AI, In Full

The world of finance is fascinated by hyperscalers and their capital expenditures. How not to be?

But the AI market must be looked at in full. In his encyclical, Pope Leo XIV noted that ‘nothing in the world of AI is immaterial or magical.’ Every response, he added, involves ‘vast networks of natural resources, energy infrastructure and, above all, people.’

To form a holistic picture of the AI market, supply (capex) dynamics must be complemented not only by a view on AI demand, but also by a perspective on the policy response that may follow.

At this stage, AI demand, measured by economically material enterprise deployment, remains underwhelming. 80% of companies report no meaningful bottom-line impact. Close to half of executives describe AI adoption as a ‘massive disappointment,’ up from a third last year.

But enterprise AI adoption will inflect over the next couple of years, as firms transition from experimentation to workflow redesign and adjust their organizational structure. Then, profit will improve, growth will accelerate, and valuation will expand in many sectors, led by ‘AI Winners.’

This is precisely when a political backlash may gain momentum. The argument that AI crowds out capital for housing, manufacturing, and the energy transition will increasingly resonate with voters. High‑profile AI failures or abuses may unsettle the public. Add labor displacement and the political case for regulating new data center supply will become irresistible. I am sensing that the divergence between AI haves and have-nots is already becoming a source of corporate angst.

Against this background, every nation will need to make difficult choices about social cohesion, global competitiveness, the level of integration in the global economy, the environment, and security. A new notion of sustainability may emerge.

Incremental data center supply may thus slow just as enterprise AI demand takes off. Those who built an installed base before the gate closed will control the most consequential infrastructure layers of the 21st century and capture a growing share of the trillions of US dollars created by AI.

For data center equipment suppliers, the game will simply change. The vast installed base will require servicing and upgrading to extract ‘more’ from a constrained supply while enhancing electricity and water efficiency: a steadier, higher-margin business with earnings visibility.

The next act belongs again to the people. As economic rents become increasingly concentrated in corporate clusters, the political pressure to tax, regulate, or break up dominant players will rise.  

The future may turn out to be different. Regardless, the better position from a business perspective is to be among those whose rents may ultimately need to be redistributed, rather than among those waiting for it.