Universities Don’t Have a Data Centre Capacity Problem. They Have a Visibility Problem.
Universities hold more infrastructure data than ever, yet many still struggle to answer fundamental questions about what they own, what is operational, where capacity exists and which constraints are real. The challenge is not the absence of data. It is whether the information describing the infrastructure estate is accurate enough to trust.
Universities are preparing for growing demand from AI, research computing and increasingly data-intensive services. Infrastructure strategies are being reviewed, new compute environments are being considered and ageing facilities require decisions about investment, consolidation and replacement.
All of these decisions depend on something more fundamental: an accurate understanding of the infrastructure the institution already operates.
In theory, universities have never had more information about their estates. DCIM platforms record assets, rack locations and power consumption. Building management systems monitor environmental conditions and mechanical infrastructure. CAFM platforms contain facilities information, while network management tools, asset registers, procurement systems, spreadsheets and technical documentation provide further views of the environment.
Yet answering relatively basic questions can still require weeks of investigation.
An institution may struggle to establish exactly what equipment remains installed, which assets are still operational, who owns them, what is consuming power and where usable capacity exists. Constraints identified in one system may not appear in another, while the people with the most accurate understanding of the estate may be relying on knowledge accumulated over years rather than information captured in institutional systems.
The problem is not that universities lack infrastructure data. It is that the information is fragmented across systems, teams and individuals, and institutions cannot always trust it enough to support significant operational and investment decisions.
Infrastructure information decays
Physical infrastructure estates change continuously.
New equipment is installed, services are migrated, projects are completed and hardware is replaced. Servers are virtualised, research systems are introduced, applications move to the cloud and equipment reaches the end of its useful life.
The physical environment evolves, but the information describing it does not always evolve at the same speed.
An asset may be removed from a rack but remain recorded in the DCIM platform. A server may be switched off but remain connected to power and network infrastructure. A project may introduce new equipment without updating the central asset register, while power allocations can remain associated with systems that no longer exist.
Documentation presents a similar challenge. Information can accurately describe a facility when it is created, but gradually become less reliable as years of changes accumulate.
This process is rarely the result of a single major failure. It happens incrementally.
Moves, additions and changes are not consistently reflected across every system. Projects use different documentation standards. Decommissioning activity removes equipment physically but does not always remove it from infrastructure records. Ownership changes, responsibilities move between teams and experienced staff leave the institution.
Over time, the difference between the physical estate and its digital representation becomes wider.
This is infrastructure data decay.
The longer it continues without active reconciliation and governance, the more difficult and expensive it becomes to reconstruct an accurate picture of the environment.
Universities have multiple versions of the same infrastructure
The organisational structure of universities makes the problem more difficult.
Estates teams understand electrical and mechanical infrastructure. IT teams manage servers, storage and networks. Research computing teams operate specialist environments. Finance and procurement systems contain information about equipment that has been purchased. Projects introduce new technology and infrastructure, often supported by external suppliers.
Each team can maintain information that is accurate within its own boundaries while the institution still lacks a complete view of the estate.
A DCIM platform may contain thousands of asset records but lack reliable ownership information.
A building management system may provide detailed environmental and plant data without identifying the services dependent on that infrastructure.
An asset register may confirm that equipment was purchased but provide little confidence about whether it remains installed or operational.
Monitoring platforms can show current electrical demand without explaining which assets are responsible for that consumption or whether those assets still provide sufficient institutional value to justify the resources they consume.
The result is not necessarily bad data.
It is fragmented truth.
Different parts of the institution understand different layers of the same infrastructure, but nobody can easily combine those layers into a single view that is sufficiently reliable to support strategic decisions.
The Capacity Confidence Gap
Most universities can produce infrastructure information.
The more important question is whether decision-makers can rely on it.
A university may know the original design capacity of a data centre, operate a DCIM platform containing thousands of assets and collect millions of monitoring data points every year.
But those systems do not automatically tell the institution how much usable capacity exists, which equipment is still required, where resources are being consumed unnecessarily or which constraints genuinely prevent additional infrastructure from being deployed.
Answering those questions requires information from multiple sources to be reconciled against the physical environment.
The difference between possessing infrastructure information and trusting that information enough to make significant operational and investment decisions is the Capacity Confidence Gap.
That gap matters because infrastructure decisions are expensive and long-lived.
Major data centre upgrades can require millions of pounds of capital expenditure. Electrical and cooling infrastructure can take years to design, procure and deliver. Decisions about AI and HPC capacity can shape institutional research capability for a decade or more.
The information supporting those decisions should meet a much higher standard than simply being available.
It needs to be accurate enough to act upon.
