The promise of artificial intelligence (AI) is enlightenment. The pressure it places on infrastructure is far less elegant.

Across every layer of the data centre stack, AI is exposing structural limits – from cooling thresholds and power capacity to build timelines and failure modes. What many operators are now discovering is that legacy models, even those only a few years old, are struggling to accommodate what AI-scale workloads demand.

This isn’t simply a matter of scale – it is a shift in shape. AI doesn’t distribute evenly, it lands hard, in dense blocks of compute that concentrate energy, heat and physical weight into single systems or racks. Those conditions aren’t accommodated by traditional data hall layouts, airflow assumptions or power provisioning logic. The once-exceptional densities of 30kW or 40kW per rack are quickly becoming the baseline for graphics processing unit- (GPU) heavy deployments.

The consequences are significant. Facilities must now support greater thermal precision, faster provisioning and closer coordination across design and operations. And they must do so while maintaining resilience, efficiency and security.

Design under pressure

The architecture of the modern data centre is being rewritten in response to three intersecting forces. First, there is density – AI accelerators demand compact, high-power configurations that increase structural and thermal load on individual cabinets. Second, there is volatility – AI workloads spike unpredictably, requiring cooling and power systems that can track and respond in real time. Third, there is urgency – AI development cycles move fast, often leaving little room for phased infrastructure expansion.

In this environment, assumptions that once underpinned data centre design begin to erode. Air-only cooling no longer reaches critical components effectively, uninterruptible power supply (UPS) capacity must scale beyond linear load, and procurement lead times no longer match project delivery windows.

To adapt, operators are adopting strategies that prioritise speed, integration and visibility. Modular builds and factory-integrated systems are gaining traction – not for convenience, but for the reliability that controlled environments can offer. In parallel, greater emphasis is being placed on how cooling and power are architected together, rather than as separate functions.

Exploring the physical gap

There is a growing disconnect between the digital ambition of AI-led organisations and the physical readiness of their facilities. A rack might be specified to run the latest AI training cluster. The space around it, however, may not support the necessary airflow, load distribution or cable density. Minor mismatches in layout or containment can result in hot spots, inefficiencies or equipment degradation.

Operators are now approaching physical design through a different lens. They are evaluating structural tolerances, rebalancing containment zones, and planning for both current and future cooling scenarios. Liquid cooling, once a niche consideration, is becoming a near-term requirement. In many cases, it is being deployed alongside existing air systems to create hybrid environments that can handle peak loads without overhauling entire facilities.

What this requires is careful sequencing. Introducing liquid means introducing new infrastructure: secondary loops, pump systems, monitoring, maintenance. These elements must be designed with the same rigour as the electrical backbone. They must also be integrated into commissioning and telemetry from day one.

Risk in the seams

The more complex the system, the more attention must be paid to the seams. AI infrastructure often relies on a patchwork of new and existing technologies – from cooling and power to management software and physical access control. When these systems are not properly aligned, risk accumulates quietly.

Hybrid cooling loops that lack thermal synchronisation can create blind spots. Overlapping monitoring systems may provide fragmented data, hiding early signs of imbalance. Delays in commissioning or last-minute changes in hardware specification can introduce vulnerabilities that remain undetected until something fails.

Avoiding these scenarios requires joined-up design. From early-stage planning through to testing and operation, infrastructure must be treated as a whole. That includes the physical plant, the digital control layer and the operational processes that bind them.

Physical security under AI conditions

As infrastructure becomes more specialised and high-value, the importance of physical security rises. AI racks often contain not only critical data but hardware that is financially and strategically valuable. Facilities are responding with enhanced perimeter control, real-time surveillance, and tighter access segmentation at the rack and room level.

More organisations are adopting role-based access tied to operational state. Maintenance windows, for example, may trigger temporary access privileges that expire after use. Integrated access and monitoring logs allow operators to correlate physical movement with system behaviour, helping to identify unauthorised activity or unexpected patterns.

In environments where automation and remote management are becoming standard, physical security must be designed to support low-touch operations with intelligent systems able to flag anomalies and initiate response workflows without constant human oversight.

Infrastructure as an adaptive system

The direction of travel is clear. Infrastructure must be able to evolve as quickly as the workloads it supports. This means designing for flexibility and for lifecycle. It means understanding where capacity is needed today, and how that might shift in six months. It means choosing platforms that support interoperability, rather than locking into closed systems.

The goal is not simply to survive the shift to AI-scale compute. It is to build a foundation that can keep up with whatever comes next – whether that is a new training model, a change in energy market conditions, or a new set of regulatory constraints.

How is the acceleration of AI development shifting into a new phase? And what effect is that having on our demand for data centre infrastructure?
We’re seeing a move from experimentation to deployment at scale. AI is no longer something that sits in a lab or a discrete cluster. It’s being integrated into core business systems and running continuously, which changes what infrastructure is expected to deliver.

The key shift is intensity. Workloads are denser, more power-hungry and less predictable. This means data centres can’t rely on older assumptions around capacity, load distribution or response time. They need to be designed for higher variability, as well as for higher volume.

It feels like data centres need to deliver more power, cooling, space – everything – faster than expected using infrastructure that is either unprepared or hasn’t been built yet. How does the industry contend with these challenges?

It starts with mindset. You can’t meet today’s pace with yesterday’s approach. Operators are moving towards prefabricated modular infrastructure, shorter design-to-deploy timelines, and more integrated delivery models. Prefabrication helps and can reduce deployment time by up to 50%. So does standardising the way cooling, power and racks are designed, manufactured and assembled in a standardised factory environment, simultaneously, rather than in sequence.

Another strategy that is key to being prepared for what’s next is collaboration across the industry. For example, our strategic partnership with NVIDIA. Vertiv has worked with NVIDIA on the end-to-end power and cooling reference design for both the NVIDIA GB200 NVL72 and the GB300 NVL72 platforms. By staying one GPU generation ahead, our customers can plan for future infrastructure before the silicon lands, with deployment-ready designs that anticipate increased rack power densities and repeatable templates for AI factories at scale. 

How do we deal with the discrepancy in development cycle speeds between AI and the infrastructure used to house it?

This is one of the biggest structural mismatches the industry faces. AI development is sprinting. Infrastructure is still built on marathon timelines. Speed is critical and densities are different. Therefore, a change of philosophy is needed when it comes to data centre design and build.

The new AI factories need to be ready much faster than we’ve ever seen before in the industry. By standardising everything including cooling and power distribution, critical infrastructure can be deployed at speed rather than needing to retrofit what already exists or build from scratch, which can reduce timelines significantly. 

On the energy side of things, do you expect data centres to take on a new role in relation to the grid, especially as some economies work to further electrify in pursuit of net zero goals?

Yes. The old model – draw power and provide backup – is shifting. It’s no secret that data centres are prioritising energy availability challenges. Overextended grids and increasing power demands are changing how data centres consume power. Many large facilities now operate as part of the wider energy system, helping manage peak demand or stabilise frequency through intelligent battery usage or flexible loads.

Data centre operators are seeking energy solutions that enable them to minimise generator starts and reduce energy costs and reliance on the grid. Microgrids integrated with uninterruptable power supply (UPS) offer a promising solution, enabling power reliability, stabilising renewable fluctuations, and protecting critical loads. They can also provide ancillary services to the main grid, such as frequency regulation and enhance grid stability by participating in demand response and load shedding.

This is being driven partly by policy and partly by economics. As electricity becomes a more valuable and volatile resource, infrastructure that can respond dynamically will be better placed to operate cost-effectively – and in some regions, to operate at all.

On the component side of things, how is the new generation of GPUs and other internal server equipment geared towards AI changing the way data centres need to be built?
Newer GPUs and high-bandwidth interconnects are driving heat and power requirements far beyond traditional design envelopes. A rack that previously ran at 10kW might now need 50kW to 100kW or more, and forecasts indicate this may increase to 300-600kW and possible 1MW by 2030 – this changes the physical reality of the room. This means that densification is required – it’s about making sure that there is more compute in as little footprint as possible.

The newer GPUs generate far more heat, so cooling systems need to become more targeted. Airflow alone is rarely sufficient, making direct liquid cooling, cold plates or hybrid systems necessary. Cable management, power infrastructure and weight loading also shift. Even the spacing between cabinets can affect thermal performance. This could involve a redesign from the inside out or layering new kit into old frameworks.

Can you talk about Vertiv’s work with Intel and NVIDIA on cooling systems? What’s the benefit of a dual system over a pure liquid-cooled facility, for example?
Vertiv has co-developed reference architectures with both Intel and NVIDIA to address next-generation AI workload demands. For NVIDIA’s GB200 NVL72, Vertiv released a 7 MW reference architecture supporting rack densities up to 132 kW. This includes a hybrid system that combines liquid cooling for prime heat sources with air cooling for supporting infrastructure.

For Intel’s Gaudi3 platform, Vertiv validated designs capable of handling 160 kW using pumped two-phase (P2P) liquid cooling, alongside traditional air-cooled setups up to 40 kW. 

Hybrid cooling systems are based on a clear set of technical and operational frameworks: 

Component-level thermal targeting

Liquid cooling – direct-to-chip cold plates or rear-door exchangers – focuses precisely on AI accelerators. This means airflow systems only need to support peripheral equipment, improving overall energy use and avoiding over-engineering the facility.

Phased deployment and flexibility

Hybrid architectures allow gradual ramping up of liquid cooling infrastructure. 

