Design strategies for efficient, high-performance data centres

The rapid expansion of artificial intelligence workloads is placing unprecedented demands on data centre infrastructure. As computer densities increase and operational expectations tighten, the need to balance performance with energy efficiency and carbon reduction has become more urgent. This shift is driving a re-evaluation of how data centres are designed, particularly in relation to cooling strategies and overall resource use. Data centres are now a critical component of global infrastructure, supporting cloud services, digital platforms, and AI applications. With increasing digitalisation, energy consumption associated with these facilities continues to rise. In the UK and globally, regulatory and market pressures are also evolving, with greater emphasis on energy performance, carbon reporting, and long-term sustainability targets. Within this context, various industry reports are suggesting that: • Data centres are estimated to account for approximately 1–1.5% of global electricity consumption• High-density AI workloads can exceed 30–80 kW per rack, significantly increasing cooling demand• Leading facilities are targeting power usage effectiveness (PUE) values of 1.2 or lower Efficient cooling system strategies As computational loads increase, cooling systems are under growing pressure to maintain stable operating conditions without excessive energy use. Traditional approaches that rely heavily on mechanical cooling are becoming less viable due to their high energy intensity. This challenge affects operators, developers, and designers, particularly as expectations around efficiency and environmental performance continue to rise. BSE|3D, a UK building services engineering and consultancy practice, says it works with organisations navigating these challenges by applying a performance-led design approach from the earliest project stages. The company notes that it has observed that early integration of simulation tools allows for more effective alignment between building form, system design, and operational performance. Solutions that focus on reducing cooling demand at source while optimising system efficiency can significantly improve outcomes. This includes evaluating environmental conditions, refining building parameters, and developing strategies that prioritise low-energy operation. A key approach involves enabling a cooling profile where approximately 70% of annual demand can be met through low-energy systems such as economisation and adiabatic processes, with mechanical systems supporting peak conditions and operational resilience. This reduces reliance on continuous compressor use and supports improved overall performance. Kriti Gupta, Sustainability Consultant at BSE|3D, explains, “As data centre loads continue to increase, the industry needs to move beyond conventional cooling approaches. By prioritising low-energy strategies and validating them through simulation, it is possible to reduce energy demand while maintaining performance and resilience. Early-stage design decisions play a critical role in achieving this balance.” Data centres are expected to play an increasingly significant role in supporting digital infrastructure. As their impact grows, so too does the importance of designing them in a way that responds to both operational requirements and environmental considerations.

1547's Orangeburg data centre reaches full occupancy

Harrison Street Asset Management and fifteenfortyseven Critical Systems Realty (1547), a developer and operator of interconnected data centres and carrier hotels across North America, have completed the latest expansion phase of their Orangeburg data centre in New York, with the facility now fully leased and operating at near-full utilisation. The colocation site, located around 18 miles (28.9 kilometres) north of Manhattan, provides capacity for tenants requiring proximity to New York City and access to established connectivity routes. Originally supporting 3.7 MW of IT load when acquired in 2021, the joint venture has since added approximately 14 MW of capacity while increasing density across the existing 232,000ft² (21,553m²) facility. A further 12MW utility feed is currently under development, with additional long-term expansion plans in place. The site has outline approval for a new 230,000ft² (21,367m²) building, supported by a planned 60MW on-site substation. Expansion driven by connectivity demand The Greater New York data centre market remains one of the largest in the US, supported by multiple terrestrial fibre routes and subsea cable landings along Long Island and New Jersey, enabling international connectivity, particularly with Europe. The Orangeburg facility now supports around 18 MW of IT load and has reached near-full utilisation following recent leasing activity. Demand is primarily driven by financial services organisations, including banks, trading platforms, and hedge funds, which require low-latency connectivity to Manhattan. Michael Hochanadel, Head of Digital Assets at Harrison Street Asset Management, comments, "The Orangeburg data centre exemplifies our approach to digital infrastructure investing, pairing strategic locations with disciplined demand-driven expansion." J Todd Raymond, Chief Executive Officer and Managing Director of 1547, adds, "From day one, our focus has been on delivering capacity in direct response to customer demand while maintaining the performance and reliability our clients depend on." Since 2018, Harrison Street Asset Management’s digital investment platform has committed more than $6.5 billion (£4.8 billion) to data centre and connectivity infrastructure, including powered shells, carrier hotels, colocation facilities, and dark fibre networks. For more from 1547, click here.

