Europe’s financial system depends on a relatively small number of technology providers. DORA has expanded visibility into those relationships, including relevant subcontractors, and supervisors are already analysing concentration risks. The next challenge is to establish how effectively that information identifies shared points of failure and supports credible recovery planning across institutions.
On Tuesday, 20 October, a power-related failure affecting Google’s London cloud infrastructure disrupted banking and other digital services across the UK. Customers of financial institutions including Revolut, Lloyds and HSBC experienced problems accessing certain services, while businesses and public institutions encountered similar disruption.
The incident was not a banking crisis. Customer deposits had not suddenly become impaired, nor was there evidence of a systemic financial failure. Yet it demonstrated how an infrastructure problem at a technology company can quickly become an operational problem for financial institutions and their customers.
The outage also exposed a question of growing importance to European financial stability: how much of the banking system ultimately depends on the same underlying technology providers?
For decades, banking regulators have concentrated on identifying financial interconnections between institutions. The global financial crisis demonstrated how exposures that appeared manageable individually could become dangerous when viewed collectively.
Today, another form of interconnectedness is developing. Banks increasingly depend on cloud computing providers, software vendors, data centres and technology subcontractors to operate essential services.
Unlike traditional financial exposures, however, these dependencies are not necessarily visible on a bank’s balance sheet. Some may be buried several layers down in its technology supply chain.
Europe has established a framework to regulate this risk and is using the information it collects. The question is how comprehensively that framework can identify and test the consequences of common infrastructure failures.
A banking system dependent on a handful of technology companies
The European Central Bank has repeatedly warned about the concentration of technology outsourcing.
In its 2025 analysis of outsourcing arrangements, based on year-end 2023 data, the ECB found that half of the expenditure on externally outsourced critical services was concentrated among just 30 providers.
The analysis also highlighted the difficulty of reducing these dependencies. Approximately 82 per cent of critical functions outsourced to external providers were classified as difficult or impossible to substitute.
Cloud adoption has been expanding alongside this dependence. The ECB’s 2024 stocktake found that cloud services accounted for approximately 20 per cent of significant banks’ outsourcing contracts, half of which involved critical or important functions.
These figures should not be mistaken for a measurement of the proportion of banking operations running on cloud infrastructure. The contractual data do not establish that. Nevertheless, they demonstrate the importance that external technology providers have acquired.
The concentration is not difficult to explain. Amazon Web Services, Microsoft Azure and Google Cloud offer infrastructure at a scale that few competitors can match. Their platforms provide computing power, storage, cybersecurity capabilities, development tools and access to increasingly sophisticated artificial intelligence services.
For banks, the commercial case is compelling. Cloud platforms can reduce the need for expensive internal infrastructure while providing flexibility and resilience that would be difficult to replicate independently.
But there is a trade-off. As more institutions use the same infrastructure providers, the benefits of scale can be accompanied by shared points of failure.
A technology arrangement that improves resilience for an individual bank does not necessarily improve resilience for the financial system as a whole.
The concentration beneath software suppliers
The more difficult challenge may not be the cloud services that banks purchase directly, but those they depend on indirectly.
Consider a hypothetical European bank that contracts with three independent technology companies.
One provides regulatory reporting software. Another supplies a loan portfolio management platform. The third operates a payment-processing application.
From a procurement perspective, the bank has diversified its suppliers. Each company has a separate commercial agreement, its own management team and responsibility for delivering its service.
But suppose all three vendors host their applications on the same cloud infrastructure.
The bank now has three different suppliers, but potentially a shared infrastructure dependency. That does not by itself establish a common point of failure: the exposure depends on the cloud services, regions, configurations and recovery arrangements involved.
A disruption affecting the underlying cloud provider could compromise several services simultaneously, notwithstanding their apparent commercial independence.
This problem becomes more complicated when the suppliers themselves use subcontractors, which may engage further technology providers.
The ECB’s February 2025 analysis found that outsourcing contracts involved an average of four subcontractors.
A bank may consequently depend on infrastructure with which it has no direct contractual relationship, limited technical familiarity and restricted ability to negotiate recovery arrangements.
The key distinction is between supplier diversification and infrastructure diversification.
They are not the same thing.
Moreover, knowing the name of an underlying cloud provider is only part of the answer. Different applications may depend on a particular cloud region, data centre, identity-management service, network component or database platform.
Two applications hosted by different suppliers may share a vulnerability. Two applications hosted by the same cloud company may, conversely, operate across genuinely independent environments.
Understanding concentration risk requires more than identifying the company issuing the invoice.
It requires understanding the infrastructure supporting the service and what would happen if an important component failed.