For smooth upgrades, it’s important to design systems that can accommodate higher liquid temperatures from the start.  Operators can begin with air cooling, introduce liquid in hot zones, and expand as capacity needs grow. 

Operational compatibility

These designs support mixed workloads – GPU clusters, CPUs, storage – in the same white space by delivering the cooling each requires without impacting others.

End-to-end deployment frameworks

Vertiv’s reference architectures include detailed layouts: fluid routing, rack spacing, containment strategies, plus commissioning protocols. The NVIDIA frameworks are factory-tested and SimReady via digital twins, significantly reducing onsite uncertainty. 

These hybrid frameworks offer precise thermal control, deployment agility, resilience, and simplified operations. Essentially, they merge the benefits of both air and liquid cooling into a scalable and AI-ready model.

How does AI change the ways in which data centres are likely to require maintenance or even fail? What kind of adjustment will this require on the part of the industry?

The criticality definitely increases. AI systems tend to concentrate compute in fewer, more critical pieces of hardware, so if one component overheats or fails, the impact can cascade faster, disrupting the computational workload it supports.  Thermal margin is tighter, fluid networks introduce new points of failure, and real-time monitoring becomes more important, not just for performance but for reliability.

This means more condition-based maintenance, more granular telemetry, and stronger alignment between IT and facilities teams. It also requires a different mindset – from reacting to faults, to proactively managing infrastructure health in real time.

In recent months, data sovereignty is once again in the spotlight for the world’s digital businesses and governments seeking to mitigate against uncertain economic and geopolitical environments. Knowing exactly where an organisation’s data is stored, and what country’s legal and compliance requirements governs this, means that a defined data sovereignty strategy should be a key business priority that warrants careful consideration at the most senior level. Failure to execute this could have wide reaching consequences including fines for non-compliance, business disruption and damage to reputation.

Currently, nowhere is more of a hotbed for debate on this than in Europe, where there is a strong drive to build a resilient and self-sufficient digital infrastructure. A key foundation for establishing this successfully is the ability to store and secure data under European jurisdiction. And with businesses of every size heavily reliant on cloud-based services headquartered outside of Europe, this is creating a sense of unease amongst leaders that they must rapidly address the operational and legal ambiguities this raises.

A European cloud for a trusted digital economy

In the UK alone, a recent survey found that more than 60% of the UK’s IT tech leaders feel the government’s use of US cloud services leaves the country’s digital economy vulnerable to a variety of risks. For example, further exacerbated by the announcements on US tariffs, a whirlwind of ever-changing trade policies and US laws such as the CLOUD Act (Clarifying Lawful Overseas Use of Data Act) that could oblige large American cloud providers to provide data to US authorities no matter the geography in which this is stored concerns over the security and sovereignty of data have been.

These sentiments are echoed across Europe, with momentum building to establish a secure, resilient and sovereign cloud for the continent. This is demonstrated by the EU’s Important Projects of Common European Interest on Cloud Infrastructure and Services (IPCEI-CIS), a notable programme to create a sovereign European cloud campus to protect data under EU regulations and ensure that data physically stored in Europe’s boundaries is far less dependent on US providers.

In today’s environment, it is no wonder that locally governed data storage services are an increasingly attractive option, with specialist European providers as well as the large hyperscalers such as Azure and AWS, actively invested in the effort to make this happen. IPCEI-CIS is backed by more than 100 organisations, not only to achieve regulatory compliance with EU laws such as GDPR, but the aim is also to support technology innovation and digital growth throughout the region.

A critical and strategic matter for all digital businesses

Data sovereignty has broad reaching implications with potential impact on many areas of a business extending beyond the IT department. One of the most obvious examples is for the legal and finance departments, where GDPR and similar legislation require granular control over how data is stored and handled. 

The harsh reality is that any gaps in compliance could result in legal action, substantial fines and subsequent damage to longer term reputation. Alongside this, providing clarity on data governance increasingly factors into trust and competitive advantage, with customers and partners keen to eliminate grey areas around data sovereignty. 

With so much at stake, it is no longer acceptable for there to be any doubt about what jurisdiction data falls under. While once perceived as an issue for large global corporates, the fact is that any size of digital business using a cloud infrastructure now needs to plan meticulously for where its data is stored, and the legal implications of this. 

Arguably, it is smaller businesses that face their own set of challenges in understanding data sovereignty requirements. Unlike multinationals, smaller organisations commonly do not have the specialist legal and IT resources at their fingertips to advise on cross-border data policies. Instead, they often turn to third party cloud providers and are reliant on these partners to provide sound counsel on data legislation and organisation.

Why repatriate data?

One way that many companies are seeking to gain more control and visibility of their data is by repatriating specific data sets from public cloud environments over to on-premise storage or private clouds. This is not about reversing cloud technology; instead, repatriation is a sound way of achieving compliance with local legislation and ensuring there is no scope for questions over exactly where data resides.

In some instances, repatriating data can improve performance, reduce cloud costs and it can also provide assurance that data is protected from foreign government access. Additionally, on-premise or private cloud setups can offer the highest levels of security from third-party risks for the most sensitive or proprietary data. 

Implementing sovereign-readiness

The rule of thumb now for any business is that if it’s not crystal clear about where your data is stored and what country governs this, it is essential to take action.

Although every organisation will ultimately choose its own path towards data sovereignty, action is needed now to fully understand where and how data is stored and how to bring it home if necessary. Many organisations will seek out a partner that can help restructure their operations to suit data storage needs and ensure this is compliant with local laws.

That partner should be able to provide transparent and specific details on data handling; for example, offering assurance that data is physically located in a UK or French data centre, and that a data centre provider is compliant with regulations such as GDPR. Providers should also offer more than basic service, with the ability to offer in-depth and proactive consultancy, and end-to-end security to protect data against external threats. 

For many companies, choosing the right partner will make all the difference to being truly sovereign ready or falling short of this. In a world beset with geopolitical and economic uncertainties, it is no surprise that Europe is heavily invested into a sovereign cloud that will underpin and enable its future digital economy. 

Every company can – and should – play its part in this now by asking tough questions about its own data. Being truly ready means knowing data location, who can access this and what legislation it is governed by. In this way, every business can align itself with Europe’s ambitions to foster the continent’s long-term digital ecosystem.

For the past decade, enterprises have been guided by a prevailing assumption. In the 2010s, conventional wisdom became that the future of IT infrastructure belonged to hyperscale cloud providers. The argument was compelling – unmatched scalability, rapid deployment and reduced capital expenditure. But as artificial intelligence (AI), high-performance computing (HPC) and cost volatility fundamentally reshape the landscape, enterprises are shifting from a cloud-first mindset to a more nuanced approach, blending public cloud with private and colocation solutions.

This is not a retreat from hyperscale cloud providers but rather an evolution in enterprise strategy. Businesses are now recognising that no single approach fits every workload. Instead, they are focusing on aligning workloads with environments that offer the best combination of cost, performance and control.

The Changing Economics of Cloud and AI Workloads

Public cloud made financial sense when workloads were dynamic and unpredictable, and when enterprises sought to avoid the capital outlays of on-premise infrastructure. However, the cost dynamics are shifting, especially for sustained, compute-intensive applications such as AI training and inference.

Hyperscale providers offer AI-optimised instances. However, enterprises are discovering that ongoing AI workloads incur high operational costs compared to predictable, long-term investments in private infrastructure or colocation. As a result, many organisations are evaluating hybrid models. These models use colocation for cost-predictable, high-performance workloads. At the same time, they leverage the public cloud for burst capacity and distributed applications.

Beyond cost, latency and data gravity, regulatory considerations are making private and hybrid environments more attractive. When data volumes are large and constantly processed – such as in AI model training, real-time analytics or financial trading – keeping workloads closer to their data sources in private or collocated infrastructure can improve efficiency and compliance.

Reassessing Private Infrastructure

The resurgence of private and hybrid cloud does not mean a return to outdated models of IT ownership. Instead, it reflects a growing emphasis on performance-driven infrastructure decisions.

Enterprises are leveraging colocation and private cloud for several reasons:

• Workload optimisation: Not all applications benefit from the shared infrastructure model of public cloud. High-performance AI training, real-time applications and compliance-heavy workloads often require dedicated, optimised resources.
• Operational predictability: Cloud pricing models, with their unpredictable egress costs and variable compute rates make budgeting challenging for enterprises running sustained workloads. In contrast, colocation and private cloud offer greater cost predictability.
• Regulatory compliance: As data sovereignty laws tighten, enterprises need to ensure data locality and compliance without sacrificing flexibility. Private environments provide greater control over infrastructure security and governance.

This shift is not about replacing hyperscale cloud, it’s about refining its role in enterprise IT. Organisations are recognising that different workloads require different environments. The future belongs to a hybrid strategy where cloud, private infrastructure and colocation work in tandem.

The Role of Colocation in AI and High-Density Computing

Colocation is evolving beyond traditional space-and-power offerings. With the rise of AI, high-performance computing, and latency-sensitive applications, modern colocation providers are becoming strategic partners in hybrid IT deployments. Some of the key developments include:

• AI-optimised infrastructure: Enterprises are deploying high-density graphics processing unit (GPU) clusters in colocation facilities designed for liquid cooling and high-power density.
• Cloud interconnection hubs: Many colocation providers offer direct on-ramps to hyperscale clouds, enabling businesses to integrate public and private infrastructure seamlessly.
• Energy and sustainability considerations: While cost and performance are primary drivers, enterprises are also under pressure to meet sustainability targets. Colocation providers are investing in renewable energy sourcing, waste heat reuse, and water-efficient cooling to align with corporate Environmental, Social and Governance (ESG) goals.

