The Modern Data Centre: Power, Progress, and Paradox

Duncan Clubb

SUBSCRIBE CONTACT US

Author



KEY TAKEAWAYS

  • Data centres are 'invisible' infrastructure – unremarkable buildings packed with servers, power backup and cooling that keep the digital world running 24/7.


  • Even with major efficiency gains, total electricity use is rising fast as AI and cloud demand accelerates – a classic Jevons’ Paradox problem.


  • Rack densities are surging (AI racks can reach 120+ kW), pushing power, heat and grid constraints to the forefront of data-centre strategy.


  • The industry is shifting from air to liquid cooling and exploring waste-heat reuse to manage high-density loads and improve sustainability outcomes.

4 MIN READ


Introduction

Duncan Clubb, Senior Partner – Data Centres, Edge & Cloud, was recently invited onto Commercis' Impactful Innovation podcast where he was asked to demystify the role of data centres in our digital society. 


You can watch the full interview here: https://www.youtube.com/watch?v=8ACXIjFyoQo


Duncan starts by pointing out that these infrastructure giants are often hidden in generic-looking warehouses in our physical landscape – most of us have driven past a data centre at some point without ever recognising what it was.


“Fundamentally, they’re just facilities designed to house computers” he says[1]. In reality, these unmarked buildings quietly keep computers running and data flowing 24/7. Behind every click, video stream or cloud backup lies rows of servers, storage arrays and networking gear, all backed by layers of batteries, generators and advanced cooling systems to ensure near‑constant uptime[1][2]. 


While early data centres might have been just a dusty closet full of cables, they have grown into hyperscale campuses operated by tech giants, with Amazon, Google and Microsoft at the forefront of a $250bn industry.


The Energy Paradox of Efficiency


Today’s data centres have become voracious energy consumers. Industry analysts estimate they already use roughly 1–2% of global electricity – a figure set to rise as AI and cloud demand grow[5]. Indeed, Goldman Sachs forecasts that share will climb to around 3–4% by 2030[5]. (For context, the International Energy Agency projects global data center power use will more than double to about 945 TWh by 2030 – slightly more than the total consumption of Japan today[6]). 


This raises a profound irony: data centres have made staggering efficiency gains in power and cooling, yet total energy use keeps climbing. Duncan calls this an example of Jevons’ Paradox – improvements in efficiency actually fuel greater demand. As he illustrates, “you add an extra lane to a motorway, the motorway doesn’t become less busy, it becomes more busy,” – meaning that each leap in hardware or cooling efficiency tends to unlock even more computing capacity and more power draw[7].


The result is that data centres now quietly consume more power than entire sectors like aviation or steelmaking. In Duncan’s view, this energy paradox is perhaps the defining challenge of modern computing: efficiency alone can’t contain our appetite for digital services and AI.

"

You add an extra lane to a motorway, the motorway doesn’t become less busy, it becomes more busy."


— Duncan Clubb

Scaling Up: Rack Densities and Infrastructure


To meet that demand, data centres keep scaling up. Server racks that once drew just a few kilowatts now routinely consume dozens. Where racks once consumed around 2 kW each, today’s high‑performance racks – especially for AI – reach upwards of 120 kW per rack. 

Industry surveys confirm this trend: the average rack density today is about 12 kW (double what it was a decade ago) and some facilities plan for 30+ kW racks[11]. In practical terms, a standard 42‑U cabinet is now packed with dozens of powerful servers, each fitted with multi‑core CPUs and GPUs. These advanced chips deliver jaw-dropping performance but also generate enormous heat.


Artificial intelligence looms large in this story. By one estimate, AI will account for about 19% of data center power by 2028[15]. In concrete terms, that could push data centre load in the US alone from about 3% of the grid in 2022 to as much as 8% by 2030[16].


From Air to Liquid Cooling


Cooling these high-density racks has long been the Achilles’ heel of data centers. Traditionally, for every kilowatt of computing power a server used, nearly another kilowatt might go into fans and chillers. In early days that 2:1 ratio made climate control one of the largest operational costs (and carbon costs) after the servers themselves. 


Over time, engineers have dramatically improved airflow, hot/cold aisle containment, and chiller efficiency. Today’s state-of-the-art data halls can achieve PUEs well below 1.2 in optimal climates. However, as racks get denser with AI chips, even these advances can struggle to keep temperatures in check.


As Duncan explains, one of the major breakthroughs has been liquid cooling. By delivering coolant directly to the hottest components, liquid systems move heat far more efficiently than air. He notes, “[liquid cooling] is beginning to come into its own now. The next generation of Nvidia chips, a lot of those will be liquid-cooled”. 


The latest GPUs and CPUs are often built with liquid cooling in mind. In fact, data centre designers estimate water’s heat capacity is roughly 3,000 times that of air, meaning liquid can whisk away heat orders of magnitude faster[19]. 