DORA has improved oversight, but the job is not finished
European regulators are not unaware of these vulnerabilities.
The Digital Operational Resilience Act, known as DORA, became applicable in January 2025. It introduced a more comprehensive framework for managing information and communication technology risks across the EU financial sector.
Among its requirements, financial institutions must maintain registers of contractual arrangements with ICT service providers, assess concentration risks and manage dependencies associated with critical or important functions.
Crucially, the reporting requirements also extend to subcontractors that effectively underpin ICT services supporting critical or important functions or material parts of them. This includes relevant subcontractors whose disruption would impair the security or continuity of the service.
In the hypothetical example above, the underlying cloud provider would generally belong in the register if its hosting meets that test. A direct contract between the bank and the cloud provider is not necessary. Banks therefore need sufficient information from their direct suppliers to identify the relevant supply chain.
It would be incorrect to suggest that European regulation ignores indirect technology dependencies, or that the register merely lists the companies with which a bank contracts directly.
The European Supervisory Authorities have also designated 19 critical ICT third-party service providers, including Amazon Web Services, Microsoft, Google Cloud and Oracle, under DORA’s oversight framework.
The ECB’s 2025 Annual Report on supervisory activities, published in 2026, confirms that the first DORA register data collection identified increased dependence on ICT providers, concentration among a small number of suppliers — particularly cloud providers — and difficulties replacing or reintegrating outsourced services.
The supervisory picture is therefore already more developed than a collection of unanalysed supplier lists. The question is how far the information and existing supervisory work can support analysis of specific, correlated failure scenarios.
DORA’s reporting framework captures material supply-chain dependencies, but it does not automatically produce a comprehensive technical map of every shared infrastructure component or the consequences of losing it.
DORA also addresses resilience testing, continuity and recovery; those subjects are not absent from the framework. Nevertheless, identifying a dependency and documenting a recovery arrangement are different from demonstrating that recovery will succeed when several institutions and suppliers face disruption simultaneously.
The quality of the underlying information is equally important. During the voluntary 2024 preparatory exercise for DORA reporting, 6.5 per cent of the registers analysed passed every data-quality check. However, half of the remaining registers failed fewer than five of the 116 checks. The authorities considered the overall quality consistent with expectations for a best-effort preparatory exercise.
These figures describe that initial exercise, not the quality of current registers or the proportion of banks that presently misunderstand their dependencies. They illustrate the practical difficulty of producing consistent reporting, but cannot establish the scale of any current supervisory blind spot.
The challenge is to maintain reliable dependency information and connect it to technical analysis and evidence from testing.
What would happen if a major cloud provider failed?
Imagine a prolonged disruption at a major cloud provider affecting an important European region.
Some banks might lose customer-facing applications. Others could experience interruptions to internal risk systems, reporting processes or operational services. Technology vendors serving multiple institutions might be affected simultaneously.
The precise consequences would depend on the services involved, the architecture of the affected applications, the availability of independent recovery systems and the duration of the disruption.
Not every cloud outage would become a systemic event. Major cloud platforms also have significant resilience capabilities, and banks can design services to withstand individual infrastructure failures.
But assessing systemic risk requires asking questions that extend beyond the individual institution.
Which banks depend on the affected infrastructure, directly or indirectly? Which critical functions would be impaired? How many institutions could lose the same service simultaneously? How long would recovery take? Could alternative providers absorb the demand if many institutions attempted to switch at once?
And perhaps most importantly, could European supervisors answer these questions before an actual disruption rather than reconstructing the dependencies afterwards?
Public sources do not establish how completely supervisors can answer these questions across all material banking services. That is a limitation of what an outside observer can assess, rather than evidence that supervisors lack the necessary information or analytical capabilities.
Supervisors hold confidential data, DORA has materially expanded their visibility, and published findings show that concentration analysis is already taking place.
The issue for public scrutiny is how the completeness, technical depth and practical validation of that picture are assessed, and what can be disclosed about the results without compromising security or confidentiality.
The ECB appears to recognise the importance of this distinction. Its supervisory priorities for 2026–2028 include an examination of selected banks’ preparedness for disruptions caused by a major cloud provider.
This is a constructive, relevant part of the existing supervisory response. A further question is how findings from such reviews are combined with provider oversight and other supervisory information to assess common failure scenarios across institutions. The published description alone does not establish the full scope of that work.
Building on existing dependency reporting
The priority should be to extract greater value from information already collected, identify any demonstrated gaps and avoid duplicating existing registers or supervisory processes.
A useful objective is a confidential, sector-wide view that connects banks’ critical functions to their relevant technology dependencies and supports analysis of common failure scenarios. This would build on DORA reporting, concentration analysis and critical-provider oversight already in place.