Strategic Workload Placement

Instead of debating whether public cloud or private infrastructure is better, leading enterprises are taking a more pragmatic approach – placing workloads where they perform best. The options to be considered, include:

• High-performance AI and HPC: Dedicated infrastructure in private or collocated environments for AI model training, large-scale simulations and mission-critical analytics.
• Cloud-native applications: Public cloud for applications requiring global scalability, rapid development cycles and dynamic elasticity.
• Regulated and sensitive data: Private cloud or colocation to ensure compliance, security, and data locality.
• Hybrid cloud interplay: Seamless movement of workloads between private and public environments, ensuring both efficiency and flexibility.

Emerging Challenges and Considerations

As enterprises adopt hybrid strategies, new challenges arise. Managing a mix of cloud, colocation and private infrastructure requires advanced orchestration tools, workload automation and robust security measures. Businesses must also invest in skills and training to enable IT teams to navigate the complexities of multi-environment management effectively.

Another growing concern is the increasing pressure on data centre power grids. AI workloads are driving up energy demands, making efficiency and sustainability critical factors. Enterprises are increasingly looking for colocation providers with strong commitments to energy efficiency and innovative cooling solutions.

Looking Ahead

The past decade’s cloud-first narrative is giving way to a more practical, workload-driven approach to IT infrastructure. The future is not about choosing between public cloud, private cloud, or colocation – it’s about using all three in the right proportions.

Enterprises that embrace this hybrid approach will benefit from performance optimisation, cost control and regulatory compliance while still retaining the agility to scale where needed.

The hyperscale cloud remains an essential part of enterprise IT, but it is no longer the default answer for every workload. Instead, businesses are moving towards a strategic, workload-optimised infrastructure model that blends cloud, colocation and private environments for maximum flexibility and performance.

As AI and high-performance computing redefine what’s possible, enterprises must think beyond infrastructure decisions in isolation. They need to consider how data flows, how latency impacts decision-making, and how evolving regulations will shape the future of IT architecture. Those who build their infrastructure strategies with adaptability in mind – prioritising flexibility, security and resilience – will not only future-proof their operations but will also be positioned to lead in a rapidly evolving technological landscape.

With technology evolving at an unprecedented rate, the enterprises that will thrive are those that embrace infrastructure as a competitive advantage, not just an operational necessity. The focus is shifting from merely accessing scalable compute power to crafting an interconnected, high-performance IT ecosystem that aligns with business goals. Those that approach infrastructure decisions strategically – rather than defaulting to one model – will be best placed to navigate the complexities of AI, high-performance computing, and the new economics of cloud.

The digital infrastructure landscape is growing rapidly. This growth is being several factors. These include the exponential rise in data and the growing adoption of artificial intelligence (AI). At the same time, data centres are also facing increasing pressure to meet stringent sustainability goals. 

Cooling, which was once an operational consideration in data centre design, has now become a strategic focus. Operators are increasingly grappling with increasing heat loads, hybrid environments and the need to balance performance with efficiency. Chilled water solutions are emerging as a vital technology to help meet these challenges. Implemented correctly, they offer a flexible, efficient and future-ready approach to cooling.

Understanding the pressures on today’s facilities

As workloads evolve, so do the demands on data centre infrastructure. AI applications are now a cornerstone of many organisations’ digital strategies, requiring vast computational resources. These applications generate significantly higher heat loads than traditional IT workloads, creating an urgent need for innovative cooling strategies.

At the same time, data centres are becoming denser, as operators strive to optimise physical space by packing more computing power into smaller footprints. This densification increases heat output per square metre, placing established air cooling methods under considerable strain. When coupled with growing regulatory and market pressures to improve energy efficiency and reduce carbon footprints, it’s clear that the status quo in cooling technology is no longer sufficient.

Next-generation chip technology is advancing at such a rapid pace that the working temperature thresholds for liquid cooling are expected to keep rising. However, the range of potential outcomes is so wide that accurately forecasting future requirements has become increasingly difficult. This creates a risk for operators; as a result, determining the precise water temperature needed from the cooling system, becomes both a challenge and a potential risk for hyperscale and colocation data centre owners. Misjudging these requirements could lead to inefficient cooling strategies, increased energy consumption, and even potential damage to critical IT equipment – while also resulting in infrastructure investments that may not meet future demands. 

Why high temperature fluid cooling systems are the solution

High temperature fluid coolers are uniquely equipped to address the challenges of high-density, hybrid data centres. Unlike traditional cooling methods, which are often limited in their ability to scale with rising thermal demands, chilled water technology provides a level of flexibility and efficiency that is unmatched.

These systems are designed to work well in hybrid environments, where air cooling can be supplemented by liquid cooling solutions such as cold plates and immersion cooling. Or, conversely, where air supplements the next generation of facilities’ design primarily for liquid cooling. This versatility allows operators to optimise their approach based on specific workloads, increasing both reliability and energy efficiency.

Higher operating temperatures to reduce the need for cooling

One of the most significant changes in the cooling landscape is the shift toward higher operating temperatures. Until now, data centres have been kept cool to maintain IT equipment reliability. However, as the industry moves toward greater efficiency, this approach is being reconsidered.

Higher operating temperatures reduce the energy needed for cooling and open the door to innovative heat recovery applications. Facilities are increasingly looking to capture waste heat and repurpose it, whether for district heating or to support industrial processes. This transition requires cooling systems that can perform efficiently under these new conditions.

Chilled water systems are particularly well-suited to this challenge. Their ability to operate at elevated temperatures without sacrificing efficiency makes them a cornerstone of efficient data centre design. This aligns with emerging metrics like energy reuse effectiveness (ERE) and heat recovery efficiency (HRE), which prioritise energy recovery alongside consumption. ERE measures the total energy recovered, while HRE looks at the percentage of waste heat that is effectively captured and used by the recovery system. A higher HRE signifies better efficiency in harnessing waste heat. 

The role of hybrid cooling in high-density environments

The shift to high-density data centres presents more significant thermal management challenges than ever before. As computing power is concentrated into smaller spaces, heat generation rises significantly, requiring cooling solutions that can scale alongside these demands.

Hybrid cooling strategies – combining air and liquid cooling – are proving effective at managing these conditions. Chilled water systems form the backbone of this approach, providing the flexibility to address both baseline and high-intensity cooling needs. For example, air cooling can handle standard loads. At the same time, liquid cooling systems can manage hot spots created by AI workloads or other intensive applications.

This hybrid approach not only enhances cooling efficiency but also helps operators to optimise energy use, tailoring their solutions to the specific needs of different workloads.

Intelligent controls: a game-changer for efficiency

But cooling isn’t just about hardware. The role of intelligent control systems in optimising performance is also crucial. These systems allow all components within a cooling network – chillers, pumps, and air handling units – to work together seamlessly.

The latest and most innovative chilled water systems are equipped with advanced control platforms that monitor workloads and adjust cooling output dynamically. This capability is especially important in hybrid environments, where cooling demands can shift unpredictably. Intelligent controls enable operators to maintain efficiency, reliability and uptime, even as conditions evolve.

Looking ahead: sustainability and heat recovery

Sustainability is no longer a ‘nice to have’ for data centres; it is a business imperative. With energy demands soaring, operators must find innovative ways to reduce their environmental impact. Heat recovery is emerging as a powerful solution, enabling facilities to repurpose waste heat for secondary applications.

Chilled water systems are integral to these efforts. By capturing thermal energy during the cooling process, operators can reduce reliance on external energy sources. This not only lowers operational costs but also supports broader sustainability goals, such as reducing carbon emissions and contributing to a circular economy.

Building for the future

The demands on data centres are only going to grow. AI workloads, densification and sustainability pressures will continue to reshape the industry, requiring operators to rethink how they design and manage their facilities. Cooling systems must be able to adapt to these changes, balancing performance with energy efficiency and environmental responsibility.

A future-ready chiller should incorporate:

Ability to work at higher water temperature

Supporting varying return water and leaving temperatures from the more traditional applications working with water at 17-27°C, to more advanced ones where supply and return water temperatures can reach up to 40 – 50°C and more. As cooling requirements evolve, this ability to be flexible is essential for accommodating future technologies, including AI and high-performance computing.

Scalable Design and Adaptability

Capable of operating efficiently across a wide range of external temperatures and compact enough to manage increased densification in facilities.

Sustainability Features

Using refrigerants with very low Global Warming Potential (GWP), approaching near-zero values, to significantly reduce environmental impact and help with compliance with both current and future regulatory standards for refrigerant use. Also using waste heat recovery to support the digital economy. 

Energy Efficiency

Offering improved operational performance compared to standard chillers, reducing energy consumption through advanced technologies such as free cooling, and improving consistently low partial Power Usage Effectiveness (pPUE).

Operational Reliability

Maintaining 100% reliability even during peak operational demands, enabling robust performance and providing strategic flexibility for diverse applications.

By addressing these critical areas, data centres will be able to support the changing needs of modern facilities. As cooling requirements continue to evolve, it’s impossible to say definitively what will be needed in future. The key to success is to deploy cooling systems available today that can cope with future demands, as well as contribute to a more sustainable and energy-efficient world.

