overheating also contributed to a temporary correction, Japan’s stock market resumed its upward trend as policy expectations rose following the formal coalition agreement on the 20th between the LDP and JIP, and the establishment of new Prime Minister Takaichi’s cabinet.

Toward the end of the month, the U.S. Federal Open Market Committee announced the expected 0.25% rate cut, while prospects for an additional December cut receded following remarks by the Federal Reserve Board’s Chair. Meanwhile, the Bank of Japan’s Monetary Policy Meeting refrained from raising rates and signaled caution regarding future hikes, contributing to continued yen weakness. Positive developments in U.S.-China trade talks, along with China’s decision to delay rare earth mineral export restrictions, further supported risk-on sentiment.

Against this backdrop, artificial intelligence (AI) and semiconductor-related stocks advanced steadily, supported by strong earnings from Advantest and a sharp share price increase for Lasertec, while the Nikkei Stock Average set new daily record highs. As a result, although the increase in gains widely varied across indices, the Japanese stock market ended October at a significantly higher level than at the end of the previous month.

The Fund’s Performance

This month, the Fund returned 3.17% (HJPIX), outperforming its benchmark, the Russell/Nomura Total Market™ Index, which returned 2.20%.

The month's positive performer among the Global Industry Classification Standard (GICS) sectors included shares of Communication Services, Industrials, and Information Technology while Financials, Consumer Staples, and Consumer Discretionary detracted from the Fund’s performance.

Among the best performers were our investments in SoftBank Group Corp., the telecom and Internet conglomerate, Hitachi, Ltd., one of Japan’s oldest electric equipment & heavy industrial machinery manufacturers and Tokyo Electron Limited, one of the world's largest manufacturer of semiconductor production equipment.

As for the laggards, Tokio Marine Holdings, Inc., Japan’s largest insurance group and  Mitsubishi UFJ Financial Group, Inc., one of Japan’s largest financial groups.

October Commentary

AI as an investment theme

The Fund does not pursue thematic investing—identifying the latest hot theme and buying stocks that fit into it. That said, guided by our investment philosophy of "Invest in a great business with exceptional management at an attractive price," our holdings often end up aligning with certain themes.

AI, which has been dominating equity market headlines, is a case in point. As noted in our August commentary, we hold SoftBank Group (SBG).

Since our initial investment in 2015, our rationale for holding SBG has consistently been twofold: first, the long-term entrepreneurial track record of its founder, Chairman Masayoshi Son; and second, the material discount of the stock price to its net asset value at the time of our purchases.

Today, SBG is widely viewed as one of the most representative AI-related pure plays in Japan. As such, we outline our views on AI as an investment theme below.

Is the AI theme overhyped?

The sharp rise in AI-related stocks reflects an explosive increase in spending on AI infrastructure, as reported in the media frequently. September and October brought a flood of notable announcements.

On September 10, OpenAI signed an ultra-large cloud agreement with Oracle, reportedly totalling $300 billion for AI data center capacity. On September 22, Nvidia decided to invest $100 billion in OpenAI to deploy AI data centers worth 10 gigawatts. On October 1, OpenAI secured a memory supply equivalent to 900,000 wafers per month from SK Hynix and Samsung Electronics, reportedly worth tens of billions of dollars. On October 6, news broke of a deal with AMD to procure GPUs equivalent to 6 gigawatts. On October 13, OpenAI and Broadcom announced joint development of custom AI chips equivalent to 10 gigawatts. Elsewhere, Microsoft committed over $33 billion to secure GPU infrastructure with so-called neoclouds, namely Nebius, CoreWeave, Nscale, and Lambda. Finally, Meta announced a $14 billion AI cloud computing agreement with CoreWeave. And the developments continue to unfold.

These are multi-year commitments. Industry estimates point to annual AI infrastructure spending exceeding $1 trillion by 2030. Remarkably, Nvidia CEO Jensen Huang projects cumulative industry-wide AI infrastructure investment of a whopping $3–4 trillion over the next five years through 2030.¹

Such unprecedented figures invite comparison to the dot-com bubble and naturally raise concerns. Yet those at the front lines of AI—those closest to demand—do not share that view.

Based on these developments, our current investment view is as follows: Pockets of excess may exist and warrant caution, but we should not underestimate AI’s potential to boost labor productivity and overall economic growth. We are therefore cautiously optimistic at present due to a possible short-term overheating in share prices, but of the view that this could be a megatrend in the years to come.

