Could Universities Share Data Centre Capacity? 

UK universities collectively own significant data centre infrastructure, but that capacity is not evenly distributed. While some institutions face constraints on power, cooling and physical space, others have invested in facilities with capacity that may remain underused for years. As infrastructure demand grows and capital becomes harder to secure, should universities make better use of what already exists within the sector?

Universities are entering a period of growing infrastructure demand. AI and research computing are increasing requirements for power, cooling and high-density compute environments, while ageing facilities are forcing institutions to consider whether existing data centres should be upgraded, expanded or replaced.

These decisions are being made at a time when higher education is under increasing financial pressure and being encouraged to collaborate more effectively. Research undertaken through the Universities UK Transformation and Efficiency Taskforce has called for a new era of collaboration across the sector, identifying shared services and infrastructure as an opportunity to improve efficiency and make better use of institutional resources.

Universities already collaborate extensively where infrastructure is expensive, specialist or capable of delivering greater value when used collectively. The Janet Network provides national digital infrastructure for research and education. Universities access national computing facilities, specialist research infrastructure and increasingly powerful shared HPC and AI resources without needing to own the underlying assets themselves.

Institutional data centres, however, have largely remained outside this model. Most were developed to serve the university that owns them, with investment, capacity planning and operational decisions based primarily on local requirements.

That approach has served institutions for decades. But as infrastructure becomes more expensive to build, electrical capacity more difficult to secure and compute demand increasingly uneven, it is reasonable to ask whether every university should continue to solve its data centre requirements independently.

Capacity is becoming harder to create

Creating significant new data centre capacity requires capital, power, time and specialist capability.

The International Energy Agency reported that global electricity demand from data centres increased by 17% in 2025, considerably faster than the 3% growth in overall electricity demand. In the UK, demand for electrical connections has become strategically important enough that the Government is reforming the connections process to prioritise projects considered important to economic growth and national infrastructure, including data centres.

Universities face the same underlying pressures, but from very different starting positions.

Some operate facilities approaching the limits of their electrical, cooling or physical capacity. Others have recently invested in resilient infrastructure designed to support institutional growth over many years. Consolidation, decommissioning and changes in workload placement can also release capacity within existing facilities.

There is no reliable sector-wide dataset showing how much suitable spare university data centre capacity exists. That absence of information matters because universities may be preparing to commit substantial capital to new infrastructure without knowing whether suitable capacity already exists elsewhere in the sector.

The proposition is therefore not that universities should stop investing in data centres. Some institutions will unquestionably require new facilities and significant infrastructure upgrades.

The more interesting question is whether every capacity requirement should automatically result in another independent infrastructure investment.

Sharing capacity is more complicated than sharing space

At first glance, the model appears straightforward. One university has available capacity, another needs somewhere to host infrastructure, and an agreement allows both institutions to benefit.

The host university receives income that contributes towards the cost of operating and maintaining its facility. The consuming institution avoids, or at least defers, the capital expenditure and time required to build or expand its own data centre.

In practice, available rack space is only the beginning of the calculation.

The host institution needs to understand how much power and cooling capacity can be committed without restricting its own future requirements. Energy consumption needs to be measured and charged appropriately, while the costs of cooling, maintenance, monitoring, security and operations also need to be recovered.

Long-term infrastructure investment creates a more difficult question. UPS systems, generators, chillers and electrical infrastructure require major maintenance and eventual replacement. Another institution’s equipment consumes part of the capacity and useful life of those assets.

A pricing model that recovers only electricity and short-term operating costs risks leaving the host university subsidising the service. But if every element of risk, operations, future capital investment and institutional overhead is passed through, the proposition may become no more attractive than commercial colocation.

A viable model therefore depends on something more sophisticated than identifying spare racks. Both institutions need to understand the genuine cost, value and risk of the arrangement.

Collaboration does not remove institutional risk

Universities collaborate extensively, but institutional trust cannot replace clear governance.

Hosting another university’s infrastructure creates responsibilities that most institutional data centres were not originally designed to accommodate. Service levels, physical access, security, change control, incident response, maintenance arrangements and capacity commitments all need to be defined.

The long-term questions are equally important. A host university may have capacity available today but require it for its own AI, HPC or institutional workloads in five years. Additional electrical or cooling investment may become necessary. Equipment may need to be relocated when the agreement ends.

The Universities UK research into shared services also identifies wider barriers to collaboration, including the capacity of institutions to implement change, lack of standardisation, difficulty identifying suitable partners and the VAT treatment of shared services.

These challenges do not make shared data centre capacity unworkable. Commercial hosting providers address many of the same issues through contracts, service definitions and mature operating models.

