The silicon chokepoint: why the 2026 AI chip shortage is different — and how to invest in it

After two years of explosive gains in AI related shares, many investors now assume the biggest opportunities are already behind us. Which is understandable given how strong the returns have already been. But despite that performance, I continue to believe the broader AI cycle is still in its early stages.

Our Rand Swiss AI Portfolio has returned approximately 160% since inception versus roughly 92% for its benchmark over the same period. Annualised returns since launch have reached 67%, compared to 55% for the benchmark. The portfolio is also up 23.80% year-to-date in US dollar terms, with April alone delivering a return of 19.53%.

But despite that performance, I continue to believe the broader AI cycle is still in its early stages.

Why the biggest AI opportunities may still lie ahead

I believe the market is still underestimating both the scale of the AI infrastructure buildout underway and the speed at which AI is moving from experimentation into real-world commercial deployment.

The world’s largest technology companies continue to commit hundreds of billions of dollars to data centres, compute infrastructure, networking, and next-generation AI models. Importantly, these companies can already see further ahead than the broader market.

The models you are using today are not the most advanced systems operating inside AI labs. There is still a significant pipeline of unreleased technology being trained and prepared for deployment.

At the same time, the economics of AI are improving rapidly as the market shifts from consumer chatbots toward enterprise AI systems capable of automating complex workflows.

We are already seeing evidence of this shift in the revenue numbers emerging from the sector. Anthropic’s annual recurring revenue reportedly increased from roughly $100 million in early 2024 to around $30 billion by April 2026. At current growth rates, some estimates suggest the company could exit the year with annual recurring revenue of between $80 billion and $100 billion.

That is not normal corporate growth.

It suggests AI may already be transitioning from a speculative technology theme into one of the fastest monetising platforms in modern corporate history.

At the same time, demand for compute infrastructure continues to exceed supply.

The bottleneck is no longer demand.

It’s infrastructure.

How I’m positioning my AI portfolio for the next quarter

The market is increasingly realising that AI is far more than a software theme. The entire ecosystem depends on semiconductors, networking infrastructure, data centres, and compute capacity. Every new AI model, enterprise deployment, and autonomous agent requires enormous processing power behind the scenes.

That is why I continue to favour companies controlling key parts of the AI infrastructure stack, particularly businesses with supply chain dominance, pricing power, or strategic bottlenecks that competitors cannot easily replicate.

• ASML (ASML | NASDAQ / Euronext Amsterdam): Few companies are as strategically important to the AI supply chain as ASML. The company produces the lithography machines required to manufacture the world’s most advanced semiconductors and remains deeply embedded in the global chip ecosystem.

• TSMC (TSM | NYSE ADR / 2330 | Taiwan Stock Exchange): This remains another major holding in the portfolio. If ASML builds the machines, TSMC is the factory producing many of the world’s leading-edge AI chips. Together, ASML and TSMC control one of the most important bottlenecks in the global AI economy.

• CoreWeave (CRWV | NASDAQ): Over the last quarter, I also added exposure to CoreWeave, which sits directly in the middle of the growing compute shortage developing across the AI industry. Demand for cloud-based GPU infrastructure continues to accelerate as enterprise AI adoption expands.

• IBM (IBM | NYSE): I also added IBM to the portfolio. While it may appear like an unusual AI investment at first glance, I believe the market may be underestimating the importance of enterprise integration as corporations begin layering AI into existing systems and workflows.

• Novartis (NVS | NYSE ADR / NOVN | SIX Swiss Exchange): At the same time, I exited Novartis and reduced broader healthcare exposure. While I remain constructive on the long-term role AI could play in healthcare and drug discovery, I believe regulatory bottlenecks may slow the speed at which those productivity gains fully translate into shareholder returns.

AI remains one of the biggest long-term opportunities in global markets

The companies dominating semiconductors, compute infrastructure, and enterprise AI integration may ultimately become some of the most strategically important businesses of the next decade.

Many of these firms already control critical bottlenecks within the AI ecosystem, positioning them at the centre of the broader infrastructure buildout now underway.

That is why I continue to believe the broader AI cycle is still in its early stages.

Infrastructure spending is accelerating, enterprise adoption is scaling rapidly, and demand for AI compute continues to outstrip supply. That combination remains one of the most compelling long-term investment opportunities in global markets today.