Daikin expands UK HVAC rental fleet

Daikin Rental Solutions, the dedicated temporary HVAC hire division of Daikin Applied UK, has expanded its UK rental fleet with additional cooling, heating, and air handling equipment, responding to growing demand for temporary HVAC systems. The investment reflects pressure on organisations to maintain operations while managing ageing infrastructure, higher temperatures, regulatory requirements, and the shift towards lower-carbon energy. The company has increased its range of high-capacity cooling systems, process cooling equipment, and heat pumps, alongside introducing UK-manufactured air handling units. Broader HVAC capability for UK industries Demand for temporary HVAC continues to rise across sectors such as data centres, healthcare, and manufacturing, where system downtime can have operational consequences. To support this, Daikin has added higher-capacity chiller units and an expanded range of heat pumps, offering scalable cooling and heating for larger or more complex environments. Additional dry air coolers have also been introduced to support process cooling requirements, including systems designed to improve energy performance under suitable conditions. The company has also expanded its airside equipment with a new range of air handling units manufactured in the UK. Produced at Daikin Applied UK’s facility in Cramlington, the units are designed for rental use, with a focus on ease of installation and suitability for environments such as cleanrooms and healthcare settings. Mike England, UK Rental Sales Manager at Daikin Applied UK, says, "In many of the sectors we support, downtime simply isn’t an option. Customers need solutions that are not only available quickly, but that perform reliably and integrate seamlessly into their existing systems." The company states that manufacturing air handling units within the UK is intended to improve equipment availability and reduce lead times, while maintaining consistent engineering standards across its rental fleet. Digital monitoring and control features are available on selected systems, supporting maintenance and operational oversight where required. For more from Daikin, click here.

€50bn Croatia AI data centre investment announced

Pantheon Atlas, a transatlantic-led investment group, has announced plans to develop a hyperscale AI data centre and innovation campus in Topusko, Croatia, with total investment expected to exceed €50 billion (£43 billion). This is reportedly the largest investment of its kind in Croatian history and among the largest private US investments in Europe. The project, known as Pantheon AI, is intended to address growing demand for AI-driven data centre capacity across Europe, where availability of power, land, and construction resources remains constrained. The development is being delivered by a transatlantic investment group combining US capital with local expertise in Croatia, including regulatory and grid access experience. The announcement was made at the Three Seas Initiative Summit in Dubrovnik. Pantheon AI is designed to meet NVIDIA’s gigawatt-scale AI factory standards and is expected to offer high levels of availability, exceeding Tier IV benchmarks. Jako Andabak, Founding Partner at Pantheon AI, comments, "Pantheon AI is a signal to the world that Croatia is open for the highest-caliber investment. "This project is the culmination of years of work to bring world-class digital infrastructure to Croatia." Addressing European data centre capacity Across Europe, established data centre markets are operating with limited vacancy, while grid connection delays continue to affect new developments. Demand in Central and Eastern Europe is expected to increase significantly by 2035, particularly as AI workloads expand and regulatory requirements encourage data to be stored within EU borders. Ryan Rich, Managing Partner at Pantheon AI, explains, "We have assembled a transatlantic partnership to solve one of the most pressing challenges in global digital infrastructure: enabling hyperscale operators to meet AI-driven demand at scale." The project is expected to support up to 5.2 GW of renewable energy integration into Croatia’s grid. It will include an on-site solar installation and battery storage, alongside multiple fibre connections across European network corridors. Joshua Volz, Special Envoy for Global Energy Integration at the US Department of Energy, says, "Critical infrastructure of this scale, built by the private sector responding to real market demand, is exactly how US interests and European security advance together." Construction of the campus is scheduled to begin in early 2027, with operations expected to start in the first quarter of 2029. The initial phase represents a €12 billion (£10 billion) investment, with additional funding anticipated as tenants deploy infrastructure. The campus will have a planned capacity of 1GW, including 800MW of usable IT load, and will span approximately 310 acres (1.2 km²), with expansion potential. The development is expected to create around 1,500 permanent roles, alongside 3,000 jobs during construction.