The liquid trend is accelerating: industry analysts predict only about 22% of current data centers use any form of liquid cooling, but that could rise to nearly 50% within five years[22]. In-rack and direct-to-chip cooling solutions are being deployed from webscale clouds to colocation facilities. Even liquid immersion (submerging servers in dielectric fluid) is gaining traction for ultra-dense clusters. 

Turning Waste into Warmth


Another innovation involves reusing waste heat. Instead of disposing of this energy, visionary operators are turning it into a community resource. In Scandinavia, for example, data centres already feed hot-water networks and district heating systems. 


As Duncan observes, “waste heat reclamation in the Nordics, where it started, is being used to provide free hot water for municipalities”. One Finnish data centre project in Mäntsälä reuses roughly 20,000 MWh per year of waste heat – enough to supply the heating needs of about 2,500 homes[24]. In Sweden’s Stockholm Data Parks initiative, multiple centres pipe residual heat into the city grid. Even outside Europe, Amazon and Microsoft are launching schemes to recycle heat: new plants in London and Finland are slated to heat thousands of homes via district networks[25].


Duncan is encouraged by these examples, noting that they underscore a shift in perspective towards sustainability. Rather than fight the byproduct of power, data centres are seeing it as a resource. This approach – along with renewable power purchases and carbon offsets – help to mitigate the energy paradox of more efficiency leads to more demand.


Conclusion 


The big challenge ahead, Duncan says, is to balance growth with sustainability. We’ve come a long way – servers are 30x more efficient than 20 years ago – but we’re still in a race against our own appetite for data and media. 


By combining smarter cooling, reuse of waste energy, and new efficiency tricks, the industry strives to meet demand without shorting out the planet. 


Watch the full podcast here: https://www.youtube.com/watch?v=8ACXIjFyoQo


Find out More about Our Services


Our data centre solutions connect cloud and data centre colocation with hybrid strategies, ensuring smoother transformations and costs benefits for clients. We support strategy, design, and construction for investors, enterprise, operators and business.


Find out more on our Data Centres, Edge & Cloud service page: https://www.cambridgemc.com/digital-infrastructure/data-centres-edge-cloud


Select Sources: 


[26][3][6][5][24] [20][22] Duncan Club’s interview on Impactful Innovation: https://www.youtube.com/watch?v=8ACXIjFyoQo


[5] [13] [14] [15] [16] AI is poised to drive 160% increase in data center power demand | Goldman Sachs

https://www.goldmansachs.com/insights/articles/AI-poised-to-drive-160-increase-in-power-demand


[6] Energy and AI

https://iea.blob.core.windows.net/assets/601eaec9-ba91-4623-819b-4ded331ec9e8/EnergyandAI.pdf


[9] [24] [25] These companies are using data centres to heat cities | World Economic Forum

https://www.weforum.org/stories/2025/06/sustainable-data-centre-heating/


[11] Exploring Data Center Rack Density | Average kW Per Rack

https://blog.enconnex.com/data-center-rack-density


[19] [22] The Future is Liquid: How In-Rack and Direct-to-Chip Cooling are Revolutionizing Data Centers

About Us

Cambridge Management Consulting (Cambridge MC) is an international consulting firm that helps companies of all sizes have a better impact on the world. Founded in Cambridge, UK, initially to help the start-up community, Cambridge MC has grown to over 200 consultants working on projects in 25 countries. Our capabilities focus on supporting the private and public sector with their people, process and digital technology challenges.


What makes Cambridge Management Consulting unique is that it doesn’t employ consultants – only senior executives with real industry or government experience and the skills to advise their clients from a place of true credibility. Our team strives to have a highly positive impact on all the organisations they serve. We are confident there is no business or enterprise that we cannot help transform for the better.


Cambridge Management Consulting has offices or legal entities in Cambridge, London, New York, Paris, Dubai, Singapore and Helsinki, with further expansion planned in future. 