The difference is that most university data centres were developed to support their own institution, not to operate as commercial colocation facilities. Any university offering infrastructure capacity to another institution needs to understand the operational maturity and long-term commitment that creates.

For the same reason, not every university data centre should become a shared facility.

Some facilities are approaching the end of their useful life. Others lack the resilience, security, operational capability or electrical and cooling capacity required to host another institution’s infrastructure. Future institutional demand may also be too uncertain to commit capacity over meaningful periods.

The sector does not need every university to become a hosting provider. It needs to identify where suitable infrastructure exists and where sharing that infrastructure could create genuine value.

The bigger opportunity is better capital allocation

The most important argument for shared data centre capacity is not improved rack utilisation or the opportunity for universities to generate hosting income.

It is the possibility of making better infrastructure investment decisions across the sector.

Consider one university facing an immediate requirement for additional compute capacity. Its existing data centre is constrained, and expanding the facility would require several million pounds of capital investment.

Another institution has recently invested in resilient infrastructure with sufficient power and cooling capacity to support growth for many years. That capacity will not be fully utilised in the short term.

Today, those universities may make decisions independently. The first institution could invest millions expanding its own facility while suitable infrastructure already exists elsewhere in the sector.

A credible shared capacity model introduces another option.

The university can compare the cost, risk and strategic value of expanding its own facility with commercial colocation, public cloud, national infrastructure and capacity available from another institution.

The answer will not always favour shared hosting. Geography, connectivity, risk appetite, technical requirements and long-term demand will influence the decision.

But the existence of a credible alternative changes the investment conversation.

Over time, greater collaboration could lead to a more deliberate distribution of infrastructure investment. Some universities may develop facilities capable of supporting regional research computing and digital infrastructure requirements. Others may decide that operating significant data centre infrastructure is no longer strategically necessary.

Capital could be concentrated where suitable infrastructure, operational capability and long-term demand already exist, rather than every institution independently investing to solve similar problems.

That is a much more significant proposition than universities renting spare racks to one another.

The sector needs evidence before it needs a marketplace

The immediate barrier is information.

There is no sector-wide view of university data centre capacity, no common definition of what constitutes available hosting capacity and no standard approach to describing the resilience, security and operational capability of university facilities.

Creating a national marketplace would therefore be premature.

A more realistic starting point would be a small group of willing institutions prepared to establish common definitions, assess facilities against agreed technical and operational criteria, share high-level information about capacity and test the commercial and governance arrangements required to support real hosting requirements.

This would allow the proposition to be tested using actual demand and actual infrastructure.

The objective should not be to prove that shared capacity works everywhere.

It should be to establish where it works, what value it creates, which barriers prevent adoption and whether the model can be expanded.

If the proposition proves viable, the sector would have evidence to support greater collaboration.

If it does not, participating institutions would still have developed a better understanding of infrastructure capacity, future investment requirements and the practical barriers to sharing critical infrastructure.

A credible opportunity, but not a universal answer

Higher education already demonstrates that shared infrastructure can work. Jisc estimates that its services save the sector up to £300 million each year, with members receiving more than £3 in savings for every £1 invested.

That does not mean university data centres can simply replicate the Janet model. The technical, commercial and institutional challenges are fundamentally different.

Shared university data centre capacity will not replace public cloud, commercial colocation, national computing infrastructure or investment in university-owned facilities.

Nor should it.

The opportunity is to create another credible infrastructure option.

Universities facing significant capacity requirements should be able to consider suitable infrastructure elsewhere in the sector alongside the commercial and technical alternatives already available to them.

The Universities UK Transformation and Efficiency Taskforce has called for greater collaboration and more effective use of shared services and infrastructure.

Data centres should be part of that conversation.

Before universities invest millions creating more capacity, the sector should at least establish whether suitable infrastructure already exists, whether institutions are willing to share it and whether collaboration could result in better use of capital across UK higher education.

The answer may only apply to a relatively small number of institutions.

But given the cost, complexity and strategic importance of the infrastructure decisions now facing the sector, it is a question worth answering.

Sources

Universities UK, Towards a New Era of Collaboration, Transformation and Efficiency Taskforce, 2025.

Jisc, New report calls for practical reform in higher education shared services, September 2025.

Jisc, The Economic Benefits of Jisc, June 2023.

Jisc, Janet Network.

International Energy Agency, Data centre electricity use surged in 2025, even with tightening bottlenecks driving a scramble for solutions, April 2026.

UK Government, Accelerating Electricity Network Connections for Strategic Demand, March 2026.

University of Birmingham, University of Birmingham to host national computing centre, February 2026.