Where provider-level information is insufficient for a particular scenario, supervisors could assess whether additional detail about infrastructure regions, services or shared technical components is necessary. The appropriate level of detail would depend on the risk being examined; a continuously complete inventory of every technical component may be impractical and unnecessary.
Any additional information request should address a specific analytical need and, where possible, reuse information available from banks, material ICT suppliers and existing oversight activities.
This would require stronger cooperation between financial institutions, software vendors, cloud companies and regulators.
It would also require a distinction between contractual substitutability and genuine operational substitutability.
A bank might possess the contractual right to terminate an arrangement and appoint another supplier. That does not mean it could migrate a critical application within the time required to prevent serious disruption.
Moving between cloud providers may involve changes to databases, application architecture, security arrangements and operating procedures. Those processes can take considerable time and may introduce further operational risks.
The solution is not necessarily to require banks to abandon cloud computing or duplicate every system across multiple providers.
Doing so could impose substantial costs, increase architectural complexity and, in some circumstances, introduce new vulnerabilities.
Instead, regulation should be proportionate to the importance of the affected banking function.
For the most critical services, supervisors should require credible evidence that continuity and recovery arrangements have been tested against realistic common-provider failures.
At a sector-wide level, regulators could then conduct coordinated scenarios that examine not merely whether an individual bank can recover, but whether many banks attempting to recover simultaneously create additional vulnerabilities.
The resulting information need not reveal commercially sensitive details to the public. Regulators could publish aggregated indicators of concentration and resilience while retaining the underlying technical data confidentially.
From financial concentration to operational concentration
Banking regulation has become increasingly sophisticated in assessing concentrations of financial risk.
Supervisors examine exposures to individual counterparties, connected borrowers, industries and geographic markets. They assess whether apparently separate exposures could deteriorate together under adverse circumstances.
DORA applies a related discipline to ICT third-party dependencies. The continuing task is to make that discipline effective for the technical and operational characteristics of common technology failures.
Two banks may have no material financial exposure to one another yet share a dependency on the same cloud platform, software provider or infrastructure component.
A failure in that shared infrastructure could affect both institutions at the same time.
The analogy with financial contagion should not be overstated. An operational interruption is not the same as a solvency crisis, and the consequences will vary substantially between institutions.
Nevertheless, common operational dependencies represent a form of correlated risk that deserves rigorous measurement.
DORA has established reporting, risk management and oversight arrangements, and the ECB is already analysing concentration and prioritising cloud resilience. The challenge is to connect this work to sufficiently detailed, tested assessments of how common failures could affect the financial system.
Europe already knows that its banks depend on a relatively small number of powerful technology providers.
The continuing supervisory objective is to establish where those dependencies converge, which critical services could fail together and whether recovery arrangements would remain credible during a prolonged, shared disruption. Public reporting of aggregated findings could make progress easier to assess without exposing sensitive infrastructure details.
The next major advance in banking supervision may not come from another financial ratio, but from understanding the technology infrastructure on which the financial system increasingly depends.
Sources and references
The following sources informed this article. All links are direct to the original publications or reporting, without ChatGPT tracking parameters.
Recent events
- Reuters — Google Cloud outage in UK affects digital banks and public-sector services, 20 October 2026
- Google Cloud — Official European service health and incident information
ECB research and supervision
- ECB — Outsourcing trends in the banking sector, 19 February 2025
- ECB — Annual Report on supervisory activities 2024
- ECB — Annual Report on supervisory activities 2025
- ECB — Guide on outsourcing cloud services, final publication, 16 July 2025
- ECB — Supervisory priorities 2026–2028
EU legislation and regulatory implementation
- European Union — Regulation (EU) 2022/2554: Digital Operational Resilience Act (DORA)
- European Union — Commission Implementing Regulation (EU) 2024/2956: ICT third-party registers and supply-chain reporting
- European Supervisory Authorities — Designation of critical ICT third-party service providers, 18 November 2025
- European Banking Authority — Findings from the DORA register reporting dry run, 17 December 2024
- European Banking Authority — Clarification on ICT subcontractors and the DORA register, Q&A 2024_7089
Editorial note: The October 2026 outage affected UK-based infrastructure, and the UK operates a distinct regulatory framework from the EU. The incident illustrates common infrastructure dependency; it does not establish that an EU-wide systemic banking disruption occurred. This article distinguishes existing DORA obligations and published supervisory activity from the author’s proposals. The absence of a publicly available technical dependency map is not evidence that supervisors lack confidential information or relevant analytical capabilities.