News & media AI and the Increasing Demand on Digital Infrastructure
At a time when banks and investment funds are looking to take lower marks on their holdings in commercial real estate, we are seeing a countertrend in data centres. Whereas demand for office space is being affected in some cities by lower uptake, as recession and new work patterns reshape the work environment, data centres are seeing accelerating demand. Demand for fibre capacity is also growing, though in a less obvious manner.
The reason behind these two emerging trends is the use of generative artificial intelligence (“generative AI”) that is washing over the industry.
Generative AI, a sub-category of artificial intelligence (AI), is designed to recognise patterns and make predictions, based on new outputs on which the algorithm has been trained. It has a wide range of applications, such as creating new images and image recognition, writing news articles and poems, or even generating new music and sound effects.
Generative AI demands significant computing capacity to work its magic. This surge in demand is reflected in rising power usage (using a density measure of kilowatts per rack of server).
More specifically, AI depends on:
- Storage of vast amounts of data;
- Access to additional, large amounts of data from different sources;
- Fast computing;
- Delivery of the resulting information as one major output (often-times through multiple parallel units producing); and
- Significant, fast connectivity.
These ripple effects spread across the digital infrastructure landscape, as higher traffic volumes ultimately reveal themselves not only in demand for more square footage of data centres, but also in the need for more fibre and upgraded mobile data networks. This represents an important opportunity for investors who are prepared to respond in a timely fashion and lends itself well to our “Buy, Build and Grow” strategy in the digital infrastructure space where we have investments in fibre-optics and cable (CRA and Emitel) and data centres (Hudson Interxchange) that are delivering long-term sustainable growth and returns.
Another important point of note is that AI adoption does not appear to progress in a linear manner: it can rise in a short period due to the nature of the software maturity cycle and end-user behaviour. This is in contrast with hardware/digital infrastructure, where deployment goes through a different time cycle of delivery, whilst additionally facing the real-world limitations of availability of power and space.
The higher power usage at data centres will have several ESG implications. Higher power densities throw off more heat and may take data centre cooling in a new direction. Direct cooling at the server or chip is increasingly being examined as a means of replacing the air-cooling methods in use today. The new approaches will likely appear first and most often in newly commissioned spaces. AI will also impose more focus on power utilisation efficiency ratios, as the imperative of efficiency will increase in tandem with overall power consumption. After all, some estimates place data centres as consuming ca. 3% of global electricity supply.
To conclude where we began, AI arrived largely unheralded and has revealed itself as a significant new demand driver in digital infrastructure. While the scale of this new demand driver is still uncertain, the direction of demand growth is certainly clear.