Bullish arguments for AI infrastructure

Industry optimism about AI infrastructure rests on several data points.

The most important issue is a severe shortage of compute capacity to run large language models (LLM). Supply of GPUs and custom ASICs from Nvidia, Broadcom, and AMD remains wholly insufficient to meet demand. These chips are essential for the computing power required in data centers when delivering AI services.

Leaders calling out compute shortage² include Jensen Huang, Jonathan Ross,³ CEO of inference-chip company Groq (formerly at Google), and Greg Brockman,⁴ OpenAI cofounder.

What is causing this?

Until recently, most AI infrastructure spend was on training—amassing GPUs for model improvement. We have since shifted into a phase where inference demand is ramping as more capable models are put to work. Both training and inference require massive computing power. As AI penetrates our society, compute needs increase further and exponentially.

Inference demand is rising because:

1.    The number of users is growing,

2.    Session lengths are getting longer, and

3.    Use cases are becoming more complex.

On the third point, this year's AI usage has moved beyond simple tasks like search, research, summarization, and translation. They have now advanced to reasoning and agentic tasks. For instance, reasoning (ChatGPT o1 was the first reasoning model released about a year ago) entails hypothesis, testing, and iteration inside the LLM, consuming a significant number of hidden tokens. AI agents can autonomously browse the web on a user’s behalf—ingest entire pages, scroll and capture content, analyze images, and store data—a process that is extremely compute-intensive.

Frontier AI model developers like OpenAI frequently complain they cannot launch or scale services due to insufficient access to compute—evidence of strong demand.

From a macro perspective, Goldman Sachs⁵ estimates that current AI-related capex remains under 1% of U.S. gross domestic product (GDP)—below the inflation-adjusted levels seen in prior productivity revolutions: roughly 3% for U.S. railroad investments in the 1880s, 2% for the auto production buildout in the 1910s, and about 2% during the 1990s Information Technology (IT) boom. The report concludes that today’s AI infrastructure spend is not excessive.

AI demand is surging in the short and long term

Current AI demand is doubling every few months

CLSA’s August 26 report, “Token’s tale,” uses token volume as a proxy for current and future AI demand. All interactions with LLMs—search, image generation, coding—are processed in tokens, the smallest units of text or symbols the model reads. Longer and more complex tasks require more tokens; token volume, therefore, approximation of aggregate AI usage system.

Google’s presentation at its I/O 2025 event indicates token processing of 120 trillion in December 2024, 160 trillion in February 2025, then a jump to 480 trillion in April—a staggering ramp that marks a clear demand inflexion in April of this year. In China, Alibaba leads in AI infrastructure. At its late-September Apsara Conference, the company also reported token consumption is doubling every two to three months. These figures point to genuine end-demand underpinning infrastructure investment.

Future demand is expected to grow exponentially.

The previously-cited CLSA report projects global token volume, including Google's, could reach 80 quadrillion (80,000 trillion) in ten years—around 80 times the level at end-2024.

The precise forecast matters less than understanding what will drive token volumes higher. Today, the dominant use cases involve text. As models become truly multimodal, usage expands to audio, images, and videos. That's what will drive token volume going forward.

The last few months have seen entertainment-focused apps launch for images (e.g., Nano Banana) and video (Sora 2, Veo 3, Kling). AI-enabled games are likely to follow, and processing beyond text will place additional strain on compute capacity.

Next comes “physical AI”: fully autonomous vehicles and general-purpose humanoid robots capable of various tasks. Using AI learning processes, humanoids can now be trained directly without explicit programming. This is considered a breakthrough in the robotics field. And general-purpose robots will require massive amounts of visual data for simulation.

Tesla and several Chinese firms (e.g., UBTech) are leading efforts to replicate complex human motions. Elon Musk notes that recreating human hand dexterity is the hardest engineering hurdle. Tesla is currently developing Optimus V3; successful mass production will require vast quantities of high-quality actuators and other components and building that supply chain will take time.⁶ Nonetheless, society will likely reach this milestone eventually.

Humanoid commercialization now looks far more realistic than it did when Japan was the global leader 15-20 years ago. In a not-too-distant future, we may start seeing humanoids in factories and even homes. Musk and leading Chinese players are seriously planning for mass production.

