Insight Type: Blog

The cost of a prompt: translating AI workloads into UK household power use

Posted on in

Artificial Intelligence is frequently perceived as a purely digital phenomenon: a series of calculations and lines of code existing outside the physical realm. While a single generative ‘prompt’ requires minimal power, typically around 0.3 watt-hours, the cumulative effect of millions of daily queries creates a substantial physical footprint. This demand places the UK National Grid under significant pressure, driven not only by the processing hardware itself but by the critical cooling infrastructure required to support it.

The primary challenge is the volume of energy lost to inefficiency. High-performance GPUs are inherently power-intensive, but the environments housing them often compound this demand through inefficient thermal management. A single high-density AI rack in a legacy facility can consume enough additional energy on cooling overhead alone to power over 50 British homes for a year. This is a substantial, measurable volume of energy that current infrastructure standards often fail to reclaim.

Estimated energy use of common AI tasks

To understand the scale of the challenge, it is necessary to contrast the negligible feel of a single AI task with the massive infrastructure footprint required to support them at scale. The following table translates common AI activities into real-world energy equivalents.

AI activity (Single Task) Est. energy use Physical equivalent
Standard LLM text query 0.1 – 2 Wh Charging a smartphone
AI image generation 0.05 – 4 Wh Running a LED bulb for a few minutes or powering a phone screen for an hour
5-minute voice interaction 10 – 40 Wh Boiling water in a kettle
Large coding/debug session 30 – 200 Wh Using a laptop for a few hours
1-minute AI video generation 50 – 1,000+ Wh Running a washing machine cycle at the high end

In isolation, these figures appear insignificant. However, enterprise AI systems are used millions of times per day. At a modest enterprise scale of 10 million prompts per day, even low-energy tasks accumulate into annual energy requirements exceeding 10.5 gigawatt-hours. In a legacy 1.8 PUE environment, this scale generates over 8,000 MWh of preventable energy waste. This cooling overhead alone is enough to supply approximately 3,000 homes in the UK for twelve months.

Why legacy infrastructure is hitting a wall

Modern AI workloads operate at rack densities that can climb to 100kW or beyond for the largest hyperscale deployments. The majority of legacy server rooms in the UK were not engineered to handle even mainstream AI workloads, which typically demand 30kW to 40kW per rack. Standard cooling methods are often insufficient: they dissipate energy into the room rather than removing it from the source, leading to systemic inefficiency while attempting to maintain optimal operating temperatures.

Datum adopts a fundamentally different architectural approach. Our facilities move away from traditional raised-floor cooling, instead distributing air evenly across the full height of each data hall, with pressure and temperature sensors continuously adjusting cooling output to match conditions. This allows enterprises to deploy AI workloads at densities of up to 30kW per rack, without committing to the vast, day-one power contracts that hyperscaler facilities require. This delivers reduced waste and more consistent, dependable results for large-scale production environments.

Paying for heat instead of compute

The choice of infrastructure becomes most apparent when evaluating the cooling overhead: the proportion of total energy that goes towards non-compute functions such as cooling, power distribution, and other systems. In a legacy facility with a Power Usage Effectiveness (PUE) of 1.8, approximately 80p of every £1 spent on energy is consumed by non-compute overheads: cooling, power distribution, and auxiliary systems. At a Datum facility with a PUE of 1.25, this overhead cost is reduced to 25p.

Efficient cooling is a rigorous engineering discipline focused on reducing the hidden charge of wasteful cooling. Every kilowatt-hour saved on infrastructure overhead reduces environmental impact directly, whilst also freeing capacity that can be redirected to active compute or returned to the national electricity grid.

Why inefficiency destroys ESG targets

When scaling from a single rack to an enterprise deployment of ten AI racks, the 2,628,000 kWh of annual IT usage creates a massive divide in sustainability profiles. In a legacy facility with a 1.8 PUE, over 2.1 million kWh are wasted on overhead. Datum’s 1.25 PUE limits that waste to 657,000 kWh.

Maintaining a 10-rack AI deployment in a legacy site results in nearly 1.5 million kWh of preventable waste every year, based on the figures above. Based on the Ofgem 2026 benchmark of 2,700 kWh per household, a 10-rack AI deployment at Datum gives back enough energy to the UK grid to power over 535 homes for a full year. This represents the difference between sustainability as a slogan and sustainability as a practical engineering solution. It transforms the data centre choice into a primary ESG decision for any UK board.

The infrastructure efficiency mandate

The UK’s power grid operates within finite capacity limits. Every watt of power lost to inefficient cooling is a watt that becomes unavailable for new AI capabilities. Inefficiency is no longer just a financial burden: it is a physical barrier to scaling and national sustainability goals.

In the current climate, a 1.25 PUE is the necessary operational standard for high-density AI. Efficient operations must move beyond software optimisation and address the physical cooling overhead at the infrastructure level. To secure a sustainable power roadmap through 2026 and beyond, UK organisations must move beyond legacy downtime metrics and begin benchmarking their AI infrastructure against modern efficiency standards.

Understanding the efficiency gap is the first step in reclaiming power that is currently being dissipated as waste. Explore the technical specifications and engineering benchmarks at Datum to learn more about high-density cooling and utility-scale efficiency.

Back to Insights

Beyond cost: five critical factors when choosing a colocation provider

Posted on in

Value-add colocation

For many organisations, colocation is the logical next step for hosting, securing, and connecting their IT infrastructure, particularly as demands grow on IT. It is often a struggle, however, to differentiate one colocation provider from another.

