Quick answer: Data residency answers where data is stored today. Data sovereignty asks a harder set of questions: who ultimately controls the infrastructure, which governments can legally compel the provider, what happens when ownership changes hands, and whether the customer can actually get its data back intact and on a workable timeline if the provider fails. History gives a clear answer to what happens when organisations only ask the first question: UK public-sector customers lost services when UKCloud collapsed in 2022; 2e2 customers were asked to pay nearly £1 million to keep their own servers running; and Nirvanix gave customers two weeks to retrieve petabytes of data before shutting down. Location was never the problem in any of these cases.
This is the fourth piece in our data sovereignty series. Part 1 covered the residency illusion at a regulatory level. This piece goes further, into the operational and corporate-structure risks that location alone can’t capture and that only show up when a provider fails, gets sold, or gets acquired by an owner nobody vetted for that risk.
The Problem With Asking Only “Where”
Most sovereignty conversations still start and end with geography: is the data centre in-country? For years, that was treated as sufficient. It isn’t. A facility can be domestically located, domestically staffed, and marketed explicitly as “sovereign” and customers can still lose operational control the moment the provider runs into financial or ownership trouble.
The clearest way to see this is to break sovereignty into five distinct layers, each of which can fail independently of the others.
The Five Layers of Sovereignty
1. Location: Where are the data and its replicas physically stored?
2. Ownership: Who owns the provider, the facility, and the underlying infrastructure?
3. Jurisdiction: Which governments can legally compel the provider, or its parent company, to act?
4. Continuity: What happens to customer service and data access after insolvency, acquisition, or divestiture?
5. Exit control: Can the organisation retrieve usable data within a realistic period and at a predictable cost?
Location is the layer everyone checks. The other four are where the real risk lives and recent history in the UK and Europe provides a working case study for each one.
Layer 3 and 4 in Practice: UKCloud’s Collapse
UKCloud built its entire brand around being a sovereign alternative for UK government and public-sector customers: UK-owned, UK-operated, and UK-based. On October 25, 2022, UKCloud and its parent company, Virtual Infrastructure Group, were placed into compulsory liquidation. Computer Weekly: UKCloud in liquidation
Central government, local government, police forces, the Ministry of Defence, the NHS, Genomics England, and university research bodies were among the customers affected. The Cabinet Office had to brief departments and begin urgent transitions to alternative suppliers, and government records later confirmed the UK Cabinet Office was left £17.5 million out of pocket after underwriting the official receiver’s costs. The Register: UK government faces £17.5M shortfall in UKCloud liquidation Reporting on the aftermath described ongoing business continuity issues across government stemming directly from the collapse. The Register: Collapse of UKCloud caused biz continuity issues for UK.gov
The takeaway: a provider can satisfy the location layer perfectly by beingBritish-owned, British-operated, and British-hosted, but still fail its customers on continuity. Sovereignty marketing is not a continuity guarantee.
Layer 5 in Practice: 2e2 and the Ransom Letter
If UKCloud shows what happens to continuity, 2e2 shows what happens to exit control when a provider fails. When UK data-centre and IT services provider 2e2 entered administration in January 2013, administrators wrote to enterprise customers demanding roughly £960,000 within a week to keep the data centres powered on. Computer Weekly: 2e2 datacentre administrators hold customers’ data to £1m ransom
Customers, including large enterprises with equipment they legally owned sitting inside 2e2’s facilities, were told that without payment, the administrators could not guarantee continued building access, power, or network connectivity. The Register: Stricken 2e2 threatens data centres
The takeaway: owning your servers doesn’t mean you control access to them. If the servers sit inside someone else’s building, that building’s insolvency becomes your operational emergency.
Layer 5, Taken to Its Extreme: Nirvanix
Nirvanix is the starkest illustration of exit control failing entirely. After a funding effort collapsed, the cloud storage provider told customers they had roughly two weeks to retrieve their data before shutdown. Some customers held a petabyte or more, effectively the only copy of that data. InformationWeek: Nirvanix Declares Bankruptcy; Customer Data Pours Out
Nirvanix filed for Chapter 11 bankruptcy on October 1, 2013, and moving that volume of data through the available export pipes. With every other customer trying to do the same thing simultaneously, it was estimated to take nine to ten months under normal conditions, not two weeks. Data Centre Knowledge: Nirvanix Collapse Provides Stress Test for Cloud Migration IBM’s SoftLayer unit stepped in to help customers migrate, and Nirvanix extended its final removal deadline, but the underlying lesson stood regardless.