In 2025, telecoms are no longer operators of unseen, underdog infrastructure — unconsidered until someone’s Netflix buffers. Telecoms are in a remarkably good position, and they’ve got the data pipelines to prove it. This is the year where telecom innovation accelerates to an almost outlandishly futuristic level. From satellites connecting the remotest parts of the world to networks so intelligent they practically read your mind, 2025 is where telecoms don’t just show up—they dominate.

In 2025, your telco might know you better than your significant other. That emergency data boost right before a cross-country road trip? Done. Latency optimisation mid-battle for your online gaming spree? Already handled. It’s like having a genie in your pocket; only this one is powered by algorithms, not wishes.

The AI Compute Hunger: Why Data is the New Lifeblood

Artificial Intelligence thrives on data, and in 2025, it’s hungrier than ever. With the explosion of connected devices, from wearables to autonomous vehicles, telecom networks are inundated with streams of data—real-time location insights, user behaviour patterns, and device health metrics. For telcos, this is an oil mine, but only if they can extract actionable intelligence from it.

It’s no longer about collecting data but orchestrating it into meaningful actions. AI-powered Next Best Offer (NBO) and Next Best Action (NBA) as a service through API workflows analyse these streams to predict and deliver exactly what the customer needs, precisely when they need it. For example:

• A hospital’s connected devices detect a critical spike in patient data usage and prioritise bandwidth for life-saving diagnostics, ensuring doctors receive real-time results, with zero lag, during emergency procedures.
•  A financial services app integrated with AI workflows, proactively notifies users of potential fraudulent activity, locks their card, and generates a secure replacement card—all before the user realises their account is compromised.
• A logistics network’s fleet management system, powered by real-time AI orchestration, reroutes delivery trucks away from severe weather conditions, ensuring vital medical supplies reach hospitals on time without disruption. This isn’t just personalisation—it’s anticipation, powered by AI’s insatiable appetite for data in exchange for its ability to make every interaction meaningful.

The Rise of the Predictive Telecom Genie

Say goodbye to boring customer interactions and hello to a world where your network knows what you want before you do. Imagine opening a streaming app, and instead of a buffering circle, you’re greeted by a hyper-personalised experience so seamless it feels like magic. This isn’t just wishful thinking; it’s powered by telecom’s newfound love affair with AI-driven predictive experiences like Next Best Offer (NBO) and Next Best Action (NBA).

In 2025, your telco isn’t just a network—it’s your digital genie, granting wishes before you even rub the lamp. Need a data boost as you zip across the country? Done. Gaming mid-battle and need lag-free magic? Sorted. Stuck in a subway and craving a seamless podcast? Stream on. Whether live-streaming a concert, hiking off the grid, or saving your online presentation from the perils of buffering, your telco has your back. No more crossed wires—this is predictive perfection, powered by algorithms that know your needs better than your best friend.

Satellites: From Niche to Mainstream Marvel

2025 is the year when telecoms finally look up—literally. Satellite technology is no longer the nerdy cousin no one talks about at family gatherings. Thanks to massive investments, satellite telecom is the cool kid on the block, beaming high-speed internet to the most remote corners of the planet.

You thought your 5G was fast? Wait until satellites deliver direct-to-device communication, which feels like it’s straight out of a James Bond movie. And if you’re thinking, “What’s the big deal about satellites?” Remember this: by the end of the year, they’ll be the reason someone in the Amazon rainforest can video chat with their grandma in real-time.

Networks That Think Faster Than You Can Blink

Remember when your network only cared about staying online? In 2025, networks have gotten smarter—like, scary-smart. These aren’t just networks anymore; they’re autonomous decision-makers. Imagine an AI-powered system detecting a potential network outage before it happens and fixing it faster than you can say, “I need to call customer support.”

This isn’t about faster internet speeds—it’s about networks with a sixth sense. They’ll anticipate failures, optimise traffic in real-time, and make sure your 4K video stream doesn’t so much as hiccup. It’s like having a network that graduated top of its class in predictive genius.

5G Gets a Real Job

Let’s be honest: the 5G hype train has been going full steam for years, but 2025 is when 5G finally stops talking big and starts delivering. This is the year it becomes the backbone of the industry, transforming everything from gaming and AR/VR experiences to industrial IoT and edge computing.

Gaming tournaments with no lag? Check. Smart cities that adjust traffic lights on the fly? Double check. 5G isn’t just a buzzword anymore; it’s the economic engine that will fuel everything from tech startups to Fortune 500 giants.

The Green Gold Rush: Recycling Is Cool Again

Who knew old copper wiring could be worth billions? In 2025, telecoms are diving headfirst into what we’re calling the Green Gold Rush. Operators decommissioning their legacy copper networks aren’t just saving money—they’re cashing in on a resource so valuable it could make Elon Musk jealous.

But this isn’t just about profits. By recycling copper and investing in energy-efficient networks, telecoms are setting new sustainability standards. Think fewer emissions, more green technology, and an industry that’s finally as eco-friendly as it is innovative.

Collaboration Over Competition: Federated Networks Take Center Stage

In 2025, telecom operators will finally figure out that sharing is caring. Federated networks—where operators team up to provide seamless, shared connectivity—are no longer just a concept; they’re the future. This means better service for customers, lower costs for operators, and a whole lot fewer headaches for everyone involved.

Imagine a world where switching between networks is so smooth you barely notice. It’s like having multiple Wi-Fi routers in your house, but on a global scale. And the best part? It’s all about giving customers what they want—reliable, uninterrupted connectivity wherever they are.

Cybersecurity Becomes Sexy

Okay, maybe it’s not sexy, but it’s a top priority. With cyber threats growing more sophisticated by the day, telecoms in 2025 aren’t messing around. AI-driven threat detection, zero-trust architectures, and ironclad data protection are the new norm.

Why the sudden obsession? Because no one wants to be the operator that lost customer data or got hit by ransomware. In this hyper-connected world, cybersecurity isn’t just important—it’s survival.

Asia Takes the Lead

Move over, Silicon Valley—Asia is where the telecom action is in 2025. With skyrocketing demand for AI-powered data centers, 5G rollouts, and high-capacity subsea cables, the region is set to become the global epicenter of telecom innovation.

India and Southeast Asia are growing so fast that it’s hard to keep up. Telcos investing here aren’t just riding the wave—they’re shaping the future. 

2025: Telecom’s Blockbuster Year

Here’s the bottom line: 2025 isn’t just another year—it’s a turning point. Telecoms are no longer playing catch-up; they’re leading the charge into a future filled with AI, 5G, satellites, and more.

And if you think this all sounds too good to be true, just wait. The telecom revolution isn’t coming—it’s already here. So, grab some popcorn, sit back, and enjoy the ride. Because in 2025, telecoms aren’t just connecting the world—they’re transforming it.

With businesses racing to harness the potential of generative artificial intelligence (AI), the energy requirements of the technology have come into sharp focus for organisations around the world. 

Training and building generative AI models requires not only a huge amount of power, but also dense computational resources packed into a small space, generating heat. 

The Graphics Processing Units (GPUs) used to deliver such technology are highly energy intensive, and as generative AI becomes more ubiquitous, data centres will need more power, and generate ever more heat. For businesses hoping to reap the rewards of generative AI, the need for new solutions to cool data centres is becoming urgent. 

Air cooling is no longer enough

Energy intensive Graphics Processing Units (GPUs) that power AI platforms require five to 10 times more energy than Central Processing Units (CPUs), because of the larger number of transistors. This is already impacting data centers. 

There are also new, cost-effective design methodologies incorporating features such as 3D silicon stacking, which allows GPU manufacturers to pack more components into a smaller footprint. This again increases the power density, meaning data centers need more energy, and create more heat. 

Another trend running in parallel is a steady fall in TCase (or Case Temperature) in the latest chips. TCase is the maximum safe temperature for the surface of chips such as GPUs. It is a limit set by the manufacturer to ensure the chip will run smoothly and not overheat, or require throttling which impacts performance. On newer chips, T Case is coming down from 90 to 100 degrees Celsius to 70 or 80 degrees, or even lower. This is further driving the demand for new ways to cool GPUs. 

As a result of these factors, air cooling is no longer doing the job when it comes to AI. It is not just the power of the components, but the density of those components in the data center. Unless servers become three times bigger than they were before, data centres need a way to remove heat more efficiently. That requires special handling, and liquid cooling will be essential to support the mainstream roll-out of AI. 

The dawn of liquid 

Liquid cooling is growing in popularity. Public research institutions were amongst the first users, because they usually request the latest and greatest in data center tech to drive high performance computing (HPC) and AI. Yet they tend to have fewer fears around the risk of adopting new technology. 

Enterprise customers are more risk averse. They need to make sure what they deploy will immediately provide return on investment. We are now seeing more and more financial institutions – often conservative due to regulatory requirements – adopt the technology, alongside the automotive industry. 

The latter are big users of HPC systems to develop new cars, and now also the service providers in colocation data centers. Generative AI has huge power requirements that most enterprises cannot fulfil within their premises, so they need to go to a colocation data center, to service providers that can deliver those computational resources. Those service providers are now transitioning to new GPU architectures, and to liquid cooling. If they deploy liquid cooling, they can be much more efficient in their operations. 

Cooling the perimeter

Liquid cooling delivers results both within individual servers and in the larger data centers. By transitioning from a server with fans to a server with liquid cooling, businesses can make significant reductions when it comes to energy consumption. 