We strongly argue that cost should not be the deciding factor when choosing a colocation provider. Seeking additional value-adds is critical to the success of any infrastructure move.

Five essential elements to evaluate before signing a colo contract

1. A long-term partnership

In theory, moving equipment away from a provider you are unhappy with should be simple. In practice, it is highly complicated and carries serious long-term implications. Due diligence before you commit is key to understanding exactly what a colocation provider has to offer your business.

It is important to consider what may happen in the future, both when times are good and when things might go wrong. This ensures you have a clear picture of how the provider can support your long-term strategy over the coming years.

2. The whole package

Given the long-term nature of colocation partnerships, it is highly beneficial to deal with a provider who lets you interact with its full team of staff. This includes professionals across:

  • sales and account management;
  • service management;
  • technical support and engineering.

Knowing who you will be dealing with after the contract has been signed is essential. It is equally important that they know exactly who you are and understand your business needs.

3. Service management

You must establish a strong relationship between your organisation and those responsible for identifying and rectifying problems within the facility. Effective service management criteria that your provider should offer include:

  • maintaining regular interaction and updates;
  • ensuring critical information is relayed to the relevant departments quickly;
  • preparing adequate, tested plans in the event of a major failure.

Do not overlook the actual migration. Securing a provider who will work closely with you to ensure a smooth transition can prove invaluable, paving the way for a successful colocation experience.

4. Physical location

The geographic location of the data centre is highly important. You must verify that the site is not prone to environmental risks, such as flooding, and that the facility maintains the exact level of physical and digital security your operations require.

Data centres located in strictly controlled environments, like our ultra secure London edge and Manchester campus sites, deliver total peace of mind for the safe-keeping and security of your IT infrastructure.

5. Value-added ecosystem

Data centre ecosystems, where available, provide tremendous value beyond basic space and power. Our partner ecosystem helps our clients connect with a network of partners offering lifecycle services that support different stages of the colocation journey.

Ready to explore your options?

Choosing a new colocation provider is not a decision that should come down solely to cost, nor to standard features that should be taken for granted. You must look for what makes the provider different, what they offer to uniquely support your organisation, and how they can help propel your business to the next level.

Why not visit our London edge and Manchester sites to see our facilities in action: MCR1 and our recently completed MCR2 new build data centre on an ultra-secure site in Manchester featuring an on-site police linked ARC, and FRN1 (our flagship data centre) and our ongoing construction of FRN2 at our London edge government grade site. Get in touch to book your tour.

Back to Insights

What to look for when evaluating a data centre partner

Posted on in

Choosing the right environment

Whether you are reviewing your current setup, planning a migration, or supporting wider digital transformation, the right data centre environment can strengthen service delivery and reduce operational risk. The wrong choice can create constraints that are difficult and costly to fix later.

Key areas to assess

1. Resilience (should be proven, not assumed)

A data centre should support continuous availability for critical systems and applications. But how is resilience is actually delivered? A resilient environment is not defined by one feature alone. It depends on how power, cooling, connectivity and operational processes work together to reduce the risk of disruption.

IT teams should assess:

  • power architecture and redundancy;
  • cooling design and operational continuity;
  • network resilience and path diversity;
  • monitoring and incident response arrangements;
  • maintenance processes and change control disciplines.

    2. Security (must cover both physical and operational risk)

    Security is often discussed in broad terms, but IT teams need to examine the detail. A strong data centre environment should protect infrastructure not only from unauthorised access, but also from operational failures and weak governance. For many organisations, a professionally managed data centre can provide stronger controls than an internal server room, particularly when infrastructure supports regulated or business-critical workloads.

    Areas to review include:

    • layered physical security controls;
    • 24/7 site monitoring;
    • access management policies;
    • security operations capability;
    • audit trails and reporting;
    • support for compliance requirements.

    3. Connectivity options (central to performance)

    For modern IT estates, connectivity is just as important as rack space and power – infrastructure needs reliable access to users, partner systems, public cloud platforms and disaster recovery environments. Good connectivity gives IT teams flexibility, and supports application performance, simplifies hybrid IT models and reduces the risk of being locked into a narrow network design.

    When evaluating a provider, IT teams should ask:

    • which carriers are available on site;
    • whether the facility is carrier-neutral;
    • what options exist for diverse network paths;
    • how easily cloud connectivity can be added;
    • what latency expectations apply for key destinations.

    4. Scalability (should match real operational needs)

    IT teams need an environment that can support change over time. Growth is rarely a linear process, and infrastructure requirements can shift due to acquisitions, new platforms, security demands or changes in application architecture. The goal is not only to meet current demand, but to avoid creating new constraints further down the line.

    A suitable data centre partner should be able to support:

    • short-term expansion needs;
    • phased migrations;
    • mixed infrastructure environments;
    • changing power density requirements;
    • future connectivity and cloud integration needs.

    5. Service quality (matters as much as the facility itself)

    Even a technically strong facility can become difficult to work with if operational support is weak. IT teams should consider what the service model looks like in practice. A data centre partner should feel like an extension of the internal IT function, not a passive landlord. Clear service management, transparency and operational responsiveness all have a direct impact on the experience of running infrastructure effectively.

    Questions worth asking include:

    • is there a dedicated engineering team on site;
    • what remote hands support is available;
    • how quickly can requests be handled;
    • what reporting is provided for power, temperature and performance;
    • how are incidents communicated and managed.