The takeaway: contractual ownership of data is meaningless without a tested, economically viable way to actually get it out on a timeline shorter than the provider’s runway.
Layer 4 Again: Sungard’s Assets Redistributed
Sungard Availability Services, a major UK and US provider of disaster recovery and managed infrastructure, entered UK administration in March 2022, followed by a US Chapter 11 filing. The Register: Sungard UK bit barns offloaded to Redcentric
Its operations didn’t simply vanish. Tthey were divided among buyers. Daisy Group took over Sungard’s UK workplace recovery customers, while Redcentric acquired three UK data centres along with Sungard’s consultancy and cloud practice; in the US, 365 Data Centres acquired the majority of the colocation and networking business. Redcentric: Acquisition of Sungard Availability Services data centres
The takeaway: even when service technically continues, the customer’s provider, contract counterparty, support organisation, and ultimate ownership can all change without the customer making or even being consulted on that decision.
Layer 2 in Practice: Global Switch and the Ownership Question
Location and ownership can diverge sharply, and Global Switch is the clearest recent example. In 2016, a consortium of Chinese investors led by Jiangsu Shagang Group acquired a 49% stake in UK-headquartered data-centre operator Global Switch for roughly £2.4 billion. Data Centre Dynamics: Chinese investors gain control of Global Switch
The building didn’t move. The facilities stayed in London, Sydney, and other major cities, fully operational. But the ownership change triggered real security scrutiny: Australia’s Department of Defence progressively wound down its use of a Global Switch facility over concerns tied to the new ownership, fully relocating its data and services by 2024, a year ahead of the lease’s scheduled expiry. iTnews: Defence vacates Global Switch a year ahead of lease expiry Earlier reporting had already flagged Australia’s difficulty in exiting a Chinese-owned data centre it considered a security risk. Caixin Global: Australian Government Struggles to Exit Chinese-Owned Data Centre
The takeaway: the address on the building tells you nothing about who holds beneficial ownership, or what happens to that ownership over the life of your contract. A change in shareholders can change your national-security risk profile without a single server moving.
A Quieter but Equally Important Example: TelecityGroup
Not every ownership change is dramatic, and that’s exactly the point. UK-based TelecityGroup, a major European colocation operator, was acquired by US-headquartered Equinix in a deal worth roughly £2.6 billion, completed in January 2016. Data Centre Dynamics: EC approves Equinix acquisition of Telecity
European regulators required Equinix to divest eight facilities in London, Amsterdam, and Frankfurt as a condition of approving the deal. Those facilities were then acquired by another US-headquartered company, Digital Realty. Equinix: Agrees to Divest Eight European Assets to Digital Realty Trust
The takeaway: a customer who deliberately chose a British or European operator could, within the same contract term, find their facility had become part of a US-controlled company twice in the case of the divested sites. No decision the customer made changed; the jurisdiction their infrastructure fell under changed anyway.
