May Wrap Up: What's driving the AI market rally?

Last month's update was entitled "Who would have predicted that?" – we could almost run the same title again. While fears of stagflation dominated sentiment early in the month and the Strait of Hormuz remains closed, markets ended the month in an exuberant mood. It was AI all the way, with AI infrastructure and the chips powering inference rallying strongly. The poster child for the performance of chip companies was the stellar rise of the Korean share market, up 28.4% for the month and dominated by memory companies SK Hynix and Samsung. The concentration of market performance on all things AI certainly has our attention. Caution is warranted. While we still see upside in this megatrend, forgetting the other dynamics playing out in the global economy, like the ongoing Iran crisis, is not wise. This is a time for smart diversification in portfolios.

AI is a dominant theme in both the media and financial markets. According to research from Goldman Sachs, AI-related exposures now account for over 40% of the US S&P 500's total market value.   

AI companies are not only dominating market value but are also delivering extremely strong earnings growth. According to Goldman Sachs, AI-related companies drove around half of the explosive, over 25%, S&P 500 earnings growth in the first quarter. At least on the surface, strong earnings growth underpins the strength of AI related stocks and makes some of the bubble conversations around the space premature.  

As the AI rally matures, where the gains are accruing is changing. While it was the hyperscalers and NVIDIA that were the primary beneficiaries of the AI boom, the market's attention has moved, increasingly focused on where there are infrastructure bottlenecks as the data centre build goes parabolic.

Coatue Management, a global investment firm, has produced insightful work in this space. They highlight the growing demand, fuelled by the AI buildout, in areas like transformers, switchgear, and high-bandwidth memory. 

The combination of massively increased demand for these products, coupled with long lead times to bring new supply on stream, results in bottlenecks. This pushes prices and margins higher and has turbocharged both the profits and share prices of companies in what had historically been deeply cyclical parts of the market. 

In May it was memory chips in the AI bottleneck limelight. As models scale and AI workflow becomes increasingly agentic, the ratio of CPUs to GPUs is evolving, putting immense pressure on High-Bandwidth Memory (HBM) and specialised DRAM. This dynamic shifts the AI cash flow windfall downstream to advanced packaging and memory leaders like SK Hynix and Samsung, as well as the specialised capital equipment makers like Applied Materials and Lam Research that build the machines to manufacture them.

The stock price moves on the back of this theme were eye-watering. SK Hynix, a memory chip manufacturer, was up 81% in May, capping off a 258% return for the year to date (as at 30 May 2026). Samsung cleared major qualification metrics for its ultra-dense High Bandwidth Memory nodes leading to a 44% rally in the month. 

There is no doubt that the market is experiencing a full AI exuberance moment. As anyone who has invested in markets for long enough knows, this exuberance is unlikely to last forever. 

The chip sector provides a useful case study on how this might play out. Historically, the memory chip industry has been deeply cyclical. It has been a brutal and oft repeated pattern; strong demand for chips results in high prices, which triggers massive capital expenditure in new supply, which leads to overcapacity, a collapse in pricing, and billions of dollars in write-downs.

While rarely said with confidence, there may be reasons to believe this cycle could be different. HBM is more complex to manufacture than traditional, DRAM memory and is intrinsically less commoditised. This reduces yields, increases the time to spin up new capacity, and has led to the buyers of memory locking in long term contracts, a practice that is very unusual in this space. 

This could elongate the cycle and give memory sellers more pricing power, but, in this analyst's view, the seeds of a cycle's destruction are always laid at higher prices. High prices inevitably attract more competition which drive down returns, unless a company has a very wide and impenetrable moat. To quote a famous Kiwi "it may not happen overnight, but it will happen." The dip will come.

Herein lies the challenge of today's AI-fuelled market. It is attracting truly spectacular sums of capital investment; the competitive moats of many players don't look that wide, and risks will build over time. 

Caution is warranted. While we still see upside in the AI megatrend, we don't focus myopically on this theme. The ongoing Iran crisis and its long-term strategic implications for supply chains and geopolitics are important considerations as we build portfolios, as are changing global demographics, the green energy transition and changes to how corporate balance sheets are funded. 

While the market is increasingly putting all its eggs in the AI basket, we prefer a more diversified approach, opting for smart diversification in portfolios. By doing this we still benefit from ongoing AI strength, but we never forget that it isn't the only game in town. 

Frank Jasper

ASB Chief Investment Officer