But this is only at device level, whereas perimeter cooling – removing heat from the data center – requires more energy to cool and remove the heat. That can mean data centres can only use two thirds of the energy it consumes on towards computing: the task it was designed to do. The rest is used to keep the data center cool.

Power usage effectiveness (PUE) is a measurement of how efficient data centers are. You take the power required to run the whole data center, including the cooling systems, divided by the power requirements of the IT equipment. With data centers that are optimised by liquid, some of them are doing PUE of 1.1, and some even 1.04, which means a very small amount of marginal energy. That’s before we even consider the opportunity to take this hot liquid or water coming out of the racks, and reuse that heat to do something useful, such as heating the building in the winter, which we see some customers doing today. 

Density is also very important. Liquid cooling allows us to pack a lot of equipment in a high rack density. With liquid cooling, we can populate those racks and use less data center space overall, less real estate, which is going to be very important for AI.

An essential tool

With generative AI’s energy demands set to grow, liquid cooled systems will become an essential tool to deliver energy efficient AI today, and also to scale towards future advancements. Air cooling is simply no longer up to the job in the era of energy-hungry generative AI. 

The emergence of generative AI has put the power demands of data centres under the spotlight in an unprecedented way. For business leaders, this is an opportunity to act proactively, and embrace new technology to meet this challenge. 

The rapid acceleration of artificial intelligence (AI), driven by GenAI, is redefining the role of data centres. As AI begins to change industries from healthcare to finance, the expectation is that the demand on data centres to support intensive machine learning processes will be unprecedented. According to analyst Gartner, spending on data centre systems is expected to increase 24% in 2024 due in large part to increased planning for GenAI.

The International Energy Agency (IEA) says that data centres are already responsible for around 1% of global electricity use, and it is expected that energy demands will grow exponentially as AI adoption increases. This highlights the increasing need for energy-efficient solutions and has prompted regulatory bodies like the European Commission to set stringent energy-efficiency targets such as the 2023 ‘Digital Decade’ policy, which aims to reduce the carbon footprint of the ICT sector by 40% by 2030.

From Stability to Agility: The New Data Centre Paradigm

Traditionally, data centres were designed for stability, focusing on consistent uptime and reliable performance for relatively predictable workloads. This model works well for traditional IT workloads but may fall short for AI, where workloads are highly variable and resource-intensive. 

Training large machine learning models (LLMs), obviously requires immense computational power and energy, while inference tasks can fluctuate based on real-time data demands. With the requirements of the digital space set to escalate, it’s crucial for data centre operators to adapt continuously, leveraging innovative solutions and operational efficiencies to meet the future head-on. 

Enhancing Energy Efficiency: A Critical Imperative

The rising energy consumption associated with AI workloads is an operational challenge as well as an environmental one. 

Data centres are already significant consumers of electricity, and the projected doubling of energy use by 2026 will place even greater strain on both operators and the grid. This makes energy efficiency and availability a top priority for operators.

Battery energy storage systems (BESS) can help to improve energy efficiency. They can store excess electricity and make it available when needed. This is critical in countries like Denmark, where the EU’s ‘Energy Efficiency Directive’ mandates operators integrate at least 10% renewable energy into their power mix by 2025. 

BESS have the potential to give data centres more control over their connection to the grid providing more autonomy. 

BESS can also be used to alleviate grid infrastructure constraints and offer equipment owners the potential to provide grid services and generate new revenue streams, as well as cost savings on electricity use. These systems can provide grid-balancing services. They enable energy independence and bolster sustainability efforts at mission critical facilities, providing flexibility in the use of utility power and are a critical step in the deployment of a dynamic power architecture. BESS solutions allow organisations to fully leverage the capabilities of hybrid power systems that include solar, wind, hydrogen fuel cells, and other forms of alternative energy.

According to Omdia’s Market Landscape: Battery Energy Storage Systems report, “Enabling the BESS to interact with the smart electric grid is an innovative way of contributing to the grid through the balance of energy supply and demand, the integration of renewable energy resources into the power equation, the reduction or deferral of grid infrastructure investment, and the creation of new revenue streams for stakeholders.”

Preparing for the AI Future: Strategic Investments in Infrastructure

As AI continues to change industries, the infrastructure that supports it needs to evolve too. This requires strategic investments not only in physical hardware but also in management systems that can optimise performance and energy use. 

AI-driven automation within data centres can play a pivotal role, enabling predictive maintenance, dynamic resource allocation, and even automated responses to security threats. For example, it is the continuous exchange of data with the critical equipment and the adoption of a monitoring system that allows the identification of potential threats and anomalies that could impact business or service continuity. The identification of patterns and anomalies in the collection of large amounts of data permits a faster and more accurate problem discovery, diagnosis and resolution. This monitoring of critical equipment adds an important layer of protection to continuity, and therefore availability of the infrastructure. 

Investment in innovative cooling solutions is also becoming essential as traditional air-cooling systems struggle to keep up with the heat generated by high-density computing environments. Although air-cooling solutions will be part of the data centre infrastructure for some time to come, liquid cooling and direct-to-chip cooling technologies offer promising additions, allowing data centres to maintain optimal temperatures without compromising performance. According to industry analyst Dell’Oro Group the market for liquid cooling could grow to more than $15bn over the next five years.

Investing in the Edge 

Edge computing is another area of infrastructure that is likely to need further investment in the AI era. Edge data centres can significantly reduce latency and bandwidth usage by processing data closer to its source, which is crucial for applications like autonomous vehicles and smart cities. This distributed approach to data management allows for more efficient processing of AI workloads, reducing the burden on centralised data centres. IDC predicts that worldwide spending on edge computing will reach $378 Billion in 2028, driven by demand on real-time analytics, automation, and enhanced customer experiences.

Collaboration Across the Ecosystem: The Path to Innovation

The future of AI-driven data centres will depond on collaboration across the technology ecosystem. Operators, IT hardware manufacturers, chip designers, software developers and AI researchers must work together to develop solutions that meet the unique demands of AI. This collaborative approach is essential for driving innovation and enabling data centres to support the next generation of AI applications. 

For instance, the integration of AI-specific processors and accelerators requires close coordination between IT hardware manufacturers and data centre operators. Similarly, the development of specialised software environments that efficiently manage data and resources will depend on ongoing collaboration between data centre operators and software developers.

Embracing the Future: A New Role for Data Centres

With increasing AI demands, power consumption challenges, and sustainability goals, the data centre industry is at a critical juncture. Implementing practical solutions like liquid cooling and battery energy storage systems (BESS) is key to addressing these issues. By investing in agile, energy-efficient infrastructures and fostering collaboration across the ecosystem, data centres can position themselves at the heart of this transformation. In doing so, they will not only support today’s AI applications but also pave the way for future innovations, helping to shape the digital landscape of tomorrow.

In the fast-paced world of digital infrastructure, data centres are expected to deliver more than just storage and processing power. As demand continues to grow, the ability to make real-time, data-driven decisions has become a cornerstone of efficient data centre operations. Artificial Intelligence (AI) is at the forefront of this transformation, automating decision-making processes and optimising operations across the board.

AI: The Brain Behind Data Centre Automation

AI is no longer just a tool for efficiency – it’s becoming the decision-making brain of modern data centres. Traditionally, data centre operations required human intervention at nearly every stage, from monitoring systems to adjusting resource allocation. While effective, this model is labour-intensive and can be prone to errors, especially as operations scale.

AI changes this dynamic by automating many of these decisions. AI can continuously monitor environmental conditions, workloads and resource consumption. By doing so, these systems can make real-time adjustments to ensure that data centres operate at peak efficiency. They can redistribute server workloads, adjust cooling systems or balance power usage. Essentially, AI is taking on the role of an intelligent, always-on operator.

Automating Workflows with AI

AI-driven automation is streamlining workflows within data centres, reducing the need for human intervention in routine tasks. For example, AI systems can automate the backup and recovery processes, ensuring that data is continuously protected without the need for constant manual oversight. 

Similarly, routine maintenance checks and system updates can be scheduled and performed automatically, allowing skilled personnel to focus on more strategic initiatives.

By automating these repetitive tasks, AI enhances productivity and reduces the risk of human error. This level of automation enables data centres to scale without a proportional increase in staffing, making operations more cost-effective and efficient.

AI’s ability to learn from previous operations means that it continuously refines its decision-making processes. The longer AI is integrated into a data centre’s operations, the more accurate and efficient it becomes, leading to further optimisation.

AI-Powered Decision-Making in Cooling and Energy Use

One of the most important areas where AI is making an impact is in cooling and energy management. Cooling systems are responsible for up to 40% of a data centre’s energy consumption, and inefficiencies in these systems can lead to substantial cost increases as operations scale. AI’s predictive analytics and real-time monitoring capabilities allow it to optimise cooling systems dynamically.

By analysing environmental conditions and server workloads, AI can adjust cooling settings to match the precise needs of the facility. For instance, during off-peak hours, AI can scale back cooling efforts, reducing energy consumption without affecting performance. This level of decision-making ensures that energy use is always optimised, reducing costs and supporting sustainability goals.

In addition to cooling systems, AI can optimise energy distribution across the entire facility. By monitoring power usage in real-time, AI can balance loads between different systems, ensuring that no single server or component is overburdened. This not only improves performance but also extends the life of critical infrastructure by preventing excessive wear and tear.