Jensen Huang’s $3–4 trillion five-year investment outlook cited above appears to encompass the ramp toward physical AI.

Who is funding the capex?

AI infrastructure spending at hyperscalers continues to remain at healthy levels.

Despite its unprecedented scale, AI infrastructure spend today is largely funded by the ample operating cash flows of hyperscalers—primarily Microsoft, Google, Meta, and Amazon. Their core businesses—Microsoft’s enterprise software, Meta’s social platforms, Google’s search and advertising solutions—continue to grow at a robust clip, providing earnings power and balance sheets to sustain elevated capex.

According to Bank of America Securities,⁷ combined 2025 capex at Amazon, Microsoft, Google, and Meta is projected around $380 billion, versus an aggregate operating cash flow of about $530 billion. Meta sharply lifted 2025 capex guidance to $66–72 billion⁸ yet expects over $100 billion in operating cash flow—leaving substantial free cash flow for buybacks. It is definitely in a position of strength. Microsoft likewise funds roughly $100 billion of investment from its operating cash flow.

In China, Alibaba has announced increases to its already planned RMB 380 billion of AI infra capex over the next three years.

While we may see increased debt issuance to fund larger-scale projects, this seems to remain a key distinction from the 2000 tech bubble.

OpenAI is the odd one out, but…

OpenAI is a partial exception. As a still-young company, it lacks the extensive resources of a hyperscaler but remains a key player in frontier LLMs. Its unique position enables it to attract external capital from hyperscalers.

While hyperscalers also develop proprietary LLMs, they simultaneously run cloud infrastructure and dominate in their respective core businesses in search and advertising, social networking, enterprise software, and e-commerce. A monopoly position by any one of them in LLMs would be formidable. Against this backdrop, supporting a relatively neutral actor like OpenAI therefore has strategic appeal to many of the incumbents.

Nvidia plans to invest around $100 billion⁹ in OpenAI, and SoftBank Group and other large corporations are also lining up funding. Commanding a dominant GPU market share and $64 billion in operating cash flow last fiscal year, Nvidia resembles a hyperscaler in financial terms. With such backing, OpenAI also seems to be in a position of strength.

Will AI infrastructure earn adequate returns?

Shifts in value from existing industries

The question is whether the returns will justify the spend. Trillions of dollars of capital must, intuitively, generate very large revenues and profits to exceed the cost of capital.

Here, we should consider not only entirely new revenue streams (AI-enabled entertainment, autonomous mobility, home-helper robots) but also the shift of value from traditional processes to AI-driven ones across existing industries.

When assessing return on investment (ROI), it is important to look at net investment: AI spending can substitute for legacy capex that becomes unnecessary. The numerator is the profit uplift from productivity and efficiency gains; the denominator is incremental net investment, not gross spend.

Global nominal GDP is roughly $110–120 trillion. If AI captures a meaningful share of value-added now produced elsewhere, then generating above-cost-of-capital returns on $1 trillion-plus annual infrastructure spend via substitution plus new demand may not be an entirely implausible scenario.

For example, companies like Meta, Google, Amazon, and Netflix have leveraged GPU-powered AI technologies since the 2010s to automatically generate personalized recommendations for users. This is essentially a shift of value within the broader advertising industry. The strong business growth these companies have seen in recent years can be attributed, in part, to their investments in AI.

Consumer surplus

Monetization of AI services is still evolving. Today’s revenues stem mostly from just consumer subscriptions (for example, as of July, OpenAI reportedly had about 700 million weekly active users with roughly 7%—around 50 million—paying) and enterprise API usage fees.

Yet the utility of AI is widely felt. In economic terms, substantial "consumer surplus" is being created—the gap between what users would be willing to pay and the current price.¹⁰ Given the remarkable pace of recent advancements in AI capabilities, consumer surplus is poised to expand further.

This implies, in principle, there is room to charge higher prices while retaining users, making the monetization path clearer. Revenue models beyond monthly subscriptions are conceivable — including transaction-based take rates, advertising-driven revenue, and outcome-based enterprise pricing, among others. We are optimistic that the monetization issue will be resolved in due course.

Conclusion

Despite signs of short-term overheating, the Fund continues to hold AI-related stocks, for reasons we believe are justified.

