The long-term growth prospects for data centres remains strong, but the sector also faces new considerations. Firstly, the question of overvaluation — driven by excitement about AI — has been raised. Secondly, concerns exist over development becoming more difficult due to land and power constraints, new regulations and community pushback. And thirdly, what does an exit look like in this environment?
When examining the first question of overvaluation, it is essential to decouple the concepts of AI demand versus data centre demand. Demand for compute capacity to run AI workloads (including model training and inference) is just one component of overall data centre demand, and while it is currently the fastest growing, it is not the largest.
Furthermore, not all AI workloads have the same demand profiles or data centre requirements, and there is much more driving data centre demand than AI model training.
Data centre demand growth today can be, largely speaking,