Data centres 2026: Energy efficiency and sustainability

Prospero Events Group says it is proud to present the 3rd Energy Efficiency & Sustainability in Data Centers 2026 conference, taking place on 26–27 May 2026 in Amsterdam, the Netherlands. The event will bring together senior decision-makers, technology innovators, and sustainability leaders to explore how data centres can become more energy efficient, resilient, and environmentally responsible. Event presentation topics and speaker panel As AI adoption, hyperscale growth, and electrification continue to increase power demand, the conference will focus on the strategies needed to balance performance with sustainability. Key discussion areas include: • Renewable energy integration• Cooling efficiency and uptime• Waste heat reuse and district heating• On-site energy generation and energy resilience• ESG integration across the data centre lifecycle• Grid constraints and power availability• AI-driven optimisation and future-ready infrastructure The conference will feature expert insights from leading organisations, including: • Günter Eggers, Director Public, NTT Global Data Centers• Vladimir Prodanovic, Principal Program Manager, NVIDIA• Pedro Filipe Barreiros, Data Center Operations, Google• Stijn Grove, Managing Director, Dutch Data Center Association• Martijn Van Wijngaarden, Global Energy Transaction Manager, Iron Mountain• Simon Muskett and Olalekan Salami, Digital Realty The event offers a valuable platform for collaboration amongst operators, investors, technology providers, and sustainability experts committed to building the next generation of low-impact, high-performance data centres. Explore the complete speaker lineup, session topics, and key discussion points shaping the future of sustainable data centres. To register and download the event agenda, click here.

Veolia, Amazon develop data centre water reuse system

Veolia, a French multinational environmental services company, is working with US technology and e-commerce company Amazon to introduce reclaimed water for cooling at a data centre in Mississippi, USA, as part of efforts to reduce water use and support long-term water resilience. The facility, expected to be operational in 2027, will be the first Amazon data centre in the state to use treated wastewater for cooling processes. The system will convert effluent from nearby treatment plants into water suitable for industrial use. Once fully operational, the project is expected to reuse more than 83 million gallons (313 million litres) of potable water each year, reducing demand on local groundwater and drinking water supplies. Veolia will deploy modular, containerised treatment systems designed for scalable use. The approach allows similar installations to be introduced at other data centres where conditions allow. Turning wastewater into cooling power for data centres The collaboration also includes the use of artificial intelligence to improve water treatment processes. Amazon Web Services (AWS) will support Veolia in developing systems for real-time optimisation, predictive maintenance, and operational analysis. These tools are intended to improve efficiency and reduce resource consumption across water treatment operations. Estelle Brachlianoff, Chief Executive Officer at Veolia, comments, “We are delighted to collaborate with Amazon to secure its water needs in Mississippi while protecting the local community's resources. This is environmental security in action. "By combining Veolia’s water expertise with Amazon’s AI technologies, we’re transforming data centres into engines of innovation for sustainability. This solution builds on our newly launched offering for data centres.” Will Hewes, Global Water Stewardship Lead at Amazon, adds, “Through our collaborative work on AI applied to water treatment, Veolia will be able to further drive innovation and enhance the efficiency of on-site teams, thanks to automated analytics, actionable recommendations, optimised inventory management, and streamlined maintenance. "We’re pleased to join forces with Veolia to advance more sustainable water use strategies while helping it pioneer more efficient water treatment solutions for customers worldwide.” The project, Amazon says, forms part of its wider aim to become water positive across its direct data centre operations by 2030.