AI and Predictive Analytics: Proactive Decision-Making

Predictive analytics, powered by AI, is also transforming how data centres make proactive decisions. By analysing historical data and real-time performance metrics, AI systems can predict when issues are likely to occur. Not only that, but they can then take pre-emptive actions to prevent these issues. For example, if AI detects that a particular server is underperforming, it can redistribute workloads to avoid potential bottlenecks or failures.

This proactive approach to decision-making helps data centres to avoid costly downtime and maintain consistent service levels. As operations scale, AI’s ability to predict and resolve issues before they escalate will become increasingly critical to maintaining performance and reliability.

Predictive analytics also plays a role in optimising resource allocation. AI systems can analyse usage patterns to determine when certain resources are underutilised and adjust them accordingly. This dynamic allocation enables data centres to operate at maximum efficiency, reducing waste and improving overall performance.

AI in Security: Real-Time Decision-Making for Threat Mitigation

Security remains a top concern for data centres, particularly as they scale and become more complex. AI’s ability to make real-time security decisions is a game-changer in this space. By continuously monitoring network traffic and access patterns, AI systems can detect and respond to threats as they arise, without the need for human intervention. 

For example, if AI detects an unauthorised access attempt or abnormal data transfer, it can automatically trigger security protocols, such as isolating the affected area or notifying administrators. This real-time decision-making capability helps data centres to remain secure, even as they expand to meet growing demands.

In addition to reacting to potential threats, AI systems learn from each incident they encounter, continuously improving their ability to detect and respond to emerging attack vectors. This adaptive learning process allows AI to stay ahead of evolving cyber threats, making it an essential part of any data centre’s security strategy. Moreover, AI can be integrated into both physical security systems – such as managing access controls to sensitive areas – and cybersecurity measures, providing comprehensive protection for the facility.

AI’s Role in Scaling and Future-Proofing Data Centres

AI’s role in decision-making extends beyond immediate operational efficiency. It’s also key to future-proofing data centres as they scale to meet increasing demands. AI helps data centres manage their growing infrastructure by enabling seamless scalability without a proportional increase in complexity or cost.

As data centres expand to include more servers, storage systems and networks, traditional management approaches can struggle to keep up. AI systems, however, can handle the increased complexity. AI can meet these challenges by automating resource allocation, predictive maintenance and security measures. In doing so, the technology allows data centres to grow while maintaining the same level of operational efficiency and reliability. This makes AI an indispensable tool for future-proofing facilities. It could, if deployed correctly, ensure that they remain agile and adaptable in the face of evolving digital demands.

The future of digital infrastructure lies in the seamless integration of AI into all aspects of data centre management. The technology has a role to play from resource allocation to security and disaster recovery. As AI technology continues to mature, it will drive greater efficiency, resilience and scalability in data centres, positioning them to meet the demands of the next generation of digital services.

The UK is one of the latest countries to benefit from the boom in Artificial Intelligence – after it sparked major investments in Cloud computing. Amazon Web Services’ recently announced it is spending £8bn on UK data centres. It is largely spending this money to support its AI ambitions. The announcement followed another that said Amazon would spend another £2b on AI related projects. Given the scale of these investments, it’s not surprising many people immediately think Cloud computing when we talk about the future of AI. But in many cases, AI isn’t happening in the Cloud – it’s increasingly taking place at the Edge.

Why the edge?

There are plenty of reasons for this shift to the Edge. While such solutions will likely never be able to compete with the Cloud in terms of sheer processing power, AI on the Edge can be made largely independent from connectivity. From a speed and security perspective that’s hard to beat.  

Added to this is the emergence of a new class of System-on-Chip (SoC) processors, produced for AI inference. Many of the vendors in this space are designing chipsets that tech companies can deploy for specific use cases. Examples of this can be found in the work Intel is doing to support computer vision deployments, the way Qualcomm is helping to improve the capabilities of mobile and wearable devices and how Ambarella is advancing what’s possible with video and image processing. Meanwhile, Nvidia is producing versatile solutions for applications in autonomous vehicles, healthcare, industry and more.

When evealuating Cloud vs Edge, it’s important to also consider the the cost factor. If your user base is likely to grow substantially, operational expenditure is likely to increase significantly as Cloud traffic grows. This is particularly true if the AI solution also needs large amounts of data, such as video imagery, constantly. In these cases, a Cloud-based approach may not be financially viable. 

Where Edge is best

That’s why the global Edge AI market is growing. One market research company recently estimated that it would grow to $61.63bn in 2028, from $24.48bn in 2024. Particular areas of growth include sectors in which cyber-attacks are a major threat, such as energy, utilities and pharmaceuticals. The ability of Edge computing to create an “air gap” through which cyber-criminals are unable to penetrate makes it ideal for these sectors. 

In industries where speed and reliability are of the essence, such as in hospitals, on industrial sites and with transport, Edge also offers an unparalleled advantage. For example, if an autonomous vehicle detects an imminent collision, the technology needs to intervene immediately. Relying on a cellular connection is not an acceptable idea in this scenario. The same would apply if there was a problem with machinery in an operating theatre.

Edge is also proving transformational in advanced manufacturing, where automation is growing exponentially. From robotics to business analytics, the advantages of fast, secure, data-driven decision-making is making Edge an obvious choice. 

Stepping carefully to the Edge

So how does an AI project make its way to the Edge? The answer is that it requires a considered series of steps – not a giant leap. 

Perhaps counter-intuitively, it’s likely that an Edge AI project will begin life in the Cloud. This is because the initial development often requires a scaled level of processing power that can only be found in a Cloud environment. Once the development and training of the AI model is complete, however, the fully mature version transition and deploy to Edge infrastructure. 

Given the computing power and energy limitations on a typical edge device, however, one will likely need to consider all the ways it can keep the data volume and processing to a minimum. This will require the application of various optimisation techniques to minimise the size of these data inputs – based on a review of the specific use case and the capabilities of the selected SoC, along with all Edge device components such as cameras and sensors that may be supplying the data. 

It is likely that a fair degree of experimentation and adjustments will be needed to find the lowest acceptable level of decision-making accuracy that is possible, without compromising quality too much. 

Optimising AI models to function beyond the core of the network

To achieve a manageable AI inference at the Edge, teams will also need to iteratively optimise the AI model itself. Achieving this will almost certainly involve several transformations, as the model goes through quantisation and simplification processes. 

It will also be necessary to address openness and extensibility factors – to be sure that the system will be interoperable with third party products. This will likely involve the development of a dedicated API to support the integration of internal and external plugins and the creation of a software development kit to ensure hassle-free deployments. 

AI solutions are progressing at unprecedented rate, with AI companies releasing refined, more capable models all the time, Therefore, there needs to be a reliable method for quickly updating the ML models at the core of an Edge solution. This is where MLOps kicks in, alongside DevOps methodology, to provide the complete development pipeline. Organisations can turn to the tools and techniques developed for and used in traditional DevOps, such as containerisation, to help owners keep their competitive advantage.

While Cloud computing, and its high-powered data processing capabilities, will remain at the heart of much of our technological development in the coming decades, expect to also see large growth in Edge computing too. Edge technology is advancing at pace, and anyone developing an AI offering, will need to consider the potential benefits of an Edge deployment before determining how best to invest. 

Often the most data rich modern organisations can be information poor. This gap emerges where businesses struggle to fully leverage data, especially where exponential data growth creates new challenges. A data ‘rich’ company requires robust, secure and efficient storage solutions to harness data to its fullest potential. From advanced on-premises data centres to cloud storage, the evolution of data storage technologies is fundamental to managing the vast amounts of information that organisations depend on every day.

Storage for today’s landscape 

In today’s climate of rigorous compliance and escalating cyber threats, operational resilience depends on strategies that combine data storage, effective backup and recovery, as well as cyber security. Storage solutions provide the foundation for managing vast amounts of data, but simply storing this data is not enough. Effective backup policies are essential to ensure IT teams can quickly restore data in the event of deliberate or accidental disruptions. Regular backups, combined with redundancy measures, help to maintain data integrity and availability, minimising downtime and ensuring business continuity.

Cyber threats – such as hacking, malware, and ransomware – is an advancing front, posing new risks to businesses of all sizes. Whilst SMEs often find themselves targets, threat actors prioritise organisations most likely to suffer from downtime, where, for example, resources are limited, or there are cyber skills gaps. It has even been estimated that an alarmingly high as 60% of SMEs wind down their shutters just six months after a breach. 

If operational resilience is on your business’ agenda, then rapid recoveries (from verified points of retore) can return a business to a viable state. The misconception, where attacks nowadays feel all too frequent, is that business recovery is a long, winding road. Yet, market-leading data storage options have evolved, like IBM FlashSystem, to address conversations around operational resilience in new, meaningful ways.  

Storage Options

An ideal storage strategy should capture a means of managing data that organises storage resources into different tiers based on performance, cost, and access frequency. This approach ensures that data is stored in the most appropriate and cost-effective manner.

Storage fits within various categories, including hot storage, warm storage, cold storage, and archival storage – each with various benefits that organisations can leverage, be it performative gains, or long-term data compliance and retention. But organisations large and small must start to position storage as a strategic pillar in their journey to operational resilience – a critical part of modern parlance for businesses, enshrined by the likes of the Financial Conduct Authority (FCA). 

By adopting a hierarchical storage strategy, organisations can optimise their storage infrastructure, balancing performance and cost. This approach enhances operational resilience by ensuring critical data is always accessible. Not only that, but it also helps to effectively manage investment in storage. 