    6. Sustainability (should be examined in practical terms)

    Sustainability is now part of many infrastructure decisions, but IT teams need measurable indicators rather than broad claims. For organisations balancing environmental goals with technical performance, these factors can help support both governance requirements and long-term cost control.

    Useful areas to explore include:

    • energy efficiency performance;
    • renewable electricity usage;
    • cooling efficiency;
    • reporting on power usage effectiveness;
    • operational initiatives such as heat recovery.

    7. Business continuity (should be built into the environment)

    A data centre decision should support business continuity planning so IT teams should evaluate how the environment contributes to continuity, recovery and operational confidence. A dependable environment helps reduce the operational burden on internal teams and strengthens continuity planning across the wider organisation.

    That means reviewing:

    • resilience of core infrastructure;
    • site procedures for incidents;
    • monitoring and escalation arrangements;
    • options for geographic diversity;
    • support for disaster recovery design.

    Final thoughts

    For IT teams, evaluating a data centre partner means considering whether the environment will help support secure, resilient and adaptable infrastructure over time.

    The best decisions are usually made when technical, operational and strategic factors are assessed together. A strong partner should offer resilience, security, connectivity, flexibility and service quality in a way that supports both current priorities and future change.

    Ready to see these capabilities first-hand? Book a tour of our Farnborough and Manchester data centres today to see our secure, resilient, and highly connected environments in action.

    Back to Insights

    Quantum computing – what it is and what it might mean for data centres

    Posted on in

    The potential of quantum computing

    You might be forgiven for thinking that artificial intelligence is the only major tech innovation happening right now, but quantum computing has the potential to completely revolutionise the technology landscape, too. While still a developing field, it promises to break through the physical limitations of traditional computing.

    New horizons in processing power

    Currently, standard CPU power is limited by the number of calculations a computer can make per nanosecond. CPU chips have a finite capacity and while they are significantly faster than the processors of the 1980s and 90s, a strict limitation remains on their maximum calculation speed. Quantum computing removes this limitation almost entirely.

    Today’s leading ‘supercomputers’ consist of millions of CPU and GPU cores, yet they still struggle to efficiently make certain complex calculations. Take a simple seating arrangement as an example. There are over 3.6 million ways to seat just 10 people at a dinner table. A traditional supercomputer must take time after each calculation to verify and analyse that its work is correct before moving to the next sequence.

    Quantum mechanics handles this much faster. It creates vast multidimensional spaces to represent problems and relies on quantum wave interference. This allows the computer to perform all calculations simultaneously, whilst translating the data back into information we can understand.

    Imagine a traditional navigation interface working out how to get from point A to B, then to C, and eventually D. A quantum computer does not just calculate each route sequentially. It looks at every single possible permutation – including how to get from D to B, C to A, and B to C – all at exactly the same time.

    How could it change our lives?

    Traditional CPUs have advanced almost as far as physics will allow. Even with the shift from 7nm to cutting-edge 3nm and 2nm chips, traditional processors still face physical boundaries. To make large-scale, complex calculations, you need to network more processors together, which takes up massive amounts of physical space.

    Quantum chips are capable of significantly more output at the same size. By working together, they can perform exponentially more calculations per second, meaning complex equations are solved instantly. For example, a quantum computer could rapidly analyse millions of variables to determine the exact optimal fuel needed to sustain a flight to Mars. It destroys the physical limitations of traditional science.

    The main hurdle we face right now is something called quantum decoherence. This makes the information received from quantum computers difficult to decode. The issue is not the information itself, but rather the blistering speed at which it returns. The quantum complexity of the output means the human mind and traditional interfaces can only translate so much at once.

    When can we expect it?

    In many forms, quantum computing already exists and has done for several years. We have seen major breakthroughs with quantum chips and the dimensional algorithms required for the data to exist. However, it remains a highly complex science. While we can ask a quantum computer for answers, we cannot always translate the returned data correctly. These ongoing translation efforts are currently slowing down practical progress.

    That said, researchers hope that massive breakthroughs in AI and machine learning over the next five to ten years will allow for rapid data translation. As hybrid systems evolve, we will likely see setups where quantum computers do the heavy lifting, while traditional supercomputers exist purely to translate that raw data into a format we can easily understand.

    Quantum’s impact on data centres

    Rapid advancements have sparked heavy investment across the sector, and this has led to serious discussions about how these powerful computers will integrate with industries like the data centre sector.

    Ultimately, data infrastructure will need to evolve to support quantum technology. While it is still developing and unlikely to replace the classical computing infrastructure that data centres currently host anytime soon, it shows promise for sustainability. If quantum computing can achieve tasks in seconds or minutes that would otherwise take hours or days, we are clearly going to need less energy to run our most intensive workloads.

    Discover our colocation data centres

    We’re building brand new state-of-the-art colocation facilities, positioning us well to cope with new technology as it develops. Come and see our data centres in action – London edge and Manchester – two of the most economically active regions of the UK and tech hubs.

    Back to Insights

    Is your data centre AI-ready? A 10 layer audit

    Posted on in

    The prevailing discourse surrounding AI in large enterprises focuses primarily on model training. Public discussion frequently highlights massive GPU clusters, complex data pipelines, and extensive cloud compute cycles. However, for organisations seeking to derive tangible business value, training is merely the initial phase. True operational value is realised when AI is deployed within functional environments, a stage often referred to as operational AI. This involves processing data at the point of origin: for example, predicting industrial equipment failure, detecting banking fraud instantaneously, or facilitating autonomous warehouse robotics.