This Isn’t a Handful of Isolated Incidents
A 2025 academic review examined 775 non-U.S. data centre projects and found that U.S.-headquartered companies operate roughly 48% of them when weighted by announced investment value. A reasonable proxy for compute capacity. The researchers’ conclusion tracks exactly with the case studies above: locating infrastructure outside the United States does not, on its own, establish sovereignty when the operator remains a U.S. entity subject to U.S. jurisdiction. arXiv: How Sovereign Is Sovereign Compute? A Review of 775 Non-U.S. Data Centres
Legal analysis reaches the same conclusion from a different direction. When a European or UK subsidiary has a U.S. parent company, the parent’s obligations under the CLOUD Act don’t stop at the subsidiary’s border. A U.S. court can compel the U.S. parent to produce data held by its European subsidiary, and the subsidiary’s operational independence doesn’t sever that obligation. SoftwareSeni: How the US CLOUD Act and FISA 702 Create Legal Exposure for EU Cloud Data This is precisely why the EU’s current Tech Sovereignty Package is focused less on where servers sit and more on preventing any foreign government from holding an effective “kill switch” over infrastructure that hospitals, energy grids, and public institutions depend on. CNBC: Europe unveils tech sovereignty package amid growing concerns over reliance on U.S. tech
What This Means for Infrastructure Decisions
Run any of these five cases back through the framework and the pattern is consistent: the organisations that got hurt weren’t careless about location. UKCloud customers picked a UK-based, UK-marketed sovereign provider. 2e2 and Sungard customers had equipment sitting in domestic facilities. Global Switch’s government tenants had chosen a UK-headquartered operator, only to see beneficial ownership shift under them. In every case, the failure showed up in ownership, continuity, or exit control. All layers that a “data stored in-country” checkbox never tests.
That’s the argument for infrastructure an organisation owns and operates directly, rather than leases from a third party whose solvency, ownership, and corporate jurisdiction are outside the customer’s control. Nexsan’s approach to on-premises storage is built around exactly this distinction: when the hardware belongs to the organisation running it, there’s no provider that can go into administration and hold access hostage, no parent company that can be acquired by an unvetted foreign buyer, and no two-week countdown to migrate petabytes off someone else’s failing platform. The system, and the exit path, both belong to the customer from day one, not contingent on a vendor’s balance sheet.
The Bottom Line
“Where is our data stored?” is the easiest sovereignty question to answer and, on its own, the least useful one. Ownership, jurisdiction, continuity, and exit control are what actually determine whether an organisation retains control of its infrastructure and its data when something goes wrong, and UK case history over the past decade shows exactly how often something does. Testing all five layers, not just the first one, is what separates sovereignty as a marketing claim from sovereignty as an operational fact.
Read the rest of the series: Part 1: The Residency Illusion · Part 2: The €890M Google Fine · Part 3: Cloud Repatriation and TCO
Sources: Computer Weekly: UKCloud in liquidation · The Register: UK government faces £17.5M shortfall in UKCloud liquidation · Computer Weekly: 2e2 datacentre administrators hold customers’ data to £1m ransom · InformationWeek: Nirvanix Declares Bankruptcy · Data Centre Knowledge: Nirvanix Collapse Provides Stress Test for Cloud Migration · The Register: Sungard UK bit barns offloaded to Redcentric · Redcentric: Acquisition of Sungard Availability Services data centres · Data Centre Dynamics: Chinese investors gain control of Global Switch · iTnews: Defence vacates Global Switch a year ahead of lease expiry · Data Centre Dynamics: EC approves Equinix acquisition of Telecity · arXiv: How Sovereign Is Sovereign Compute? A Review of 775 Non-U.S. Data Centres · SoftwareSeni: How the US CLOUD Act and FISA 702 Create Legal Exposure for EU Cloud Data · CNBC: Europe unveils tech sovereignty package
Quick answers
Data residency describes only the physical location of data. Data sovereignty covers who owns and controls the infrastructure, which laws can reach the provider or its parent company, what happens to service continuity if the provider fails or is sold, and whether the customer can retrieve its data within a workable timeframe and cost.
UKCloud, a UK-based provider marketed specifically as a sovereign cloud for government and public-sector customers, was placed into compulsory liquidation on October 25, 2022. UK government departments had to migrate to new suppliers, and the Cabinet Office was left £17.5 million out of pocket covering liquidation costs.
Not fully. The 2e2 case showed that customers with legal ownership of their own servers still lost practical access when the hosting provider entered administration and building access, power, and network connectivity were put at risk pending payment.
Yes. Under the U.S. CLOUD Act, a U.S. court can compel a U.S. parent company to produce data held by its foreign subsidiaries, regardless of where that data is physically stored, and the subsidiary's operational independence does not remove this exposure.
When an organisation owns and operates its own infrastructure directly, there is no third-party provider whose insolvency, acquisition, or foreign ownership change can disrupt access, no lease or hosting contract that can be renegotiated under duress, and no data-egress countdown imposed by someone else's shutdown timeline.