For several years, European banks have been told that artificial intelligence will transform their operations. Most have responded with careful pilots, mainly in fraud detection and back-office automation, while a harder question went unanswered. If an AI system works on the most sensitive records a bank holds, whose servers does it run on, and whose laws apply to it?
In September 2026, two Stockholm-based technology companies announced a partnership that tries to answer that question directly. Vilja, a cloud-native core banking vendor, and evroc, a European-owned cloud and AI infrastructure provider, are joining forces to deliver agentic AI to financial institutions. The aim is for all the data and AI processing to stay under European control.
What was announced
The partnership connects Vilja’s core banking platform with evroc’s sovereign cloud and AI infrastructure. In the companies’ framing, European banks need the efficiency gains AI can deliver, and they also need certainty about where sensitive financial data is processed and which jurisdiction governs it. For agentic AI, that means keeping both the data and the AI processing inside a sovereign European infrastructure.
Agentic AI differs from the chatbots and recommendation engines most people know. It refers to autonomous software agents that can carry out multi-step tasks, make decisions within defined limits and act on them, rather than just answering a question. In a bank, that could mean an agent that gathers documents for a loan file, checks them against policy, flags exceptions and moves routine cases through the workflow without a person handling each step.
For such agents to be useful, they need access to the core banking system and the data inside it. That access is exactly what makes them sensitive. The partnership is built around this idea: the agents sit next to the core platform, on infrastructure that European banks and regulators can trust.
“Banks shouldn’t have to choose”
Mattias Åström, founder and CEO of evroc, said the partnership shows how European banking can stay both competitive and “in control of itself.” His wider argument is that banks should not have to choose between modern AI and authority over their own data, and that evroc was built to remove that trade-off.
Fredrik Ulvenholm, CEO of Vilja, put the case in practical terms. In his view, AI delivers its real value only when it connects to the core banking platform and the data behind it. Working with evroc, Vilja wants agentic AI to become an integrated part of everyday banking operations, while banks keep control of where their data and AI workloads reside. He described evroc as one of Europe’s leading innovators in sovereign cloud and AI infrastructure.
The message from both executives is consistent. AI that sits on the edge of a bank’s operations, disconnected from core data, offers limited benefit. AI that is deeply connected needs a trusted place to run.
Why sovereignty has become a banking issue
“Digital sovereignty” was once a phrase for policy papers and conference panels. It has become a procurement criterion. European banks operate under dense regulation covering data protection, operational resilience and outsourcing, and supervisors expect firms to understand and manage their dependence on third-party technology providers. When the third party is a large non-European hyperscaler, questions follow about foreign legal reach, concentration risk and the ability to switch providers if circumstances change.
Industry coverage of the announcement noted that the large non-European hyperscalers are under growing pressure to prove they can meet local data autonomy standards. Native European players such as evroc are positioning themselves to benefit from the trust that comes with European ownership and jurisdiction.
Agentic AI sharpens the concern. A traditional analytics tool may read data and produce a report. An autonomous agent can read, decide and act, and it may process customer records, transaction histories and internal risk assessments as it goes. The more autonomy a system has, the more a bank needs to know exactly where it runs, who can inspect it and under which legal framework it operates.
The market context
Adoption is still early, but the direction is clear. Analysis from GlobalData, published through Retail Banker International, estimated Europe’s enterprise agentic AI market at $634 million in 2024 and projected it to pass $5.5 billion by 2030. The same analysis found that adoption among European banks is at an early but progressing stage, driven by relatively low-risk, high-impact uses such as back-office automation and fraud detection.
The report was also candid about the obstacles. Implementing agentic AI is more complex than earlier generations of AI, and institutions should expect heightened data privacy and security risks, opaque “black-box” systems, ethical concerns and talent shortages. Its advice was for banks to proceed thoughtfully and to make decisions within a responsible AI framework, so that value is unlocked in a secure, compliant and ethical way.
The Vilja and evroc partnership speaks mainly to the first of those challenges. It does not remove the need for governance, human oversight or careful rollout, but it addresses a question that often stalls projects early: whether the underlying infrastructure is acceptable to compliance teams and regulators at all.
What each partner brings
Vilja supplies the banking side. The company describes its technology as having originated in one of Europe’s first digital banks, and it is now positioned as a leading Nordic cloud-native core banking platform. Its deposit and lending products are composable, so institutions can adopt an end-to-end solution or select specific components. The platform is designed to be highly automated and configurable, which suits agentic AI: the more consistent and well-structured the underlying processes, the easier it is for an agent to operate on them safely.
Vilja is headquartered in Stockholm and also has a UK headquarters in Leeds, where it serves building societies and digital banks. It has been building partnerships to widen its reach, including a 2024 agreement with global IT company FPT Software to help bring digital banking solutions to customers worldwide. Earlier in the AI journey, Vilja’s chief product officer took part in a 2025 roundtable on agentic AI-driven core banking transformation in Leeds, which suggests the company has treated the topic as strategic for some time.
evroc supplies the infrastructure side. Founded in 2022 and branded “The European Cloud,” the company set out to build what it calls Europe’s first secure, sovereign and sustainable hyperscale cloud. When it unveiled its plans in 2023, it pointed to a striking market trend. Between 2017 and 2022 the European cloud market grew five-fold, yet the share held by European providers fell from 27 percent to 13 percent, by evroc’s own figures.