Achieving operational resilience with storage 

1. Protection – a protective layer in storage means verifying and validating restore points to align with Recovery Point Objectives. After IT teams restore operations, ‘clean’ backups ensure that malicious code doesn’t end up back in the your systems.   
2. Detection – does your storage solution help mitigate costly intrusions by detecting anomalies and thwarting malicious, early-hour threats? FlashSystem, for example, has inbuilt anomaly detection to prevent invasive threats breaching your IT environment. Think early, preventative strategies and what your storage can do for you. 
3. Recovery – the final stage is all about minimising losses after impact, or downtime. This step addresses operational recovery, getting a minimum viable company back online. This works to the lowest possible Recovery Time Objectives. 

Storage can be a matter of business survival. Cyber resilience, quick recovery and a robust storage strategy help circumvent the following:

• Reduce inbound risks of cyber attacks. 
• Blunt the impact of breaches.
• Ensure a business can remain operational. 

It’s helpful to imagine whether or not your business can afford seven or more days of downtime after an attack. 

Advanced data security 

Anomaly detection technology in modern storage systems offers significant benefits by proactively identifying and addressing irregularities in data patterns. This capability enhances system reliability and performance by detecting potential issues before they escalate into critical problems. By continuously monitoring data flows and usage patterns, the technology ensures optimal operation and reduces downtime. 

But did you know market-leaders in storage, like IBM, have inbuilt, predictive analytics to ensure that even the most data rich companies remain informationally wealthy? This means system advisories with deep performance analysis can drive out anomalies, alterting businesses about the state of their IT systems and the integrity of their data – from the point where it is being stored.   

Selecting the appropriate storage solution ultimately enables you to develop a secure, efficient, and cost-effective data management strategy. Doing so boosts both your organisation’s and your customers’ operational resilience. Given the inevitability of data breaches, investing in the right storage solutions is essential for protecting your organisation’s future. Storage conversations should add value to operational resilience, where market-leaders in this space are changing the game to favour your defence against cyber threats and risks of all varieties.

In the landscape of AI development, computing capabilities are expanding from the cloud and data centres into devices, including vehicles. For smart devices to improve and learn, they require access to data, which must be stored and processed effectively. Embedded AI computing can facilitate this by integrating AI into an electronic device or system – such as mobile devices, autonomous vehicles, industrial automation systems and robotics. 

However, for this to happen, the need for ample storage capacity within the device itself is increasingly important. This is especially so when it comes to smart vehicles and traffic management, as these technologies are also tapping into the benefits of embedded AI computing. 

Smarter vehicles: Better experiences

By storing and processing data locally, smart vehicles can continuously refine their algorithms and functionality without relying solely on cloud-based services. This local approach not only enhances the vehicle’s autonomy but also ensures that crucial data is readily accessible for learning and improvement.

Moreover, as data is recorded, replicated and reworked to facilitate learning, the demand for storage capacity escalates. In this case, latency is key for smart vehicles as they need access to data fast – especially for security features on the road. This requires the integration of advanced CPUs, often referred to as the “brains” of the device, to enable efficient processing and analysis of data.

In addition, while local storage and processing enhance device intelligence, data retention is essential to sustain learning over time. Therefore, there must be a balance between local processing and cloud storage. This ensures that devices can leverage historical data effectively without compromising real-time performance.

In the context of vehicles, this approach translates into onboard systems that will be able to learn from past experiences, adapt to changing environments, and communicate with other vehicles and infrastructure elements – like traffic lights. Safety is, of course, of huge importance for smart vehicles. Automobiles equipped with sensors and embedded AI will be able to flag risks in real time, such as congestion or even obstacles in the road, improving the safety of the vehicle. In some vehicles, these systems will even be able to proactively steer the vehicle away from an obstacle or bring the vehicle to a safe stop.

Ultimately, this integration of AI-driven technology will allow vehicles to become smarter, safer, and more responsive, revolutionising the future of transportation. To facilitate these advanced capabilities, quick access to robust data storage is key.

Smart cities and traffic management

Smart cities run as an Internet of Things (IoT), allowing various elements to interact with one another. In these urban environments, connected infrastructure elements such as smart cars will form part of a wider system to allow the city to run more efficiently. This is underpinned by data and data storage. 

The integration of AI-driven technology into vehicles has significant implications for smart traffic management. With onboard systems capable of learning from past experiences and adapting to dynamic environments, vehicles can contribute to more efficient and safer traffic flows.

Additionally, vehicles will be able to communicate with each other and with infrastructure elements, such as traffic lights, to enable coordinated decision-making. This communication network facilitated by AI-driven technology will allow for real-time adjustments to traffic patterns, optimising traffic flow, reducing congestion and minimising the likelihood of accidents.

For any central government department of transport and local government bodies, insights from connected vehicles can better prepare a built environment to handle peaks in traffic. When traffic levels are likely to be high, management teams can limit roadworks and other disruptions on roads. In the longer term, understanding the busiest roads can also inform the construction of bus lanes, cycle paths and infrastructure upgrades in the areas where these are most needed. 

Storage plays a foundational role in enabling vehicles to leverage AI-driven technology for smart traffic management. It supports data retention, learning, communication, and system reliability, contributing to the efficient and safe operation of smart transportation networks.

Final thoughts

Ultimately, the integration of AI into vehicles lays the foundation for a comprehensive smart traffic management system. By leveraging data-driven insights and facilitating seamless communication between vehicles and infrastructure, this approach promises to revolutionise transportation, making it safer, more efficient, and ultimately more sustainable – all made possible with appropriate storage solutions and tools.

In the dynamic landscape of modern technology, artificial intelligence (AI) has emerged as a transformative force. The technology is revolutionising industries and creating an unprecedented demand for high performance computing solutions. As a result, AI applications are becoming increasingly sophisticated and pervasive across sectors such as finance, healthcare, manufacturing, and more. In response, data centre providers are encountering unique challenges in adapting their infrastructure to support these demanding workloads.

AI workloads are characterised by intensive computational processes that generate substantial heat. This can pose significant cooling challenges for data centres. Efficient and effective cooling solutions are essential to facilitate optimal performance, reliability and longevity of IT systems. 

The importance of cooling for AI workloads

Traditional air-cooled systems, commonly employed in data centres, may struggle to effectively dissipate the heat density associated with AI workloads. As AI applications continue to evolve and push the boundaries of computational capabilities, innovative liquid cooling technologies are becoming indispensable. Liquid cooling methods, such as immersion cooling and direct-to-chip cooling, offer efficient heat dissipation directly from critical components. Thishelps mitigate the risk of performance degradation and hardware failures associated with overheating.

Deploying robust cooling infrastructure tailored to the unique demands of AI workloads is imperative for data centre providers seeking to deliver high-performance computing services efficiently, reliably and sustainably.

Advanced cooling technologies for AI

Flexibility is key when it comes to cooling. There is no “one size fits all” solution to this challenge. Data centre providers should be designing facilities to accommodate multiple types of cooling technologies within the same environment. 

Liquid cooling has emerged as the preeminent solution for addressing the thermal management challenges posed by AI workloads. However, it’s important to understand that air cooling systems will still be part of data centre’s for the foreseeable future. 

Immersion Cooling

Immersion cooling involves submerging specially designed IT hardware (servers and graphics processing units, GPUs) in a dielectric fluid. These fluids tend to comrpise mineral oil or synthetic coolant. The fluid absorbs heat directly from the components, providing efficient and direct cooling without the need for traditional air-cooled systems. This method significantly enhances energy efficiency. As a result, it also reduces running costs, making it ideal for AI workloads that produce substantial heat.

Immersion cooling facilitates higher density configurations within data centres, optimising space utilisation and energy consumption. By immersing hardware in coolant, data centres can effectively manage the thermal challenges posed by AI applications.

Direct-to-Chip Cooling

Direct-to-chip cooling, also known as microfluidic cooling, delivers coolant directly to the heat-generating components of servers, such as central processing units (CPUs) and GPUs. This targeted approach maximises thermal conductivity, efficiently dissipating heat at the source and improving overall performance and reliability.

By directly cooling critical components, the direct-to-chip method helps to ensure that AI applications operate optimally, minimising the risk of thermal throttling and hardware failures. This technology is essential for data centres managing high-density AI workloads.

Benefits of a mix-and-match approach

The versatility and flexibility of liquid cooling technologies provides data centre operators with the option of adopting a mix-and-match approach tailored to their specific infrastructure and AI workload requirements. Integrating multiple cooling solutions enables providers to:

• Optimise Cooling Efficiency: Each cooling technology has unique strengths and limitations. Different types of liquid cooling can be deployed in the same data centre, or even the same hall. By combining immersion cooling, direct-to-chip cooling and / or air cooling, providers can leverage the benefits of each method to achieve optimal cooling efficiency across different components and workload types.
• Address Varied Cooling Needs: AI workloads often consist of diverse hardware configurations with varying heat dissipation characteristics. A mix-and-match approach allows providers to customise cooling solutions based on specific workload demands, ensuring comprehensive heat management and system stability. 
• Enhance Scalability and Adaptability: As AI workloads evolve and data centre requirements change, a flexible cooling infrastructure that supports scalability and adaptability becomes essential. Integrating multiple cooling technologies provides scalability options and facilitates future upgrades without compromising cooling performance. For example, air cooling can support HPC and AI workloads to a degree, and most AI deployments will continue to require supplementary air cooled systems for networking infrastructure. All cooling types ultimately require waste heat to be removed or re-used, so it is important that the main heat rejection system (such as chillers) is sized appropriately and enabled for heat reuse where possible.  