Vertiv acquires Strategic Thermal Labs

Vertiv, a global provider of critical digital infrastructure, has acquired Strategic Thermal Labs (STL), a company specialising in liquid-cooling technologies for high-density computing environments. The acquisition is intended to strengthen Vertiv’s engineering capabilities in managing the interaction between server-level liquid cooling and supporting infrastructure, which is becoming increasingly important in AI and high-performance computing deployments. Strategic Thermal Labs brings experience in cold-plate design, server-side liquid cooling, and thermal validation for high-density systems. This is expected to support Vertiv’s ability to simulate real-world operating conditions and improve the integration of thermal and power systems. Acquisition targets high-density cooling challenges As computing workloads become more intensive, thermal management at chip level is playing a greater role in overall system performance and reliability. The addition of STL is aimed at improving design, integration, commissioning, and long-term operation of liquid-cooled environments. Scott Armul, Chief Product and Technology Officer at Vertiv, comments, “As AI and high-performance computing push power densities to unprecedented levels, understanding and solving heat challenges at the chip level becomes critical to system design, performance, and reliability. "STL brings deep expertise and proven capability in addressing some of the industry’s most demanding chip-level density and thermal problems, strengthening Vertiv’s ability to emulate and validate system-level solutions and enabling customers to improve performance and lifecycle outcomes in liquid-cooled environments.” Vertiv states that the acquisition will not change its approach to supporting interoperable infrastructure and the company will continue to work with a range of server and silicon platforms. It adds that the move forms part of its wider strategy to address increasing infrastructure complexity through integrated power, thermal, and lifecycle capabilities. For more from Vertiv, click here.

Fluke warns AI boom exposes 'confidence crisis'

As artificial intelligence (AI) demand accelerates, new research from Fluke Corporation, a US manufacturer of electronic test and measurement tool, suggests a growing confidence crisis among data centre professionals, raising concerns about the sector’s ability to scale reliably. A survey of more than 150 data centre professionals, conducted at Data Centre World London 2026, found that only 22% fully trust that their test and measurement data reflects real-world operating conditions. Confidence drops further under pressure, with just 19% expressing full trust in data accuracy during peak load or failure scenarios. Several factors are driving this lack of confidence in infrastructure data. Skills and training gaps were cited as the biggest barrier (43%), followed by time pressures during commissioning (16%), inconsistent testing processes (11%), and budget constraints (10%). The operational impact is seemingly already being felt. Half of respondents reported experiencing unplanned outages or major performance disruptions at least annually, with nearly one in five experiencing disruptions as frequently as monthly (10%) or weekly (8%). Outdated testing equipment is compounding the issue, with nearly two thirds (65%) saying legacy tools increase the risk of downtime and compliance failures within their organisation. Speed vs compliance trade-offs emerge The research exposes a widening gap between intent and execution. While almost all respondents agree that regular maintenance is critical to reducing downtime, only 28% have real-time or predictive monitoring in place across critical infrastructure such as power, cooling, and networks. One fifth admit maintenance is conducted quarterly at most. Adoption of advanced technologies also reportedly remains limited. Just 10% have fully implemented automation, AI diagnostics, or predictive monitoring, while many remain in pilot (22%) or early-stage (19%) phases. Pressure to deliver data centre capacity faster is also creating new risks. 42% of respondents said time pressures create occasional compliance risks, while 17% said they make it significantly harder to meet evolving connector and certification requirements. Mike Slevin, Director of EMEA Market at Fluke Corporation, comments, “What’s striking here is that organisations already know what needs to be done. There’s broad recognition that regular maintenance and better monitoring are critical to reducing downtime, yet, in practice, adoption is lagging. “That gap between awareness and action is where risk builds. When testing isn’t consistent and monitoring isn’t real-time, small issues can quickly escalate into outages.” UK readiness in question as AI ambitions grow The findings also cast doubt on the UK’s ability to support its ambitions to become a global AI leader. Only half of respondents believe the UK data centre sector is operationally ready to scale for AI, cloud, and hyperscale demand over the next five years. Additionally, just 7% believe the UK currently has the infrastructure resilience and operational standards required to support its “AI superpower” ambitions, with 28% pointing to significant infrastructure gaps. Mike continues, “AI is redefining the demands placed on data centre infrastructure. With higher-density architecture and increasingly complex fibre environments, multi-fibre testing has become paramount as the margin for error narrows. “If organisations can’t confidently validate performance under real-world conditions, they risk building AI on unstable foundations. The challenge now is ensuring that capacity is resilient and ready for sustained demand.” For more from Fluke, click here.