    As the focus shifts from training to deployment, the critical performance metrics change. Success is no longer defined solely by raw compute power or model scale: it depends on latency, system integration, and proximity to the operational environment. Establishing a presence at the network edge is vital. AI operating locally often outperforms distant cloud-based models because critical information must be processed in milliseconds rather than minutes.

    This ten-part evaluation is one way to look at the AI stack, though others in the industry define it using varying numbers of layers. What follows demonstrates that achieving readiness for operational AI requires a sophisticated, multi-layered infrastructure.

    Domain 1: The intelligence foundation

    Layer 1: Data layer

    To generate accurate predictions, AI requires high-quality, structured data. Transmitting raw data from IoT devices or sensors directly to the network often creates unnecessary volume. An initial phase of data cleansing, reduction, and standardisation at the local site ensures the AI only receives the most relevant datasets.

    This initial stage is critical at the network edge. Uploading all raw data to a remote cloud server can result in significant network congestion and latency. By performing data preparation at the source, the AI accesses usable data more rapidly, while the organisation reduces costs associated with storage and bandwidth.

    Layer 2: Training layer

    Even operational AI models require continuous optimisation. Utilising local data to adjust a model enhances its accuracy and adaptability. However, large-scale retraining is typically better suited for central cloud environments where vast processing power is available.

    The role of the edge in this layer is one of refinement: learning from the local environment to prevent model drift. This ensures that predictions remain accurate and appropriate for the specific deployment site.

    Layer 3: Model layer

    An AI model’s weights and parameters represent significant intellectual property. This data must be kept in storage that is secure, high-speed, and situated near the point of decision-making. Localised storage ensures the AI can respond instantaneously without waiting for data to complete a round trip to a central server.

    Proximity also relates to governance. By maintaining the model alongside the data, organisations can more easily comply with data residency regulations and mitigate the latency risks that threaten mission-critical applications.

    Domain 2: The execution hub

    Layer 4: Inference layer

    The inference layer is where the model processes data to produce actionable outputs: such as identifying equipment anomalies or dynamically adjusting inventory levels. The speed at which these conclusions are reached defines the system’s utility.

    To meet enterprise standards, the inference engine must process high volumes of information with minimal delay. Running inference on local systems reduces reliance on remote cloud services, prevents network saturation, and enables the near-instantaneous responses required for daily operations.

    Layer 5: Agentic layer

    The agentic layer involves AI that acts autonomously rather than merely responding to prompts. These self-operating agents can modify processes, trigger alerts, or execute procurement tasks independently. Their effectiveness depends entirely on the speed of information retrieval.

    If an agent receives a forecast from one location and must act in another, any delay in data transmission can degrade performance or result in system failure. For autonomous AI to be dependable, all components must communicate with negligible latency.

    Layer 6: Application layer

    The application layer serves as the bridge between AI decisions and the end-user. This may take the form of a visual dashboard, an API for other software, or a direct integration into an existing operating system. Users must receive AI insights without lag to ensure they remain actionable.

    Situating the application layer at the physical site of compute reduces delays and simplifies troubleshooting. It allows engineers to monitor the AI’s output alongside the source data in real time, rather than navigating distributed data across multiple cloud regions.

    Domain 3: The operational engine

    Layer 7: Deployment layer

    The deployment layer involves more than simple model distribution. It requires a framework for safe, predictable updates and version control. While tools like Kubernetes and CI/CD pipelines facilitate this, a resilient compute foundation is necessary to manage these heavy workloads. Without specialised edge locations, deployment can become inconsistent or suffer from performance degradation.

    Layer 8: Integration layer

    AI does not operate in isolation. Middleware, APIs, and carrier-neutral connectivity allow models to integrate with existing enterprise systems and IoT devices.

    Applications where speed is paramount, such as autonomous shipping or predictive maintenance, require reliable and swift connections. Standard enterprise networks often lack the necessary bandwidth: therefore, the quality of your connectivity is a primary element of AI readiness.

    Layer 9: Hardware layer

    AI workloads depend on high-performance GPUs. Traditional server architectures often struggle with the thermal output and electrical demands of modern graphics processors. An AI-ready facility must support 30kW or more per rack and provide specialised cooling to prevent thermal throttling.

    Edge processors, such as NVIDIA Jetson, allow for high-performance computing at the point of data collection. These configurations provide the necessary compute density while remaining functional in industrial settings. Datum’s facilities are engineered to manage these high-density requirements, ensuring that hardware remains at optimal temperatures even under sustained heavy load.

    Domain 4: Trust & governance shield

    Layer 10: Security & governance

    An AI model represents significant capital investment and intellectual property. Simple software encryption is insufficient for high-value assets. The physical location of the hardware is a critical security consideration: this requires stringent access controls, constant monitoring, and auditable compliance.

    NSI Gold Approved BS 5979 monitoring provides a comprehensive security framework for confidential data. Physically safeguarding your hardware ensures that models cannot be tampered with or exfiltrated. Datum’s data centres provide audited security, 24/7 surveillance, and rigorous access protocols to protect your proprietary models at the point of compute.

    The readiness verdict

    AI readiness is not defined by a single component or a prominent GPU cluster. It requires a balanced approach to data, compute power, connectivity, and governance. A successful operational AI strategy must reconcile performance, security, proximity, and scalability.

    While massive centralised facilities handle global training, the success of AI in practice depends on the localised layers: preprocessing, inference, and secure edge storage. By bringing these specific capabilities closer to the source of data, organisations can implement AI that is privacy-first, resilient, and sustainable..