A cooler future

Effective cooling solutions are paramount if data centres are to meet the ever-growing demands of AI workloads. Liquid cooling technologies play a pivotal role in enhancing performance, increasing energy efficiency and improving the reliability of AI-centric operations.

The adoption of advanced liquid cooling technologies not only optimises heat management and reuse but also contributes to reducing environmental impact by enhancing energy efficiency and enabling the integration of renewable energy sources into data centre operations.

The worldwide scramble to adopt and monetise generative artificial intelligence (AI) is accelerating an already bullish semiconductor market, according to new data gathered by Gartner. 

According to the company’s latest report, the global AI semiconductor revenues will likely grow by 33% in 2024. By the end of the year, the market is expected to total $71 billion. 

“Today, generative AI (GenAI) is fueling demand for high-performance AI chips in data centers. In 2024, the value of AI accelerators used in servers, which offload data processing from microprocessors, will total $21 billion, and increase to $33 billion by 2028,” said Alan Priestley, VP Analyst at Gartner.

Breaking down the spending across market segments, 2024 will see AI chips revenue from computer electronics total $33.4 billion. This will account for just under half (47%) of all AI semiconductors revenue. AI chips revenue from automotive electronics will probably reach $7.1 billion, and $1.8 billion from consumer electronics in 2024.

AI chips’ biggest year yet 

Semiconductor revenues for AI deployments will continue to experience double-digit growth through the forecast period. However, 2024 is predicted to be the fastest year in terms of expansion in revenue. Revenues will likely rise again in 2025 (to just under $92 billion), representing a slower rate of growth. 

Incidentally, Garnter’s analysts also note coprorations currently dominating the AI semiconductor market can expect more competition in the near future. Increasingly, chipmakers like NVIDIA could face a more challenging market as major tech companies look to build their own chips. 

Until now, focus has primarily been on high-performance graphics processing units (GPUs) for new AI workloads. However, major hyperscalers (including AWS, Google, Meta and Microsoft) are reportedly all working to develop their own chips optimised for AI. While this is an expensive process, hyperscalers clearly see long term cost savings as worth the effort. Using custom designed chips has the potential to dramatically improve operational efficiencies, reduce the costs of delivering AI-based services to users, and lower costs for users to access new AI-based applications. 

“As the market shifts from development to deployment we expect to see this trend continue,” said Priestley.

Public cloud spending by end-users is on the rise. According to Gartner, the amount spent worldwide by end users on public cloud services will exceed $675 billion in 2024. This represents a sizable increase of 20.4% over 2023, when global spending totalled $561 billion. 

Gartner analysts identified the trend late in 2023, predicting strong growth in public cloud spending. Sid Nag, Vice President Analyst at Gartner said in a release that he expects “public cloud end-user spending to eclipse the one trillion dollar mark before the end of this decade.” He attributes the growth to the mass adoption of generative artificial intelligence (AI). 

Generative AI driving public cloud spend

According to Gartner, widespread enthusiasm among companies in multiple industries for generative AI is behind the distinct up-tick in public cloud spending. “The continued growth we expect to see in public cloud spending can be largely attributed to GenAI due to the continued creation of general-purpose foundation models and the ramp up to delivering GenAI-enabled applications at scale,” he added. 

Digital transformation and “application modernisation” efforts were also highlighted as being a major driver of cloud budget growth. 

Infrastructure-as-a-service supporting AI leads cloud growth

All segments of the cloud market are expected to grow this year. However, infrastructure-as-a-service (IaaS) is forecast to experience the highest end-user spending growth at 25.6%, followed by platform-as-a-service at 20.6% 

“IaaS continues at a robust growth rate that is reflective of the GenAI revolution that is underway,” said Nag. “The need for infrastructure to undertake AI model training, inferencing and fine tuning has only been growing and will continue to grow exponentially and have a direct effect on IaaS consumption.”

Nevertheless, despite strong IaaS growth, software-as-a-service (SaaS) remains the largest segment of the public cloud market. SaaS spending is projected to grow 20% to total $247.2 billion in 2024. Nag added that “Organisations continue to increase their usage of cloud for specific use cases such as AI, machine learning, Internet of Things and big data which is driving this SaaS growth.”

The strong public cloud growth Gartner predicts is largely reliant on the continued investment and adoption of generative AI. 

Since the launch of intelligent chatbots like Chat-GPT, and AI image generators like MIjourney in 2022, investment exploded. Funding for generative AI firms increased nearly eightfold last year, rising to $25.2 billion in 2023. 

Generative AI accounted for more than one-quarter of all AI-related private investment in 2023. This is largely tied to the infrastructural demands the technology places on servers and processing units used to run it. It’s estimated that roughly 13% of Microsoft’s digital infrastructure spending was specifically for generative AI last year.

Can the generative AI boom last? 

However, some have drawn parallels between frenzied generative AI spending and the dot com bubble. The collapse of the software market in 2000 resulted in the Nasdaq dropping by 77% drop. In addition to billions of dollars lost, the bubble’s collapse saw multiple companies close up, and widespread redundancies. “Generative AI turns out to be great at spending money, but not at producing returns on investment,” John Naughton, an internet historian  and professor at the Open University, points out. “At some stage a bubble gets punctured and a rapid downward curve begins as people frantically try to get out while they can.” Naughton stresses that, while it isn’t yet clear what will trigger the AI bubble to burst, there are multiple stressors that could push the sector over the edge. 

“It could be that governments eventually tire of having uncontrollable corporate behemoths running loose with investors’ money. Or that shareholders come to the same conclusion,” he speculates. “Or that it finally dawns on us that AI technology is an environmental disaster in the making; the planet cannot be paved with data centres.” 

For now, however, generative AI spending is on the rise, and bringing public cloud spending with it. “Cloud has become essentially indispensable,” said Nag in a Gartner release last year. “However, that doesn’t mean cloud innovation can stop or even slow.”

Automatons have operated within warehouses and factories for decades. Today, however, companies are pursuing new forms of automation empowered by artificial intelligence (AI) and machine learning. 

AI-powered picking and sorting 

In April, the BBC reported that UK grocery firm Ocado has upgraded its already impressive robotic workforce. A team of over 100 engineers manage the retail company’s fleet of 44 robotic arms at their Luton warehouse. Through the application of AI and machine learning, the robotic arms are now capable of recognising, picking, and packing items from customer orders. The AI directing the arms relies on AI to interpret the visual input gathered through their cameras.

Currently, the robotic arms process 15% of the products that pass through Ocado’s warehouse. This amounts to roughly 400,000 items every week, with human staff at picking stations handling the rest of the workload. However, Ocado is poised to adjust these figures further in favour of AI-led automation. The company’s CEO, James Matthews, describes their approach for the future, wherein the company aims for robots to handle 70% of products in the next two to three years.

“There will be some sort of curve that tends towards fewer people per building,” he says. “But it’s not as clear cut as, ‘Hey, look, we’re on the verge of just not needing people’. We’re a very long way from that.”

A growing sector

Following in the footsteps of the automotive industry, warehouses are a growing area of interest for the implementation of robots informed by AI. In February of this year, a group of MIT researchers transposed their work in using AI to reduce traffic congestion in order to mitigate issues that arise in warehouse management. 

Due to the high rate of potential collisions, as well as the complexity and scale of a warehouse setting, Cathy Wu, senior author on a paper outlining AI-pathfinding techniques, discusses the imperative for dynamic and rapid artificial intelligence operations.

“Because the warehouse is operating online, the robots are replanned about every 100 milliseconds,” she explained. “That means that every second, a robot is replanned 10 times. So, these operations need to be very fast.”

Recently, Walmart also increased their AI systems in warehouses through the introduction of robotic forklifts. Last year, Amazon, in partnership with Agility Robotics, undertook testing of humanoid robots for warehouse work.

Words of caution

Developments in the fields of warehouse automation, AI, and robotics are generating a great deal of excitement for their potential to eliminate pain points, increase efficiency, and potentially improve worker safety. However, researchers and workers’ rights advocates warn that the rise in robotics negatively impacts worker wellbeing.  

In April, The Brookings Institution in Washington released a paper outlining the negative effects of robotisation in the workplace. Specifically the paper highlights the detrimental impact that working alongside robots can have upon workers’ senses of meaningfulness and autonomy. 

“Should robot adoption in the food and beverage industry increase to match that of the automotive industry (representing a 7.5-fold increase in robotization), we estimate a 6.8% decrease in work meaningfulness and 7.5 % decrease in autonomy,” the paper notes, “as well as a 5.3 % drop in competence and a 2.3% fall in relatedness.”

Similar sentiments were released in another paper published by the Pissarides Review regarding technology’s impact upon workers’ wellbeing. It is uncertain what the application of abstract terms like ‘meaningfulness’ and ‘wellbeing’ spell for the future of workers in the face of a growing robotic workforce, but Mary Towers of the Trades Union Congress (TUC) asserts that heeding such research is key to the successful integration of AI-robotics within the workplace.

“These findings should worry us all,” she says. “They show that without robust new regulation, AI could make the world of work an oppressive and unhealthy place for many. Things don’t have to be this way. If we put the proper guardrails in place, AI can be harnessed to genuinely enhance productivity and improve working lives.”