Report finds what's slowing down DC planning permission

A new report by Hoare Lea, a UK engineering consultancy, has found that data centre planning applications in the United Kingdom are being delayed by an average of 490 days, driven largely by objections related to inadequate community engagement, unclear community benefits, design, infrastructure constraints, and energy use. Hoare Lea’s Societal Insights team analysed 33 disputed applications to understand the underlying reasons for rejection and delay. Rejections frequently cited policy non-compliance, unsuitable locations, and insufficient energy strategies. These findings identify the tension between the growing demand for data centres in the UK - infrastructure that is supported by the Government’s AI Growth Zones - and the realities of navigating the planning system. Carl Walker, Head of Societal Insights at Hoare Lea, comments, “Success will be judged not only by new infrastructure, but also by the skills, growth, and opportunities delivered to local communities, [whose needs] must be recognised in the planning process.” Ambitions to build data centres in the UK already face significant challenges. Electricity grid capacity is already under strain, particularly in London and the M4 corridor, where data centres have delayed housing developments. Concerns also exist around environmental impacts, green belt land, and controversial government interventions overriding local planning decisions. The report findings underscore the need for integrated approaches that combine renewable energy planning, transparent governance, and meaningful community dividends. It suggests that by engaging communities early, supporting local skills and infrastructure, and embedding social and environmental value, data centres can become catalysts for sustainable local growth rather than sources of conflict. To view the full report, click here.

