AI - a win for data centres?
19 June 2024
We're seeing a huge number of opportunities emerging from the AI driven demand for digital infrastructure.
Data centres in particular are feeling the benefits, both on the demand and the supply side.
Digital Bridge's CEO Marc Ganzi has said that "data centres are becoming AI factories", describing it as "a global opportunity at scale" with AI pushing demand into a "whole new gear".
But why is this? We've dived into some of the key drivers, trends and risks below.
AI is cloud trained and delivered, and needs an unprecedented amount of compute and storage in a way that traditional data centres don't provide. As we use more AI, the demand for AI suitable data centre capacity is exploding.
GPUs: Unlike traditional software, AI relies on powerful graphics processing units (GPUs) due to their "parallel computing" abilities (i.e. doing multiple calculations at once). GPU dedicated data centres are dominating CapEx plans for providers. Microsoft is planning to expand its AI data centre footprint through €4bn investment into France and $3.2bn into Sweden, deploying up to 25,000 and 20,000 GPUs respectively.
Meanwhile, GPU providers are seeing record profits and share price highs. As at mid June 2024, NVIDIA (the world's largest GPU provider) has a $3.34 trillion market capitalisation and is currently the world's most valuable publicly listed company (passing both Apple and Microsoft). In its Q1 FY25 results, NVIDIA reported record quarterly revenue of $26 billion, with data centre revenue of $22.6 billion, up 23% from Q4 and 427% year on year.
Edge delivery: AI can be 'trained' remotely where power, space and operating costs are cheaper, but usually needs to be delivered at a closer proximity - especially so for autonomous vehicles or industrial process AI systems. This requirement is driving growth in edge, and even 'far edge', data centres to serve the AI demand with minimal latency.
While driving demand, AI also presents significant opportunities for operators to deliver capacity in more streamlined and efficient ways.
Controlling cooling: Data centres consume a significant amount of power and produce a lot of heat, particularly those designed for AI. Dedicated temperate and environment sensors, coupled with trained AI systems, can make real-time decisions to intelligently direct cooling resource - significantly reducing power OpEx.
Smart maintenance: Trained AI systems can predict maintenance issues in advance, promoting strategic preventative maintenance, early fault detection and more accurate fixes. Operational tasks such as patching and updating, server upgrades and decommissioning processes can all be streamlined by AI. This reduces downtime and increases profitability.
Improving security: AI systems can be deployed to learn new threats and implement defences in a much quicker timeframe, spotting abnormal traffic and behaviour quickly and taking informed preventative security measures. AI may also help providers to comply with applicable cyber security regulations through smarter implementation of risk management measures and reporting systems.
While undoubtedly a huge benefit, AI presents new risks to providers which will need close management.
Regulatory regimes: Regulators worldwide are scrambling to respond to the rise of AI, with different countries at different stages of their journey. The EU's AI Act (the first comprehensive AI regulation to be enacted globally) will affect many data centre providers – who, if in scope, will be required to identify 'high-risk' AI systems deployed in their facilities and comply with the impact assessment, oversight, monitoring and record keeping obligations (amongst others) set out in the EU AI Act. Under the EU AI Act, AI systems intended to be used as safety components in the management and operation of critical digital infrastructure (which includes cloud computing and data centre services) are expressly classified as high-risk. See our article here for further information.
Chip shortage: To remain competitive, providers will need reliable and secure supply chains for the latest GPU components, however the raw materials required to produce GPUs are predicted to be in short supply over the next decade. Certain countries are already contemplating regulations to secure their supply and production of semiconductors (such as the EU's Chips Act).
Power demands: The significant power demand of AI data centres risks overwhelming the constrained power grids and transmission lines which supply them. It also risks pushing power utilities back towards fossil fuels, which in turn presents challenges for data centre providers' (and also their customers') ESG credentials. Additionally, the EU AI Act will require developers of large foundation AI models to report on energy consumption, so these customers may be particularly focused on procuring "green" AI capacity. Microsoft has started a collaboration with Vattenfall for the sourcing and supply of renewable energy to its new AI data centres in Sweden.
Data centre providers who best place themselves to serve the AI's insatiable hunger for space and power, while leveraging the efficiencies AI offers and managing the risks it presents, will make the most of the "global opportunity" presented by AI in the mid to long term.
The major players are aware of the huge opportunities afoot and we're seeing a number of strategic collaborations – with NVIDIA entering into high profile partnerships with Google, AWS, Microsoft, Equinix and others. We will be keeping a close eye on the market and legal issues surrounding AI and data centres as they develop.
Authors: Rebecca Clarke (Counsel), William Barrow (Senior Associate)
Visit our Data Centres hub for a list of all articles in this series
The information provided is not intended to be a comprehensive review of all developments in the law and practice, or to cover all aspects of those referred to.
Readers should take legal advice before applying it to specific issues or transactions.
Sign-up to select your areas of interest
Sign-up