From the CTO’s corner: Simon Gardner, Oper8 Global Group

Global spending on edge computing services is expected to be more than US$260bn in 2025, and projected to reach almost US$380bn by 2028 – a CAGR of 13.8% – according to IDC. That growth is being driven by a number of factors, which has seen organisations moving some strategic workloads away from centralising infrastructure and services with the big hyperscalers or data centre providers.

While we expect to see massive megawatt-scale AI factories in the near future, at the same time, smaller AI models – including those embedded into everyday applications – are reducing computational overheads and can be run locally in much smaller edge data centres. We are also seeing distributed computing clusters designed to run complex AI applications where processing loads can be spread to provide greater resilience and latency can be reduced with computing much closer to both data sources and end users of the applications.

Training and inference for larger and more complex models can also be distributed across multiple clusters to spread processing loads, reduce latency, and improve resilience by being closer to data sources or end users. Low latency is critical for any number of AI, ML or IoT applications where milliseconds count, from autonomous systems like self-driving cars and drones to AI-powered financial fraud detection or market analysis.

One of our own implementations of edge computing are the trackside data centres we have deployed for our Formula 1 clients. Rapid capture trackside, processing and delivery of data from the team’s headquarters to the decision-makers trackside is essential, with a latency of no more than 30ms expected.

We are also seeing new technologies – such as GPUs, direct-to-chip cooling, etc. – that provide support for higher density, AI-optimised facilities for specific applications in much smaller form factors. Cutting edge hardware and smarter design can deliver facilities that are not only quicker and easier to build, but more efficient to run. That’s making it easier for data centre operators to retrofit existing facilities and deploy services more quickly and sustainably at the edge.

On the topic of sustainability, the trend towards edge computing is also being driven by power availability, or lack thereof. The rapid growth of data centres is already putting a strain on existing grid networks around the world, and also limiting the ability of new data centres to be built in regions with unreliable or at-capacity grids. “Every single data centre in the future is going to be power-limited, and your revenue is limited if your power is limited,” said Wade Vinson, Nvidia’s Chief Data Centre Distinguished Engineer.

Instead, deploying more edge computing allows you to spread those power requirements, and reduce the risks associated with energy supply. At the same time, smaller facilities like micro or modular data centres enable you to more cost-effectively incorporate UPS devices, backup generators or battery storage – rather than attempting to provide power redundancy to keep a hyperscale data centre operational when supply is (inevitably) compromised.

Demand for locally hosted data and compute has also increased due to data centre operators’ concerns about security, sovereignty, privacy, geographic and geopolitical risks.

We’ve deployed modular micro data centres to support banking and mining operations in remote regions that represent some, or all of these concerns!  Mining companies are increasingly reliant on digital technologies to optimise their operations, so the need for localised, low-latency edge computing for real-time data processing has never been more pressing. However, mine sites (particularly in parts of Africa) often face a multitude of challenges including remoteness, lack of security, harsh environments, poor connectivity and limited on-site technical resources.

The geography itself is often the biggest risk. Our micro data centre supporting Bank of South Pacific’s operations in Vanuatu’s capital Port Vila had to deal with a magnitude 7.3 earthquake in December last year. Despite widespread power, internet, and satellite outages the data centre’s robust design mitigated potential damage – despite being physically shifted 200mm both vertically and horizontally! The UPS kicked in, ensuring critical systems remained powered while the generator started and stabilised, allowing banking operations to resume as soon as communication services could be restored.

So where do you start if you are embarking on an edge computing strategy? The number one obstacle clients raise with us is visibility and control. A centralised data centre architecture looks so much easier to manage and monitor on paper. You have fewer facilities to worry about, you can have your key skilled resources close by, and (especially if you are outsourcing to the hyperscale providers) relying on third party management services is more cost-effective and straightforward.

However, even with a distributed architecture and facilities located in remote areas, gaining the visibility and control of your data centre assets and performance is achievable with the right mix of security measures, automated tracking, proactive maintenance, and environmental monitoring solutions in place.