    Transitioning from a pilot project to a functional edge AI infrastructure requires specialist architectural insight. Contact Datum’s team to evaluate your infrastructure and plan a system designed for the rigorous demands of real-world AI.

    Back to Insights

    The problem of data centre ‘phantom investments’

    Posted on in

    Announced but not seen through

    It’s not uncommon for governments and developers to announce massive deals to build new data centres, promising thousands of jobs and vast economic growth. However, recent media investigations (Guardian, 2026; Financial Times, 2025) have painted a very different picture, claiming that many of these promised projects simply do not exist. Economic experts have labelled these projects that don’t come to fruition ‘phantom investments’. They serve primarily to artificially inflate economic impact figures, creating an illusion of growth without delivering any physical infrastructure. Furthermore, they cause grid connections to stall.

    The strain on utilities

    Data centres are critical infrastructure – technological and digital development cannot continue without them. When developers make bogus claims about upcoming projects, they muddy the waters for the entire industry. Often, developers rush to stake claims on power capacity long before they secure the necessary capital or develop a concrete strategy for construction. Estimates suggest that a staggering 90% of these power connection requests are actually false.

    This flood of speculative requests places an enormous burden on utility companies. Tasked with managing finite power resources, grid operators must somehow determine which projects are real and which are merely speculative.

    Faced with this uncertainty, utilities understandably hesitate to allocate power. They cannot risk committing grid capacity to a project that will never materialise. This hesitation slows down the entire development cycle, creating bottlenecks that hinder the companies ready to build actual infrastructure.

    According to Oxford Economics, two main factors explain the massive discrepancy between initial connection requests and the actual power draw expected from mature data centres:

    1. Speculative requests: the removal of connection requests that are unlikely to ever proceed accounts for almost 90% of the difference between total requests and actual power draw.
    2. Capacity overestimation: the remaining gap is explained by the difference between the massive capacity developers initially request and the actual, much lower power draw a facility will need at maturity.

    Genuine construction projects

    It takes immense effort, rigorous planning, and substantial time to bring a data centre online. Constructing a data centre is much like an iceberg. The final physical build – the part everyone sees – is just the tip. All the critical work that happens before a shovel ever enters the ground remains invisible to the public.

    Long before developers can make responsible public announcements, they must complete an exhaustive list of preliminary tasks. This includes detailed feasibility studies, ensuring stringent regulatory compliance, and drafting both conceptual and detailed designs. Teams must conduct thorough risk assessments, secure numerous planning approvals, and carefully select vendors. Only after finalising complex utility contracts and sourcing materials can the actual physical work begin.

    Understanding this complexity highlights why responsible planning matters. Navigating the extensive groundwork is a challenging but necessary part of the process. For instance, the completion of our MCR2 facility in Manchester in June 2025, alongside the start of construction of FRN2 at our site in Farnborough towards the end of last year, represents the culmination of this invisible iceberg. These milestones were achieved only after an extensive period of rigorous preparation.

    Responsible planning

    To ensure the continuous growth of our digital economy, the industry must prioritise genuine development over speculative projects.

    As for us, we’re focusing on what we do best: growing our data centre infrastructure so we can continue delivering high quality service, resilient colocation, and bespoke solutions with a personal touch – choosing to let our services speak for themselves rather than joining the race to make ever-bigger promises. To find out more about our colocation offering, get in touch or come and visit one of our sites in Manchester or Farnborough.

    Back to Insights

    How data centre efficiency solves the 2026 ESG reporting challenge

    Posted on in

    ESG reporting is moving from the back office to the boardroom. Under the latest UK mandates, sustainability disclosures must be integrated into the annual financial report. This transition ensures they carry the same weight of director accountability and governance as traditional financial statements. These disclosures will be signed off by directors, with a new requirement to explicitly state the level of third-party assurance and audit treatment applied to the data.

    For many businesses, this presents a fresh set of issues, and data centres have become pressure points. Ineffective or secretive practices can bring about legal, reputational, and financial risk. Beyond environmental concerns, this is a fundamental governance issue. Your data infrastructure management directly impacts ESG credibility, investor confidence, and potential personal liability.

    The 2026 ESG reporting landscape

    Fully in line with ISSB guidance, the final UK SRS S1 and S2 standards were issued in February 2026 for immediate voluntary use. The ISSB provides the global baseline for these disclosures to certify that sustainability data is as consistent and reliable as financial accounting. These new standards are designed to modernise and eventually succeed the SECR framework. While SECR focused primarily on reporting energy totals, the new UK SRS requires companies to demonstrate how climate risks impact their long-term financial value. This shift introduces mandatory, director-signed and assurance-graded ESG reporting from January 2027. Corporations can no longer rely on high-level narratives. Every disclosure now requires traceable, defensible evidence.

    There is a transition window. Companies get a two-year reprieve from climate-related reporting obligations, allowing them to focus on energy and emissions data. This makes infrastructure energy reporting the most quantifiable and auditable starting point, and the most efficient path to becoming fully ESG-compliant.

    Datum is a year-1 pragmatic solution. We are a verified partner for ESG directors, offering GRESB-rated data infrastructure and a low design PUE. Our systems provide meter-verified energy data at source, enabling truly credible emissions reporting from the start. In practice, Datum’s facilities ensure energy use is measurable, traceable, and defensible under scrutiny. This removes one of the most significant layers of ESG risk.