Poor information creates expensive decisions
When infrastructure information becomes unreliable, institutions can make poor decisions in both directions.
Universities can underestimate the capacity they already own. Equipment that has been decommissioned may remain recorded as operational, rack space can appear occupied when hardware has been removed and power allocations can remain attached to systems that no longer exist.
Without physical audit and reconciliation, that capacity remains hidden. Institutions can then begin planning upgrades or additional infrastructure while existing resources could have been released, consolidated or used more effectively.
The opposite problem is equally significant.
Available rack space can create the impression that a facility has capacity to grow, while electrical distribution, cooling, resilience or infrastructure condition make additional deployment difficult or expensive. Monitoring systems may accurately report current consumption without providing sufficient information about future headroom or the impact of infrastructure failures.
Poor visibility also creates operational inefficiency. Projects can be delayed while teams manually reconstruct the estate. New compute requirements can be rejected because capacity appears unavailable. Equipment can remain powered and cooled because ownership is unclear and nobody is prepared to authorise its removal.
Perhaps most importantly, institutions can become dependent on knowledge held by a small number of experienced individuals.
In many infrastructure environments, the most accurate representation of the estate is not contained within the systems intended to manage it. It exists in the accumulated knowledge of engineers, infrastructure managers and facilities teams who have worked with the environment for years.
That knowledge is valuable, but it is also fragile.
When people change roles or leave the organisation, institutional understanding can disappear with them.
The financial and operational consequences are real. In one university environment, structured decommissioning and reconciliation activity improved confidence in DCIM asset records to approximately 95%. The improvement came not from implementing another software platform, but from physically validating the estate, removing equipment no longer required and reconciling infrastructure records against the real environment.
The exercise demonstrated a wider point: infrastructure visibility is created through active management of the estate, not through the existence of infrastructure data alone.
The institution owns the infrastructure, but who owns the complete picture?
Universities invest heavily in systems designed to manage infrastructure information.
A DCIM platform may have an owner. The building management system has an owner. Asset registers, network management platforms and CAFM systems each have teams responsible for maintaining them.
But ownership of individual systems is not the same as ownership of the overall infrastructure picture.
Someone needs to be accountable for ensuring that physical assets, capacity information, monitoring data, infrastructure condition and operational changes remain sufficiently aligned to support decisions.
Without that accountability, information begins to fragment again.
A university can invest significant time and money auditing its estate, improving asset records and implementing new tools, only to see information quality deteriorate once the initial project ends.
The objective should not be to achieve perfect infrastructure data at a single point in time. Large and complex estates will always contain uncertainty, and maintaining every piece of information with absolute accuracy would be expensive and unrealistic.
The objective is to establish the level of information quality required to make good decisions and create the operating capability needed to maintain it.
That is fundamentally a governance challenge.
Infrastructure intelligence is an operating capability
Improving infrastructure visibility does not begin and end with implementing a DCIM platform.
Technology is important. Accurate asset records, power monitoring, environmental data and capacity models all contribute to better infrastructure management.
But software cannot compensate for weak operating processes.
Equipment needs to be physically audited and infrastructure records reconciled against the real environment. Ownership needs to be established, while equipment that is no longer required needs to be formally decommissioned and removed from both the facility and the systems used to manage it.
Moves, additions and changes need to update the infrastructure record as part of normal operations rather than through periodic corrective exercises.
Power and cooling information needs to be understood in the context of the assets consuming those resources. Infrastructure condition and lifecycle need to inform capacity decisions, and responsibility for maintaining the integrity of the information needs to be clear.
This is not a one-off data cleansing exercise.
It is an operating discipline.
Audit. Reconcile. Baseline. Govern. Measure. Improve.
That is what turns infrastructure data into infrastructure intelligence.
Visibility before investment
Universities will make significant infrastructure decisions over the next decade.
AI, research computing, energy constraints, ageing estates and changing cloud economics will influence where workloads run and where capital is invested.
The quality of those decisions will depend partly on the quality of the information describing the infrastructure estate.
Universities do not necessarily lack infrastructure data. They lack confidence that the information they hold accurately represents the infrastructure they operate.
Closing that gap can reveal unused capacity, identify genuine constraints, reduce operational risk and allow capital investment to be targeted more effectively. It can reduce dependence on institutional knowledge and provide a stronger evidence base for decisions about consolidation, cloud, colocation and future infrastructure requirements.
Some institutions will need more data centre capacity. Others will need to upgrade ageing facilities, expand electrical and cooling infrastructure or consume capacity elsewhere.
But before universities commit significant capital to solving those problems, they should establish whether the decisions they are making are based on information they can trust.
Because the first infrastructure problem to solve may not be capacity.
It may be visibility.
