Thought Leadership article by atNorth Director of HPC Operations Rui Gomes for Data Centre News UK, published 10 July 2023
There are currently two megatrends that are creating rip tides within global businesses. The first is the explosion in using Artificial Intelligence (AI) as businesses strive to automate and optimize key processes that can increase productivity and operational efficiencies. The second is the worsening climate crisis and the financial impact of increasing energy costs.
At the intersection of these two mega trends is the humble data centre, and choosing the right one can not only future-proof against technological advancements but also significantly reduce a business’s environmental impact.
Digital Infrastructure and the Impact on the Environment
AI technology requires a significant investment in digital infrastructure to allow for the storage and almost instantaneous processing of huge amounts of data. AI algorithms, particularly deep learning models, require significant substantial computational resources such as Graphics Processing Units (GPUs) and multiple software frameworks. To train just one computer to behave in a humanlike way via machine learning (ML) on a GPT‑3 large language model can use upwards of 12MW of data and cost up to $3 million dollars.
The resulting demand on digital infrastructure resources, capacity and space is significant. Many legacy data centers built for general-purpose computing are not able to accommodate the equipment necessary to cope with such high-density workloads, and so businesses are having to carefully consider their data center provider to future-proof their requirements.
Additionally, data centers that are built to accommodate the high-performance workloads that AI technology necessitates require powerful systems with significant cooling capabilities. These cooling systems have been estimated to be responsible for up to 40% of total energy costs within traditional data centers — a substantial drain on resources amidst the tipping point of an energy crisis fueled by political uncertainty that has caused the cost of energy to skyrocket.
It appears businesses have an almost impossible task to balance the need to stay competitive, remain solvent — if not profitable — and meet carbon reduction targets too. This is where the modern data center comes into play.
The Data Center Solution
Choosing a data centre provider is a long-term decision, and there are many factors to consider should businesses wish to mitigate the financial, technological and ecological fallout that seems likely to occur as the demand for AI continues.
Modern, purpose-built, ‘next generation’ data centers that are designed for High-Performance Computing (HPC) should have a high-density strategy in place and be fully equipped for high-capacity services with second-to-none connectivity and security to cater for evolving market changes. These optimised environments can deliver more computing power and tailor-made high-density hosting to support its customers with increasing workloads and applications at speed and scale.
Yet, there are other elements to comprehend, and in this pivotal time of environmental uncertainty, the geographical location of digital infrastructure is becoming more important than ever.
Countries that benefit from a surplus of renewable energy sources such as hydro and geothermal energy can offer a stable long-term power supply that is significantly cheaper compared to its natural gas counterparts. Similarly, countries with a cool and more consistent mild climate will ensure that steady temperature and humidity levels within the data center are maintained more efficiently, reducing energy outputs and pollution and ultimately decreasing carbon emissions.
Newer data centers in cooler climates can be designed with more energy-efficient cooling systems. For example, direct liquid cooling which leverages the higher thermal transfer of water to more efficiently cool equipment or heat capturing technology capabilities so that all warm air exhaust from the data center is recycled and the excess heat can potentially be used by local energy providers to heat homes in local communities. These are highly energy-efficient processes that drive down the total cost of ownership and will go some way to future-proof businesses against any sustainability or carbon reduction standards that may be introduced in the future.
In summary, the colossal requirement of energy required to fuel the rise in AI technology is completely at odds with any ESG and related sustainability targets a business might have. In the middle of this disconnect is the data center, and those companies wishing to embrace AI, reduce their energy costs and their carbon footprint would do well to choose their site wisely.
Read the complete article online here.