    Why data centres are ESG pressure points

    Data centres are long-duration emitting assets. Cooling systems and power supplies run continuously, while backup generators are regularly tested and occasionally activated, contributing to overall scope 1 and 2 emissions. The new SRS S1 and S2 UK ruleset brings purchased infrastructure into Scope 3. This is compounded by the fact that your third-party data centres are also directly affecting your ESG audit risk.

    Auditors need an emissions lineage. They want to see where the numbers come from, how they are measured, and whether the methodology is independently verifiable. Without that, infrastructure is a blind spot in your reporting.

    Datum eliminates that blind spot. Our real-time, meter-level reporting gives you a complete audit trail from energy consumption to emissions estimates. Estimation risk disappears, and ESG teams end up with data that is not just accurate but also audit-ready. Transparent and efficient data centres reduce reporting overheads while mitigating potential liability.

    Efficiency is a balance sheet solution

    ESG compliance carries direct financial consequences rather than being a purely regulatory exercise. Soaring numbers of British firms are adopting internal carbon pricing (CFP Energy, 2025), meaning that inefficient third-party infrastructure now carries a direct, measurable cost on a client’s internal balance sheet. Every tonne of CO₂ can be priced, which, in turn, translates into financial exposure to emissions. Inefficient infrastructure increases the cost, whereas efficient systems reduce it.

    Datum’s performance illustrates this clearly. As an active participant in the Global Real Estate Sustainability Benchmark (GRESB), we were ranked number 1 in our European peer group in 2025 with a score of 93/100. Governance came in at 23/24 and management at 37/40, which are significantly above sector benchmarks. This shows that Datum offers IFRS S1-aligned governance strength alongside engineering excellence.

    For CFOs, this matters. Efficient, well-governed infrastructure reduces reporting overhead, minimises estimation risk, and boosts credibility with both auditors and investors. GRESB validation also provides the third-party assurance most ESG auditors prefer. In practice, this allows you to reduce time spent defending numbers and increase focus on strategic carbon reduction.

    Tackling greenwashing and data integrity

    Legal compliance is a critical factor when reporting on ESG. According to CMA guidance and the DMCC Act 2024, businesses may be liable for repeating false information about a supplier’s environmental sustainability claims. Penalties can be as high as 10% of worldwide turnover. Even nonprofits and “100% renewable” declarations based solely on REGOs are coming under increasing challenge.

    Estimation risk is a legal liability as much as a logistical inconvenience. Datum solves this by enabling the use of real-time, verified, metered energy usage instead of soft estimates. With our systems, companies can provide proof for any claim they make and eliminate the risks of greenwashing or audit non-compliance. Audit-ready infrastructure safeguards the company and its directors, transforming legal liability into a manageable part of your ESG strategy.

    Sovereign sustainability: solving the dual mandate

    The UK’s Cyber Security and Resilience Bill has added an extra layer of obligation. Data centres with a capacity greater than 1MW are now classified as Critical National Infrastructure, so firms must adhere to sustainability and security standards. Older sites have struggled to meet the dual demands of energy efficiency and regulations.

    Datum’s London–Manchester presence serves this two-pronged requirement. We provide data residency, infrastructure sovereignty, and strong carbon reporting in secure, resilient facilities. When constructing MCR2, we integrated closed-loop heat reuse technology to allow the capture of waste heat, which can be used to warm a development of new homes planned adjacent to our data centre. This approach demonstrates sovereign sustainability, showing that energy efficiency and compliance are compatible with addressing a country’s digital infrastructure requirements.

    Strategic selection of the location also serves as a solution. By placing facilities on key transit lines, Datum minimises carbon footprint and regulatory overhead, ensuring that firms’ ESG and security requirements are met.

    Future-proofing your data strategy

    Data centre strategy now influences ESG audit results, director liability, and the credibility of carbon reporting. These factors far exceed simple questions of uptime or cost. Efficient and transparent infrastructure is an essential control lever for ESG and a competitive advantage.

    You do not want your data centre to be an auditing bottleneck. Ensure your 2026 infrastructure data is audit-ready and compliant with the latest UK standards. Book an appointment with a Datum expert today to learn how our GRESB-rated facilities can help you de-risk your upcoming sustainability disclosures.

    Back to Insights

    Our sponsorship of the local pan-disability football club

    Posted on in

    Supporting community and inclusive sport

    Community sports teams offer a vital support network and a sense of belonging, so we’re proud to sponsor the Wythenshawe & Norbrook Pan-Disability Football Club.

    About the team

    The Wythenshawe & Norbrook Pan-Disability Football Club was formed seven years ago following the closure of Wythenshawe United, a team previously run by a local community centre. Determined to keep the players together, Tommy, a former assistant manager at the Royal Oak community centre, took the reins.

    Today, the club supports 22 players, operating as two distinct but united squads: Norbrook FC and Wythenshawe Adult Disability Football Club. The structure is entirely inclusive. All players are over the age of 18 and live with various disabilities, including ADHD and specific learning difficulties. It is a mixed squad featuring both men and women, with ages ranging from 18 right up to their 61-year-old goalkeeper who continues to be unphased by throwing himself onto the ground in a bid to save goals. In some cases, the players’ carers also attend the sessions.

    Overcoming funding challenges

    The team is entirely self-funded. The players pay weekly subscriptions to cover basic operating costs. When the team travels for away fixtures or tournaments – such as matches in Northwich or Stockport – the players often have to contribute additional funds to cover minibus hire.