How 5G-A GigaUplink kept a robot half-marathon on track

Chinese telecommunications operator China Unicom and Chinese multinational technology company Huawei's landmark deployment at the 2026 Beijing E-Town event demonstrates what high-uplink network architecture can deliver when the stakes - and the robot runners - are at full sprint. When 21 humanoid robots lined up alongside human runners at the 2026 Beijing E-Town Half-Marathon on 19 April, the spectacle represented far more than a novelty sporting occasion. Behind the scenes, a sophisticated 5G-A network was working hard to ensure that each machine could navigate, sense its environment, and communicate in real time - all while travelling at speeds of up to 10 metres per second. The event, held at Tongming Lake Park and Nanhaizi Park in Beijing's E-Town (Yizhuang) district, marked a genuine global first: human runners and humanoid robots competing simultaneously on the same course. It also served as a compelling proof of concept for how high-uplink 5G-A connectivity can solve some of the most demanding real-world engineering challenges in embodied AI. The engineering challenge Running a humanoid robot half-marathon is not simply a mechanical feat. Each robot must receive and transmit a continuous stream of HD video, environmental sensor data, and positioning information. Remote-control robots require stable, low-latency command links, while the autonomous navigation group demands even more: real-time decision-making support, precise positioning to sub-decimetre accuracy, and dynamic obstacle avoidance - all at pace, over a 21km course, in a densely populated public space. Add to that the media requirements - 4K and 8K live broadcasting, real-time data feeds, and tens of thousands of spectators uploading content simultaneously - and the uplink demands become extraordinary. Each robot alone requires approximately 10 Mbps of sustained uplink bandwidth, and the network must support more than 40 concurrent service scenarios without degradation. In short, this was not a use case that a conventional network architecture was built to handle. The network solution China Unicom Beijing, supported by Huawei's radio access technology, deployed a 5G-A premium network along the full length of the course. The solution centres on 5G-A three-component carrier (3CC) aggregation across 3.5GHz and 2.1GHz bands, using Huawei's Extended Large Aperture Array (ELAA) technology. This combination delivers a peak uplink speed of 677 Mbps and an average of 155 Mbps, with end-to-end latency averaging just 30 ms. Crucially, uplink speeds of 20 Mbps or above were maintained more than 99.6% of the time - a figure that underlines the reliability required for safety-critical autonomous systems operating at speed. Dedicated uplink network slices were reserved along the course to prioritise robot connectivity and guarantee positioning accuracy to sub-decimetre level. At the high-density start and finish areas, carrier priority technology ensured that robot and media traffic was not crowded out by spectator demand. The result was consistent, stable connectivity for both the autonomous navigation robots and the remote-control group throughout the full 21km route. Underpinning the physical network is an intelligent operations and maintenance (O&M) platform, which uses AI algorithms to monitor equipment remotely, predict faults at millisecond timescales, and optimise resource allocation dynamically. By combining dynamic and static management, the system enables continuous, unattended operation and real-time adaptation to shifting traffic patterns across the course. China Unicom has built this into a broader ICT-integrated smart operations architecture - one that gives the network, as the company puts it, a "smart brain". What it delivered The results speak clearly to what a well-engineered, high-uplink network can achieve. Robots in the autonomous navigation group executed precise perception and rapid decision-making throughout the course, supported by sub-decimetre positioning accuracy and guaranteed low-latency data paths. Live broadcast teams delivered 4K and 8K feeds without interruption, and spectators experienced consistently smooth connectivity despite the event's scale. Field testing recorded a peak uplink speed of 677 Mbps, while the 99.6% fulfilment rate at 20 Mbps confirmed that the headline performance figures translated into reliable real-world delivery. The network also supported innovative spectator applications, including AR and VR viewing modes that allowed fans to experience the marathon from the robots' own perspectives - all enabled by the same high-uplink infrastructure. In their own words Qin Yang, Deputy General Manager of China Unicom in Beijing, spoke proudly of the excellent connections his company provided for the 2025 World Humanoid Robot Games, as a global strategic partner, and for the 2026 Beijing E-Town Half-Marathon event, as the exclusive official communications sponsor. He said, "5G-A and AI are essential digital infrastructure, enabling us to bring embodied AI to sports. Given the new dynamics of AI development, we will double down on our priorities over connectivity, computing power, services, and security to sharpen our competitive edges as a preferred telecom partner for intelligent sports and a core enabler for intelligent industry transformation. "Moving forward, we will accelerate our innovation-driven push to strengthen our digital infrastructure and drive the high-quality growth of embodied AI in China and beyond." Samuel Chen, Vice President of Huawei's Wireless Network Business Marketing, added that this humanoid robot half-marathon offers a good example of deepening integration between mobile technology and embodied intelligence. He commented, "Beyond redefining connectivity for sports, it has shown us what intelligent production and life will look like in future. We are always dedicated to building excellent 5G-A networks together with operators based on user-centred innovation to ensure GigaUplink, low latency, and high reliability for differentiated mobile AI services. This will enable us to continuously drive the high-quality growth of the digital economy." The broader picture The Beijing event is significant not only as a spectacle, but as an indicator of where network engineering is heading. The shift from downlink-dominant 5G towards symmetrical, high-uplink architectures reflects a fundamental change in how AI-driven applications interact with mobile infrastructure. Embodied AI, industrial data backhaul, and immersive real-time services all place enormous demands on uplink capacity - demands that traditional LTE and early 5G networks were not designed to meet. The 5G-A GigaUplink standard, with its target of ubiquitous 20Mbps uplink availability, positions the technology as a credible foundation for the next generation of intelligent applications. China Unicom's deployment at the E-Town marathon demonstrates that this is not merely theoretical; with the right network architecture, AI-powered autonomous systems can operate reliably at scale in complex, real-world environments. For network engineers and infrastructure professionals, the lesson from Beijing is clear: As embodied AI moves from the laboratory to the streets, connectivity infrastructure must evolve to match - and the 5G-A deployment at the E-Town marathon offers a practical, field-tested blueprint for how that evolution can be achieved. For more from Huawei, click here.