Nervous about an October crash? Diversify your AI exposure

HISTORICALLY, stock market crashes tend to happen in October. However, this year I sense that investors are less apprehensive than in previous years.

Since bottoming out in early April, the increase in the S&P 500’s market capitalisation has exceeded US$15 trillion, which is more than a third of US GDP. This has had an incredible impact on the wealth of American consumers, with the top 10 per cent of earners driving 49.2 per cent of consumption, said Moody Analytics.

Understandably, the US Federal Reserve’s decision to restart interest rate cuts has given market bulls more confidence that the current positive trend in global equity markets will continue. This seems especially likely given that – despite the obvious weakness of the US labour market – the risk of a US recession appears minimal, although non-farm payrolls have remained at levels typically associated with a recession since May.

What has driven the resilience of the US economy? Investment in artificial intelligence (AI) has played a pivotal role in this, contributing nearly 1 per cent to the 1.4 per cent GDP growth achieved in the first half of 2025. This surpasses the peak spending seen in the late 1990s, when companies were forced to invest to overcome the millennium bug.

The turbocharged levels of AI spending reflect the significant capital expenditure boom taking place in order to develop Al infrastructure such as data centres, power utility networks, expanding and upgrading fibre optic networks, among others.

This appears to be the missing ingredient in solving multiple macroeconomic puzzles and explains why global trade has stayed resilient, despite the rather substantial tariffs imposed by the Trump administration, as most of the required infrastructure parts are manufactured outside the US.

It also explains the discrepancy between the technology sector and the wider economy, where sentiment indicators reflect soft demand and lower trending inflation.

In the second quarter of 2025, the top 10 tech companies in the US spent US$426 billion on AI investment, which was up 73 per cent in just a year. Since the launch of ChatGPT, the market capitalisation of the top 10 US tech companies has increased by US$10 trillion – around a third of US GDP.

Clearly, the US economy is now geared to the US stock market, with the outcome of the AI capital expenditure cycle the key determinant of the economy’s long-term health.

Most interestingly, the AI “contagion” has found its way to China, where the AI theme has also been driving share prices but from a much lower valuation than the US. The Hang Seng Tech Index, comprising 30 of the largest tech companies, is up more than 40 per cent this year, significantly outperforming the Nasdaq Composite, which is up 18 per cent.

The strong performance of the Hang Seng Tech Index is largely due to mainland institutional investors buying Hong Kong-listed Chinese stocks, particularly AI-related technology stocks. This is reflected in the continued inflows into the Southbound Connect in recent months, where net inflows have risen from HK$42 billion in September 2024 to HK$167 billion in April, the biggest monthly net inflows since January 2021, and a monthly average of HK$137 billion (S$22.7 billion) so far this quarter.

Strong upward momentum in the Hong Kong-listed Chinese stocks has spread into the mainland markets as well, with the CSI 300 up 18 per cent over the past three months. Data released by the Shanghai Stock Exchange indicated that more than 2.65 million new A-share accounts were opened in August, a 35 per cent increase from July and a 165 per cent year-on-year increase.

Although AI-related technology spending and breakthroughs have boosted stock market performance in both the US and China this year, the two countries’ approaches to AI are fundamentally different.

China’s approach focuses on inexpensive, practical open-source applications, such as DeepSeek and Qwen. Many of these applications have the potential to be commercially viable and generate new demand. AI in healthcare is one good example.

In contrast, OpenAI and the hyperscalers in the US are still seeking to build superior, proprietary large language models (LLMs) in pursuit of artificial general intelligence (AGI).

Neither approach is superior to the other. However, DeepSeek’s approach is clearly more likely to generate commercially viable use cases for AI technology more quickly than the US hyperscalers’, whose AI business model has shifted from asset-light to asset-heavy. This puts pressure on them to monetise their much heavier AI capital expenditure. For example, OpenAI is reportedly losing around US$1 billion a month despite generating approximately US$10 billion in annual revenues, and it is expected to have burned US$115 billion in cash by 2029.

While there is little doubt that AI has the potential to transform many sectors of the economy, historical evidence suggests that it may take longer for meaningful productivity gains to materialise. It is important to distinguish between people benefiting from the increased convenience of AI and genuine, substantial productivity gains achieved through its use in industry.

Therefore, despite the growing momentum of AI, investors are primarily concerned about sluggish productivity growth, although there are indeed sporadic signs of promise.

Investors should think about spreading their money around a bit and not put all their eggs in one basket with AI investments.

The writer is managing director and chief investment officer, South Asia-Pacific, UBS Global Wealth Management. He is also adjunct associate professor, Nanyang Business School, Nanyang Technological University.