The company’s cloud platform officially went live in July 2025, offering compute, storage and networking tailored to sensitive workloads in sectors such as defence, government, financial services, healthcare and critical infrastructure. It operates from partner data centres in Stockholm, Paris and Frankfurt and is building its own facilities, including a flagship site near Stockholm’s Arlanda airport designed for up to 10,000 GPUs, liquid cooling and very high rack power density. By 2030 it aims to run ten hyperscale data centres.
It has also built an AI layer on top. evroc has introduced roc, an enterprise AI agent powered by large models running on its European GPU infrastructure, and evroc Think, a sovereign AI platform for organisations that want to run advanced AI workloads without depending on non-European cloud providers. For a bank, that means the models, the compute and the data can all sit within one jurisdiction.
A track record in sensitive environments
A bank weighing a new infrastructure partner will want evidence that the provider can handle serious workloads. evroc’s recent announcements offer some. The Swedish Armed Forces selected the company to help develop the Swedish Interoperable Combat Cloud, a command-and-control platform that will handle classified data under Swedish, NATO and EU frameworks, with all information remaining under Swedish jurisdiction. In June 2026 it also announced a partnership with KPMG Sweden to help organisations assess regulatory exposure and resilience risks and then operate sovereign cloud and AI environments.
None of that guarantees success in banking, which has its own regulatory demands. But it shows a company accustomed to environments where where data lives matters as much as how fast it can be processed.
evroc is also well funded for a young infrastructure company. In March 2025 it raised €50.6 million in a Series A round, with backers including EQT Ventures, Norrsken VC, Giant Ventures and blisce. Åström has said publicly that he simply wants Europe to control its own destiny, a sentiment that runs through the company’s positioning.
What this could mean for banks
For banks, the partnership suggests several practical possibilities, though the companies have not published detailed product specifications or customer deployments, and readers should treat the following as areas of potential rather than announced features.
The first is operational automation. Tasks such as onboarding checks, document handling, payment exception management and routine servicing are natural candidates for agents that can work through multi-step processes. Because the agents would be connected to the core platform, they could act on live data rather than on exports or copies.
The second is regulatory comfort. A bank that can show its AI workloads run on European-owned, European-governed infrastructure may find internal approvals and supervisory conversations easier, particularly around outsourcing and data residency.
The third is strategic flexibility. Institutions worried about concentration risk can see a European alternative that integrates with their core system, rather than a bolt-on that adds yet another dependency on a non-European provider.
Customers may eventually see the results as well: faster responses, shorter waits for decisions and fewer manual errors. Whether those benefits arrive will depend on execution, on how carefully banks scope their first agentic use cases, and on how well human oversight is built into automated workflows.
The harder questions still ahead
Sovereign infrastructure is necessary for trust, but it is not sufficient. Several questions will determine whether agentic AI delivers on its promise in European banking.
Accountability is one. When an autonomous agent makes or recommends a decision, banks must be able to explain it to customers and regulators. Opaque systems sit uneasily with a sector built on audit trails. Model choice is another. Sovereign infrastructure helps only if banks can still access capable models, and the sector will watch how quickly European-hosted options keep pace with the frontier. Cost and scale matter too, since hyperscalers have enormous scale advantages, and a challenger must prove it can match their reliability and price over time.
There is also the human dimension. Automation changes jobs, and banks will need to retrain staff and redesign processes rather than simply switch agents on. The GlobalData analysis flagged talent shortages as a real constraint, and that applies on the bank side as much as on the vendor side.
A signal for the wider market
Even with those caveats, the announcement is notable because it joins two layers of the stack that have often been treated separately: the banking platform and the AI infrastructure beneath it. By pairing them from the start, Vilja and evroc are betting that European institutions will increasingly ask where their AI runs before they ask what it can do.
If that bet is right, the partnership could become a template for other combinations of European software and European infrastructure, in banking and beyond. If it is wrong, and banks continue to prioritise raw capability and cost above jurisdiction, the pair will have to prove that sovereignty can be delivered without a performance penalty.
For now, the companies are offering European banks a simple proposition: modern AI, applied close to the core, on infrastructure they can govern themselves. In a region where trust in data handling is a regulatory and commercial imperative, that is a proposition worth watching.
Vilja is a Stockholm-headquartered provider of cloud-native core banking technology. Its platform originated in one of Europe’s first digital banks and is today a leading Nordic core banking platform, serving banks, building societies and digital banks. Vilja’s composable deposit and lending solutions can be adopted end to end or in part, and the platform is built to be highly automated and configurable for rapid time to market and stronger operational efficiency. The company also has UK headquarters in Leeds.
evroc, “The European Cloud,” is a sovereign cloud infrastructure and AI services company headquartered in Stockholm, Sweden. Founded in 2022 and fully European-owned, it provides compute, storage, networking and AI services, including the enterprise AI agent roc and the sovereign AI platform evroc Think, for organisations with sensitive workloads. It operates cloud regions in Stockholm, Paris and Frankfurt, has development teams across Sweden, France and the United Kingdom, and is backed by European investors including EQT Ventures, Norrsken VC, Giant Ventures and blisce.
Cosmopolitan The Daily is a daily news and features publication that brings readers timely reporting and analysis on the stories shaping modern life, from culture and lifestyle to business, technology and the wider world. Its mission is to deliver clear, engaging and trustworthy coverage that helps a curious, globally minded audience stay informed.