    These financial hurdles can place a strain on the players and the volunteers who dedicate their time to running the club. Despite these obstacles, the management team has consistently worked hard to ensure that financial limitations do not prevent the players from doing what they love.

    Our sponsorship

    Our contribution has directly funded brand-new kits for the entire squad. The club now proudly wears two distinct strips: a red kit for the A-team and a blue kit for the B-team. We’re pleased that the kits have had a substantial effect on team morale and have generated some healthy competition within the ranks, with players highly motivated to score goals and earn a promotion to the red A-team.

    Training, competing and celebrating together

    The team trains every week at the Woodhouse Park Lifestyle Centre in Wythenshawe and competes in the local Power League, which means regularly facing off against other disabled teams across Greater Manchester, including Stockport School, Oldham FC, Altrincham FC, and Stockport Summer FC. A highlight is the annual end of season tournament – the Cheshire FA tournament held in Northwich. The team received a winner’s trophy last year and was a runner up in the previous year.

    Beyond the medals and competitive fixtures, the club’s true success lies in its community spirit. Because many of the players cannot easily visit traditional social venues to celebrate their victories, Tommy opens up his own home. After a successful match, he invites the team back to his custom garden bar. The players bring their own drinks and celebrate their achievements in a safe, welcoming, and familiar environment.

    Speaking to Tommy and seeing the team in action really brings home how this team exemplifies the very best of grassroots sport, and we’re honoured to stand beside them and support their ongoing journey. We wish the team all the best in their upcoming matches and are looking forward to hearing about their progress.

    Back to Insights

    The UK’s AI corridor: why a London–Manchester infrastructure strategy is now essential

    Posted on in

    The UK’s AI growth map: why demand clusters around Manchester and London

    The SAS AI Cities Index 2025 highlights a clear concentration of AI activity across two primary UK hubs. London remains at the forefront, sustained by enterprise adoption, financial services, and a high density of R&D labs. However, Manchester has emerged as the fastest-growing AI city outside the capital. This is backed by a 184% increase in AI company registrations over a five year period leading into 2026, signalling a shift toward production-scale deployments, analytics platforms, and industrial AI applications.

    This clustering of demand is intentional, driven by connectivity, talent, and network effects. Enterprises place their R&D and high-compliance workloads in London & the South East to be near clients, regulators, and research facilities. Manchester, however, offers a production-focused ecosystem connected to northern talent pools with lower congestion.

    Infrastructure demand follows these patterns. AI workloads are data-intensive, have tight latency requirements, and are power-hungry. Facilities capable of handling such workloads must be close to where the AI is developed and used. This is especially crucial for decision-makers considering their cloud versus colocation options. While the cloud is a flexible option, high-density AI-ready colocation near innovation hubs often offers lower latency, greater control, and stronger reliability.

    As a result, Manchester and London are creating the UK’s AI corridor, playing complementary roles within the national infrastructure.

    Manchester: the connectivity powerhouse for production AI

    Manchester serves as a critical AI network and performance powerbase. In fintech, automation systems, and analytics platforms, production workloads are dominated by applications requiring real-time capabilities. For end-users, this latency is offset by consistent experiences, faster decisions, and clear business value.

    Datum’s MCR2 facility positions Manchester at the heart of this network. With LINX Manchester direct peering, AI inference traffic can stay local without traversing London. This provides a tangible performance and user-experience benefit. MCR2 is one of a handful of recently constructed northern facilities designed for high-density requirements, offering up to 30kW of rack density, carrier-neutral connectivity, and a resilient design.

    The facility acts as a benchmark for teams scaling production AI, providing high-density infrastructure, operational flexibility, and proximity to the businesses rolling it out.

    London & the south east: enterprise scale and sovereign AI

    Where Manchester powers production, London & the South East fuel enterprise AI, R&D, and regulated work. High-compliance industries, such as finance and defence supply chains, require infrastructure that supports sovereignty, control, and stringent operational stewardship. This is where Farnborough, acting as a London-edge location, becomes a significant strategic asset.

    Farnborough offers sub-millisecond latency to London without the grid restrictions and space constraints of inner-city sites. With its second Farnborough data centre under construction (FRN2), Datum’s scaling allows its client organisations to grow without capacity restraints. For businesses, this provides the physical control and jurisdictional influence required for sensitive tasks.

    At the Farnborough campus, businesses can utilise high-performance AI infrastructure near London’s commercial ecosystem without the limitations of urban sites. This high-security environment is built for real-world AI workloads, enabling resources to be deployed without excessive capital expenditure. As regulatory compliance and business operations grow more complex, combining proximity, speed, and sovereign control at a single site provides a decisive competitive advantage.

    Establishing Farnborough as the London-Edge gives enterprise AI teams the freedom and resilience to meet current demands while accommodating future scale.

    AI-ready infrastructure: the shift from legacy to high-density colocation

    AI growth is currently shifting away from legacy infrastructure and toward specialised colocation. Legacy 5kW-era data centres cannot achieve the sustained density required by contemporary GPUs. Datum’s operational MCR2 and the currently under construction FRN2 data centres offer the AI-ready infrastructure these workloads require. High-density construction, carrier-neutral access, and free-cooling design work together to ensure maximum performance with minimal power consumption.

    Many legacy data centres struggle with constraints on power and cooling. Datum’s solution enables high-density AI workloads to be scaled efficiently, providing an environment where resilience, efficiency, and future-proof scalability are realised. This transition from low-density legacy sites to AI-ready infrastructure is fundamental for organisations moving from pilot projects to enterprise-scale AI.

