Across the United States, hospital operators are reconsidering ownership of their data centre infrastructure as growing demand for digital healthcare services, rising costs and ageing facilities drive a shift towards specialist colocation providers.
Across the United States, hospital operators are confronting a question that rarely features in public debates about healthcare: why should they own and operate the buildings housing their computing infrastructure? As medical records, diagnostic imaging and digital services expand, maintaining private data centres demands capital and specialist expertise that health systems could deploy elsewhere.
The shift is not necessarily dramatic. Hospitals are not abandoning their technology or handing over responsibility for patient information. Instead, some are reconsidering ownership of the physical infrastructure beneath their applications, using specialist colocation providers to deliver capacity, power, cooling and connectivity.
The University of Pittsburgh Medical Center (UPMC) offers a revealing example. The integrated healthcare provider and insurer operates 40 hospitals and more than 800 clinical locations, employing over 100,000 people. Years of expansion placed mounting pressure on ageing on-premises facilities, prompting UPMC to seek an alternative to another costly construction programme.
Its decision to deploy infrastructure with Ark, a specialist data centre operator, reflects a wider strategic proposition: hospitals can remain accountable for critical systems without owning every building supporting them.
For healthcare executives, the attraction is not simply outsourcing. It is redirecting investment and attention towards clinical services, digital capabilities and resilience.
The infrastructure burden
Healthcare computing has become harder to accommodate within facilities designed for an earlier generation of workloads. Electronic health records must be accessible across departments and locations. Imaging generates large files, research programmes add demanding storage requirements and expanding networks of clinics multiply the connections that infrastructure must support.
When capacity runs short, adding servers is only part of the answer. Organisations must also find electrical capacity, increase cooling, maintain physical security and build network redundancy. Each extension brings construction costs, operational risk and another layer of maintenance.
For a health system growing through new services and facilities, infrastructure planning can become a recurring constraint on its broader ambitions.
UPMC faced precisely this combination of pressures. According to the Ark case study, data growth was stretching its existing data centres, while limited connectivity options complicated plans for hybrid cloud architectures. Its leaders wanted to reduce the expense and complexity of managing facilities while preserving the performance and availability of systems used throughout the organisation.
That balance is unusually consequential in healthcare. A disrupted clinical application is not merely an inconvenience for office staff. Delays accessing records or diagnostic information can interfere with care delivery.
Infrastructure choices therefore require more than a comparison of rental costs against construction budgets. They demand scrutiny of continuity arrangements, security controls and accountability during failures.
From ownership to access
Following a competitive request for proposals, UPMC selected Ark to provide a customised colocation deployment. The arrangement gave the health system one megawatt of capacity in an Ark-owned Tier III facility, with room for future expansion. UPMC became an anchor tenant rather than financing a new data centre itself.
This changes the economics and responsibilities of infrastructure without removing the need for internal technical leadership. Ark manages the underlying facility environment, including redundant power, cooling and physical safeguards. UPMC retains responsibility for decisions about applications, information governance, access and the way its systems are designed and operated.
That distinction is central to understanding why colocation appeals to healthcare organisations. They are not necessarily seeking to outsource IT strategy. They are separating the ownership of buildings from control over technology, allowing specialised operators to handle facilities that require substantial engineering investment and constant attention.
Ark’s carrier-neutral connectivity and cloud on-ramps also addressed a weakness in UPMC’s previous arrangements. Rather than treating the data centre as an isolated destination, the model provides connections to external networks and cloud services. Applications can potentially move between dedicated infrastructure and public or private cloud environments as requirements change.
For hospitals with deeply integrated legacy systems, that flexibility matters. A wholesale move to the public cloud may be impractical, while continuing to host everything on-site can restrict expansion. Colocation provides an intermediate model, although successful migration still depends on careful testing, network design and the management of interdependent applications.
Security does not move with the building
Moving systems outside a hospital’s own property can sound like surrendering control, particularly when patient information is involved. But the relevant question is whether the new operating model provides appropriate protections and clear responsibilities, not simply who holds the title to the building.
Ark’s facility includes biometric access controls and round-the-clock monitoring. Redundant power, cooling and network infrastructure are intended to protect availability, while high-density racks accommodate changing workloads. These measures can improve the physical foundation supporting healthcare applications, especially when compared with ageing facilities facing investment constraints.
The case study describes Ark’s facilities as HIPAA compliant and certified. That language requires qualification: HIPAA does not offer an official government certification for data centres. Compliance depends on the applicable safeguards, agreements and ongoing conduct of healthcare organisations and their service providers. Physical security alone cannot guarantee that patient information is protected.
UPMC must still address encryption, identity controls, cybersecurity monitoring, incident response and access to sensitive records. Its applications also remain exposed to risks beyond the colocation facility, including software failures and problems with external networks.
The move changes the operating environment; it does not transfer away ultimate responsibility for patient data or clinical continuity.
The numbers behind the decision
Ark reports that the arrangement allowed UPMC to avoid US$100 million in capital expenditure associated with building and maintaining its own data centre infrastructure. The supplier’s published account does not provide a detailed calculation or measurement period, so the figure should be treated as a reported outcome rather than an independently verified saving.
Even so, the scale illustrates why the ownership question matters. Hospitals face competing demands for investment, from clinical equipment to digital tools and facilities used directly by patients. Spending heavily on another computing building can restrict choices elsewhere, particularly when capacity requirements are difficult to predict years in advance.
UPMC also reports 50% more infrastructure capacity than under its previous in-house arrangement. That headroom can support growing medical datasets, additional services and phased application migrations without an immediate return to construction planning. It is flexibility purchased through access to a specialist facility rather than ownership of another asset.
Ark cites 99.999% infrastructure availability, a five-nines benchmark associated with very limited downtime. It should not be read as a promise that every UPMC application will always remain available. Clinical service continuity depends on software, networks, backup arrangements and operating practices extending beyond the data centre walls.
This shift also changes how success should be measured. A modern facility is valuable only if it helps clinicians access dependable systems, supports new services and gives technology teams room to respond to changing demands. Gains must be assessed across total costs, operational performance and patient care outcomes.
A quieter transformation
The significance of UPMC’s move extends beyond one procurement decision. It illustrates how healthcare infrastructure strategy is evolving from a question of what organisations can build to what they genuinely need to own. The distinction becomes more important as hospitals confront growing data volumes, ageing facilities and pressure to demonstrate value from technology expenditure.
This is not evidence that every American health system is exiting the data centre business, nor that colocation is always cheaper or safer. Geography, existing investments, application dependencies and contractual terms can change the calculation. Some organisations may retain substantial on-premises capacity while outsourcing selected workloads or using cloud services for others.
The transition also creates new dependencies. Healthcare customers must examine service agreements, disaster recovery provisions, network diversity, exit arrangements and the financial stability of their partners. Outsourcing a facility without preserving operational visibility could replace one category of risk with another.
Nevertheless, the strategic direction in UPMC’s case is clear. Infrastructure is being treated less as a property portfolio and more as a service supporting clinical and organisational priorities. Its leaders can concentrate on modernising applications rather than repeatedly expanding the physical estate beneath them.
“We appreciated the elasticity of Ark’s colocation model,” said McCallister in the case study. “It gave us the flexibility we required to modernize applications and to once again lead the industry in our approach to IT.” For health systems weighing their next major infrastructure investment, that may be the most important lesson. The future of hospital computing need not depend on hospitals owning more data centres. It may depend on knowing which responsibilities must remain inside the organisation and which buildings no longer need to.
