The next phase of the AI trade

By Joss Murphy on 12 August 2026 in Specialist investing

While few doubt the potency of the AI trade, it is becoming increasingly difficult to know how to benefit from it in a portfolio. The more obvious routes – through semiconductors or the hyperscalers – look crowded, but the next generation of winners is not yet clear. At the same time, investors can’t afford to ignore the most significant technological shift in a generation.

Are the obvious AI winners running out of road?

The traditional route to AI has been through the US technology giants. This includes some of the semiconductor groups, but many investors have believed that a decent holding in the Magnificent Seven would probably give them the AI exposure they needed. However, there are growing doubts about this route. The hyperscalers have spent eye-watering amounts on their AI build-out, to the detriment of their cash flow and balance sheets. Investors have started to worry that they are not worth their lofty valuations.

Aditya Khowala, manager at Premier Miton, agrees: “Some headwinds to the AI trade have started to surface, namely switching to open-source models over frontier models, and SpaceX and Meta offering their cloud services to compete with AWS, Azure, GCP and Oracle, thereby creating concerns of over-investment. This remains our core view – that the current AI investments will struggle to generate the required return on invested capital and will therefore lead to substantial slowdown and losses.”

Looking beyond Silicon Valley

If not via this route, how should investors be taking exposure to the AI trade? The semiconductor and memory behemoths have been seen as a key beneficiary. Much of the work done by the Silicon Valley giants relied on infrastructure provided by these companies. The market has woken up to this over the past 12 months. TSMC has seen its share price grow over 60% over the past 12 months. Samsung is up 220% over the same period**.

For Ben, these companies are supported by the growth of agentic AI – this is the “non-human scaling” part of the AI story. AI can now synthesise information in a way humans cannot and this should usher in an age of ‘cognitive abundance’. He remains focused on semiconductors/hardware assets in Japan, South Korea and Taiwan, as well as in the US, with Nvidia, Micron and Lam Research in his top 10 holdings*.

However, some of the Asian giants have started to look like a crowded trade. There have been concerns over the leverage building in the Korean market, where loose regulation has allowed the widespread use of leveraged ETFs. This has created significant volatility.

That said, those valuations are anchored to exceptionally strong and potentially peak-cycle earnings. James adds: “What matters, therefore, is not simply the multiple, but the sustainability of the earnings base to which that multiple is applied. In our view, current pricing in parts of the AI supply chain increasingly reflects expectations that are both elevated and fragile. Should demand normalise, supply respond, or capital intensity rise, there is a risk that earnings expectations are reset, potentially sharply.” This argues for some selectivity on the AI infrastructure names.

James sees more value in China. Whereas the Korean and Taiwanese technology giants are ‘ambiguously cheap’, Chinese AI companies are ‘unequivocally cheap’. He points out that some of the Chinese companies equivalent to the US companies trade on significantly different multiples***.

China is taking a different approach to AI, championing ‘open source’ technology rather than the closed system used by the Silicon Valley giants. Open source can be used by international governments and companies, and adapted to their own needs. It becomes ‘theirs’. In contrast, US closed AI systems can be turned off at the whim of a disgruntled chief executive or political leader. This has made the Chinese option more appealing to many countries.

French think tank IRIS says China’s open-source strategy serves three key objectives. “It undermines US tech dominance by providing a viable alternative to proprietary systems. It builds a new tech alliance, particularly among developing nations that lack the compute resources for expensive models. And it lays the groundwork for a parallel tech ecosystem.

The next generation of AI winners

Selected commodities are another option. Ben Rogoff says that their AI team identified the power complex as being essential for AI, “given how power-intensive AI is relative to traditional compute…areas such as power, transmission, cabling.”

He believes that it will also start to become clear which companies are using AI effectively to improve their businesses. This could power a new wave of ‘AI winners’ beyond the more obvious technology giants. “These are companies using AI to reduce costs, improve workflows, increase pricing power, launch new services or deepen customer relationships. This part of the story is still underappreciated.” He believes smaller companies, with fewer legacy processes to unpick, should be able to move faster than sprawling incumbents and may be in a better position to benefit^.

The market is starting to be more nuanced in its understanding of the winners and losers from AI. This is likely to continue as AI adoption builds momentum. This should present opportunities for investors willing to look beyond the obvious AI players.

*Source: Investment manager’s report, 30 April 2026

**Source: Yahoo finance, at 10 August 2026

***Source: Federated Hermes, 14 July 2026

^Source: Montanaro, June 2026