In May 2026, the investment landscape on Wall Street remains dominated by a single theme: Artificial Intelligence. However, the focus has shifted dramatically from the software applications and chatbots of 2023-2024 to the hard infrastructure required to power them. Recent analysis from Morningstar Canada poses the critical question: can the top US stocks, which have ridden the wave of data center construction and semiconductor demand, maintain their momentum?

The Shift from Training to Inference

For nearly three years, the market was fueled by the need for training large language models (LLMs). This meant an insatiable demand for Nvidia’s GPUs. But 2026 marks a maturation of the market. Demand is now migrating toward "inference" – the process where an AI model is executed for the end-user. This shift favors companies offering more specialized and energy-efficient solutions.

Analysts point out that while Nvidia remains the leader, competition from custom silicon solutions (like those from Broadcom and Marvell) has begun to eat into profit margins. Infrastructure is no longer just about the chip; it’s about the entire ecosystem: from server cooling to the advanced fiber optic networks connecting GPU clusters.

The Energy Wall and the Utilities Winners

One of the most fascinating phenomena of 2026 is the emergence of energy companies as "tech stocks." AI has proven to be extremely power-hungry, leading to an electricity supply crisis in the US. Companies managing the grid and those investing in Small Modular Reactors (SMRs) have seen their valuations skyrocket.

  • Electricity demand from data centers is expected to triple by 2030.
  • Utilities with stable contracts with tech giants (Hyperscalers) show the lowest volatility.
  • Sustainability and ESG metrics are now a central part of AI infrastructure evaluation.

Morningstar emphasizes that investors must look beyond the obvious players. Companies providing liquid cooling systems, such as Vertiv, have become indispensable, as traditional fan-based methods fail to manage the heat generated by next-generation processors.

Valuations and the Bubble Risk

Despite the optimism, skeptics warn of excessive valuations. Many infrastructure stocks are trading at price-to-earnings multiples reminiscent of the dot-com era. The difference, according to Morningstar, is profitability. Unlike 2000, today's infrastructure leaders are generating massive free cash flows.

"The question is not whether AI is real, but whether the return on investment (ROI) for the companies buying this infrastructure will justify the continued spending of billions," the report states.

If major software firms fail to turn AI infrastructure into profitable subscription services for the general public, the pace of infrastructure investment may slow down abruptly by 2027.

Conclusion: Selective Positioning

2026 requires a more refined investment approach. The era where "all boats rose with the AI tide" has passed. Investors are now turning to companies with "economic moats" – those with proprietary technology or critical cost advantages that cannot be easily replicated. Infrastructure remains the backbone of the digital economy, but selecting winners is now done with stricter financial criteria.