Contact Form

Contact - Craig Devolution Blog

Subscribe to our Newsletter

Blog Subscribe

SHARE CONTENT

Wind farms  and solar panels in the countryside at dawn
by Scott Armstrong 27 March 2026
Sustainability | Energy, risk and competitiveness – find out why sustainability is no longer just about reporting, but about resilience, cost control and long-term advantage | READ FULL ARTICLE
Yello and turquoise neon lights.
24 March 2026
International consulting firm, Cambridge Management Consulting has acquired telecommunications cost-reduction specialist, The Carrier Club, strengthening its ability to help organisations reduce their telecoms and network infrastructure costs.
Pembroke College lawn bathed in sunlight
by Tim Passingham 12 March 2026
CAMBRIDGE | See how Cambridge MC and Pembroke College are creating mutual value through a unique corporate partnership spanning student opportunities, academic collaboration and industry events | READ FULL CASE STUDY
Neon sharks made out of code.
by Simon Crimp 9 March 2026
Cyber Security | Ransomware in 2026 is a board-level resilience issue. Learn the key risks, weak spots and practical questions boards should ask to improve readiness, recovery and response.
The Top 21.2026 at the awards event in Cambridge, UK.
6 March 2026
The #21toWatch Top21.2026 winners have been announced at an awards ceremony at The Glasshouse innovation hub in Cambridge.
Asian business woman near a long window and looking at a tablet.
by Arianna Mortali 6 March 2026
BLOG | A student’s perspective on why women shouldn’t have to ‘play masculine’ to succeed at work – and how valuing empathy, confidence and inclusive leadership can help close gender gaps and build healthier organisations.
Abstract squiggle of circles
by Simon Crimp 19 February 2026
Where should leaders start with AI in 2026? A practical guide to moving beyond pilots, clarifying risk appetite, strengthening governance, improving data readiness, and delivering measurable enterprise value from AI at scale | READ FULL ARTICLE
Close up of a data centre stack with ports and wires visible
12 February 2026
We were approached by one of the fastest growing data centre providers in Europe. With over 20 data centres throughout the continent, they are consistently meeting the need for scalable, high-performance infrastructure. Despite this, a key data centre in Scandinavia had become reliant on a single, non-redundant 1 Gbps internet service from a local provider, posing significant risks to operational continuity. To enhance the reliability of its network and resolve these risks, our client needed to establish additional connectivity paths to ensure the redundancy of its infrastructure. The Ask Cambridge Management Consulting was engaged to address these connectivity challenges by identifying and evaluating potential vendors and infrastructure options to create second and third connectivity paths. This involved exploring various types of connectivity, including internet access, point-to-point capacity, wavelengths, and dark fibre. Additionally, Cambridge MC was asked to provide recommendations for building a local fibre network around the data centre to control and maintain diverse paths. This would allow the data centre to connect directly to nearby points of presence (PoPs) and reduce dependency on external providers, thereby enhancing network resilience and operational control. The goal of this project was to ensure that the Nordic data centre could maintain continuous operations even in the event of a failure in the primary connection. Approach & Skills Cambridge MC approached the project with a focus on ensuring operational continuity and resilience for the data centre. By identifying multiple connectivity paths, we aimed to mitigate the risk of network failures and ensure that the data centre could maintain continuous operations even in the event of a failure in the primary connection. This approach allowed Cambridge MC to provide a comprehensive solution to address both immediate and long-term connectivity needs. We employed a combination of Agile and Waterfall methodologies to manage the project. The initial investigative phase allowed a Waterfall approach, in which our team conducted thorough research and analysis to identify potential vendors and connectivity options. This phase involved detailed interviews with various telecommunications providers and an assessment of publicly available information. Once the initial analysis was complete, the workflow transitioned to an Agile approach for the implementation phase. This allowed Cambridge MC to adapt to new information and feedback from stakeholders, ensuring that the final solution was both flexible and robust. Challenges Lack of information: One of the primary obstacles we faced was the lack of detailed network maps and information from some of the potential vendors. To overcome this, the team conducted extensive interviews with contacts at these companies and leveraged its existing network of industry contacts to gather as much information as possible. Remote location: Another challenge was the remote location of the data centre, which limited the availability of local infrastructure and required us to explore creative solutions for connectivity. Cambridge MC addressed this by proposing the construction of a local fibre network around the data centre, which would allow for greater control and flexibility in connecting to nearby PoPs. Fragmented factors: Additionally, coordinating with multiple vendors and ensuring that their services could be integrated seamlessly posed a logistical challenge. We mitigated this by recommending a phased approach to implementation, starting with the most critical connectivity paths and gradually expanding to include additional options. Outcomes & Results Increased Connectivity: Cambridge MC successfully identified and evaluated multiple connectivity paths for the data centre. By exploring various types of connectivity, including internet access, point-to-point capacity, wavelengths, and dark fibre, we provided a comprehensive solution that significantly enhanced network resilience and reliability. Greater Control & Flexibility: Our recommendations for building a local fibre network around the data centre allowed for greater control and flexibility in connecting to nearby points of presence, ensuring continuous operations even in the event of a failure in the primary connection. New Vendors: The team’s extensive network of industry contacts and deep understanding of the regional telecommunications landscape allowed for a thorough and nuanced evaluation of potential vendors and connectivity options. Scope for Future Work: Cambridge MC identified several future developments with the potential to further enhance international connectivity and provide additional redundancy for the data centre. We also proposed further assistance, including a site visit for a more in-depth analysis of options, issuing RFI/RFP to vendors for capacity and fibre, and conducting similar connectivity studies for other candidate sites in the region.
Neon discs fading from blue to green to purple, cascading diagnolly across the screen.
by Cambridge Management Consulting 28 January 2026
Thames Freeport this week revealed the eight companies selected to participate in the Freeport’s Connectivity Lab, an initiative focused on validating commercially proven technologies in live port and logistics environments.
More posts