Polen Capital · 2026 · Published 12 August 2026
Simon Rafferty argues that AI spending is moving from chips to the land, substations, transformers, cooling and labour needed to bring compute online, and that small and mid-cap companies gain more because a $1 billion revenue opportunity moves their share prices 20-44% in Polen's model against about 1.6% for large caps.
Simon Rafferty, Senior Investment Specialist on the 5Perspectives Growth team, writes on 12 August 2026 that the AI opportunity is broadening from platform development and model training to the physical infrastructure that brings compute online. Hyperscalers including Alphabet, Amazon, Microsoft, Oracle and Meta spend hundreds of billions of dollars a year on data centres, chips, networking and power. He names power and grid access, skilled labour and electrical and thermal equipment as the main bottlenecks. Small and mid-cap indices hold more of the industrial, electrical, construction, materials and energy companies doing this work, while large-cap indices hold the hyperscalers that pay for it. He adds that SMID allocations have fallen for two decades, their valuations stand at a discount to large caps, and fund flows remain low.