    Why dual-site strategy is becoming the default for AI resilience

    Depending on a single site for mission-critical AI introduces significant operational risk. Localised outages, regional grid pressure, and latency fluctuations are direct threats to high-density loads. A dual-site approach between Manchester and Farnborough provides the resilience and flexibility required for modern workloads. This architecture allows teams to balance tasks between production and development, or between inference and support modes, across two distinct geographic regions.

    Regulatory mandates are now the primary driver for these infrastructure decisions. Under the Cyber Security and Resilience Bill, data centres with a capacity of 1MW or more are classified as Critical National Infrastructure (CNI). This 2026 status requires organisations to demonstrate high levels of operational redundancy and geographic failover. Furthermore, frameworks like DORA (Digital Operational Resilience Act) have made high availability and operational stability the baseline for enterprise AI.

    The pairing of Manchester and Farnborough offers a strategic solution with both sites delivering resilience, low latency, and enterprise-scale sovereign infrastructure for sensitive or regulated workloads. Together, they form a robust, high-speed UK AI corridor. As AI use cases expand, organisations are increasingly adopting this dual-site strategy to ensure they meet 2026 compliance standards without compromising on speed or control.

    Connecting the UK’s high-performance AI corridor

    Demand for UK AI is accelerating across the Manchester and London corridors, increasing the need for high-density, AI-ready data centre capacity. The performance, scale, and resiliency needs of production AI can no longer be met by legacy facilities. Datum supports both regions with purpose-built infrastructure that enables organisations to deploy, scale, and secure their AI workloads.

    Is your infrastructure AI-ready? Contact us to explore how our Manchester and Farnborough facilities can support your AI and HPC workloads.

    Back to Insights

    Data centres in a circular economy

    Posted on in

    Balancing data centre demand and sustainability

    Despite their status as critical national infrastructure (or, perhaps, because of it), data centres face heavy and constant scrutiny regarding their power consumption and environmental impact. As such, the data centre industry must increasingly demonstrate how it can reliably support digital demand whilst doing everything possible to operate sustainably. Might the solution lie in adopting the principles of the circular economy?

    What is the circular economy?

    A circular economy is an economic model designed to minimise waste and make the most of available resources – it offers a sustainable approach to managing resources, reducing waste, and improving operational efficiency. Instead of the traditional linear model of ‘take, make, and dispose’, the circular approach focuses on keeping materials, products, and resources in use for as long as possible.

    Because data centres are energy-intensive facilities that rely heavily on complex hardware, infrastructure, and energy resources, they are perfect candidates for adopting these circular principles.

    Maximising resource efficiency

    Power and cooling are two of the most resource-hungry elements of data centre operations. Transitioning to renewable energy sources, such as wind, solar, and hydroelectric power, allows facilities to grow without reliance on carbon-intensive fossil fuels, whilst deploying energy-efficient cooling systems can dramatically reduce the power required to maintain optimal server temperatures, directly lowering facilities’ overall energy demands.

    Lowering operational costs

    By implementing energy-efficient systems, data centre operators can significantly reduce their long-term operational costs. Heat recovery systems like the heat reuse capability we built into our recent Manchester data centre construction project (MCR2), serve as an excellent example of circularity in action. Servers naturally generate heat, which facilities traditionally vent outside. A circular approach captures this waste heat and repurposes it to warm nearby commercial buildings, homes, or agricultural facilities. This turns a costly waste byproduct into a valuable resource and creates new efficiencies.

    Designing for lifecycle management

    Effective lifecycle management is crucial for maintaining a sustainable data centre, and involves carefully managing how physical assets, from server racks to cooling infrastructure, are renewed and written off. The best-case scenario involves designing data centres from the outset with appropriate, forward-looking technology that prioritises sustainability. Doing so ensures the facility can handle increasing computational loads without requiring expensive, highly disruptive capital upgrades later.

    Meeting rigorous sustainability standards

    Building a new greenfield data centre enables operators to incorporate cutting-edge sustainable technologies and circular designs from the outset. However, it demands significant raw materials and land. Upgrading a legacy brownfield site, meanwhile, comes with its own challenges but fits well into a circular economic system. Retrofitting older facilities with modern, energy-efficient technology requires meticulous planning but repurposes existing buildings and infrastructure. For example, at MCR1 in Manchester, we upgraded an existing office building, while MCR2 involved demolishing a derelict job centre on our secure campus. Similarly, our ongoing FRN2 project in Farnborough is redeveloping a car park within our existing boundary, which helps reduced embodied carbon through fewer enabling works and shared infrastructure. We conduct full embodied carbon assessments for all data centre constructions to measure and minimise emissions associated with the build in order to move towards a more circular and resource-efficient system, and we are part of the GRESB scheme, which allows us to measure, benchmark, and improve our ESG performance.

    Even backup systems can align with these sustainability goals. While data centres must maintain backup power to guarantee uptime, they no longer need to rely on diesel fuel. We’ve been using hydrotreated vegetable oil (HVO) in our backup generators since 2022 to cut greenhouse gas emissions and support our broader sustainability targets.

    The path forward

    The demand for digital infrastructure will only continue to accelerate. By embracing the circular economy, data centre operators can continue to provide the critical digital infrastructure global economies require whilst minimising their impact on the environment.

    If you would like to find out more or arrange a tour of our Manchester and Farnborough data centres, please get in touch with our team today.

    Back to Insights