First, we purchased AI-related names at attractive valuations:

SoftBank Group shares were bought at an average discount of over 40% to net asset value (NAV). We also hold Tokyo Electron (TEL), which benefits from tight memory demand amid the AI infrastructure boom. The position in TEL was established during the sharp global sell-off in semiconductors following U.S.–China trade frictions in autumn 2022. Both positions were initiated with a sufficient margin of safety at the time of purchase. Hitachi, a purveyor of power grid solutions critical to AI data centers, was initiated in 2021 at a forward price to equity (P/E) of 10x—well before the AI theme became apparent.

AI is only one of several themes in our "concentrated diversified portfolio":

The portfolio is invested in a spectrum of businesses ranging from non-life insurance, entertainment, convenience stores, homebuilding, inbound tourism, to general trading companies. Geographically, these holdings span Japan, the U.S., Europe, China, India, Southeast Asia, Oceania and beyond. Valuations range widely, too. Alongside higher-P/E growth names, we own underappreciated compounders trading around 10x P/E with total shareholder yield above 5%. Overall, the portfolio should be well diversified.

Among a plethora of investment opportunities, AI remains attractive:

By contrast, for example, large-cap automakers face U.S. tariffs and intensifying competition from Chinese electric vehicles (EVs). Structural headwinds are mounting, while Japanese OEMs have been slow to adapt. On the other hand, AI as a "theme" offers a brighter long-term outlook and greater excitement.

To be clear, we remain vigilant and have not increased our AI exposure merely to chase momentum since this summer. The uptrend will not be linear for sure; unnerving market events like January’s “DeepSeek shock” can still occur. On balance, however, we believe maintaining our current allocation is the most prudent course at this time.

Addendum: Japan's new prime minister

Corporate governance reform will continue irrespective of political leadership changes.

Lastly, on the political front, this month brought news that Sanae Takaichi was selected to be Japan’s first-ever female prime minister.

Seen as a successor to Abenomics, she advocates proactive fiscal policy and is viewed as dovish on monetary policy. She also supports strengthening domestic capabilities in AI, semiconductors, space and defence so Japan can stand on its own.

Near-term policy is likely to focus on demand stimulus, while longer-term risks could include side effects from inflation.

Unlike at the start of Abenomics, today’s Japanese economy faces supply constraints arising from labor shortages, etc., more than a shortfall in demand. It remains to be seen whether “Sanaenomics” is well-suited to current conditions and what it implies for fiscal discipline.

The long-term market impact of a change in government is uncertain. However, we are confident about one thing: the steady progress in corporate governance reform since around 2014 should continue regardless of political shifts.

That continuity may seem obvious, but recent experience in neighboring South Korea shows how a change in leadership can alter course, as it was seen to affect the distance between the government and powerful chaebol groups, raising questions about the momentum of market reforms. In Japan, by contrast, we see little risk that governance progress will reverse, no matter which political party comes into power—an important source of comfort for equity investors in Japan.

While a country's prime minister serves as its public face and shapes its global image, what truly matters in stock investing is the strength of individual companies in the private sector. Although President Trump's inauguration may have “tarnished” the image of the U.S. to some extent, the continued rise of the U.S. stock market is a good example of how corporate performance can prevail.

Large-scale foreign direct investment into the U.S.

As part of Japan's commitment to direct investment in the U.S. following bilateral tariff negotiations, the Japanese and U.S. governments jointly released a 
"Joint Fact Sheet for Japan-U.S. Investment" on October 28th. While full details remain undisclosed, Japanese companies are set to play a broad and significant role in this endeavor.

According to media reports, key participants include two of our major holdings, SoftBank Group—tasked with developing and operating major energy infrastructure—and Hitachi, which is expected to contribute to long-distance power transmission networks.

The U.S. has faced decades of industrial hollowing-out, leaving it in a position where independently revitalizing its domestic infrastructure and defense industries is increasingly difficult. In this context, Japan is expected to play a central role.

Despite prolonged deflation following the burst of its 1989 economic bubble, Japan has managed to preserve much of its manufacturing infrastructure. There was a time when Japan’s slow pace of restructuring and reluctance to fully exit certain sectors were criticized. However, in the evolving geopolitical regime defined by the U.S.–China rivalry, Japan’s manufacturing strengths may once again come into the spotlight.

Click here for a full listing of Holdings.