When Amazon Web Services (AWS) set out to build its European Sovereign Cloud, the ambition was clear: deliver the full AWS hyperscale experience – every API, every service, every performance benchmark – while guaranteeing that customer-created metadata, operations and infrastructure remain entirely within the EU. 

The AWS European Sovereign Cloud represents a massive engineering undertaking.  It replicates AWS' entire infrastructure stack within Europe, for Europe, and was completed in half the time it took to build regions just three years ago. 

Automation is the engine behind accelerated data centre deployment

Colm MacCárthaigh, VP and Distinguished Engineer at AWS, has spent 18 years at the company and several of those years leading the design and delivery of the European Sovereign Cloud. His background is in networking and data centres, and he draws on a bovine analogy to explain the philosophy underlying modern data centre construction at AWS scale.

"There's a saying when it comes to servers – that you want to manage servers more like cattle than like pets," he says. "When you get to a certain scale, you don't want to know the individual names of your servers anymore. You instead want to have automated everything, templated everything, and you're not thinking about naming individual servers. That's a sign of when you're at scale."

That principle now extends to the data centres themselves. AWS operates 39 geographic regions globally, a number Colm acknowledges has grown well beyond what any individual could catalogue from memory. "We have to manage and build those now through automation first," he says. "Consistency and templatising – all of that becomes very, very important, and it makes a huge difference for our customers, because we're also able to build much more quickly because of that automation."

The payoff for the AWS European Sovereign Cloud was measurable. AWS built the new offering in half the time previously required for a standard region – and this despite the fact that the AWS European Sovereign Cloud demanded additional services and back-end systems that a typical region would not include. "We actually built even more than we usually do, in half the time," Colm says. "All of that is driven through automation. That's the main thing that gets us there."

AWS’ engineering strategy combines proven design with flexible infrastructure

Beneath every cloud deployment is a physical reality – racks, cables, power and cooling – and the AWS European Sovereign Cloud is no exception. The engineering strategy AWS applied was to replicate the entire stack: services, operational systems and data centre technician tooling, assembled as a fully autonomous copy running within Europe.

"We decided with the AWS European Sovereign Cloud that we would replicate the entire stack – the full AWS hyperscale services and cloud, our full-scale systems and services, and everything it takes to operate those, including the back-end systems and everything our operators or data centre technicians need to rely on," Colm explains. "We would take that entire set of systems and build it again as a fully autonomous, separate copy running in Europe."

That approach relied on proven infrastructure designs, applied consistently across physical deployments, while retaining enough flexibility to respond to demand that could not have been predicted at the time of planning. The rapid evolution of AI workloads is the clearest example. "When we decided to build the AWS European Sovereign Cloud, there would have been no way to predict which AI instances were going to be the most popular, or how many customers were going to need them," Colm notes. "Those are fields that change so quickly, so we also have to have enough flexibility and agility in our designs to be able to pivot at a moment's notice and respond to what our customers need."

That balance between consistency and adaptability – templated enough to deploy quickly, flexible enough to absorb shifts in workload demand – is what Colm describes as central to the entire build strategy.

Localisation and replication are central to meeting sovereignty requirements

For regulated customers across Europe, assurance is not a secondary concern – it is often the deciding factor in cloud adoption. AWS has developed a Sovereignty Reference Framework, translating compliance requirements from customers and regulators across Europe into hundreds of documented evidence points.

"In the case of the AWS European Sovereign Cloud, our localisation and replication really focused on meeting our customers' assurance needs," Colm says. "We wanted to be sure that customer data and customer-created metadata always stayed within the EU, and that only our EU staff had any permissions or authority over that data."

The physical architecture supports this: data centres and availability zones built exclusively for the AWS European Sovereign Cloud, logically and physically separate from the broader AWS estate, operated exclusively by EU residents, and with no critical dependencies on systems outside the EU. Colm is careful to note that AWS’ standard regions are already sovereign by design, but the AWS European Sovereign Cloud went considerably further. "We also built additional back-end systems, as well as billing, identity and access management systems," he says.

The compliance-focused evidence packaging matters because many of the customers the AWS European Sovereign Cloud is designed to serve are themselves regulated entities – public sector bodies, financial institutions and healthcare organisations that must demonstrate adherence to specific frameworks. 

By building the evidence packages in advance, AWS has sought to reduce the compliance burden on customers rather than leaving them to map their own requirements against the infrastructure independently.

The level of detail AWS has applied extends well beyond the headline architecture. Colm points to Route 53, the company's domain name system (DNS) service, as an illustration of how deep the EU-specific engineering runs. "Internally, that also has to use its own DNS names. Well, in the EU, we've made sure those are EU-based DNS names – they end in .eu, or .de, and so on." 

Certificates issued to AWS services or customer endpoints are generated by AWS’ first European Certificate Authority, with keys managed entirely within the EU. "That fine level of detail extends to the access control systems we use, which are all maintained inside the EU," Colm adds.

Sustainability performance improves as customers move to the sovereign cloud

Data centre efficiency has become a significant factor in procurement decisions, particularly among European public sector and regulated enterprise customers. AWS cites a global power usage effectiveness (PUE) of 1.15 across its data centres, with its best-performing European sites reaching as low as 1.04. The AWS European Sovereign Cloud uses the latest generation of AWS instance types, including Graviton processors, which Colm identifies as central to those efficiency gains.

"Systems like our Graviton processor have been a huge part of those efficiency gains, which all feed directly into sustainability," he says. Research by Accenture puts AWS data centres at up to 4.1 times more efficient than on-premises alternatives – a figure that becomes particularly relevant in the European sovereign market, where many prospective customers have historically been constrained to on-premises infrastructure or, in some cases, equipment housed in office data closets.

This is where Colm sees a compounding benefit specific to the sovereign cloud context. Many of the organisations the AWS European Sovereign Cloud is designed to reach have not previously been able to migrate to public cloud at all, having been held back by regulatory constraints or a lack of sovereign assurance. Their baseline is often ageing, inefficient, on-premises hardware. The move to the AWS European Sovereign Cloud therefore delivers a sustainability dividend on top of the operational and security improvements it offers.

"As they're now able to adopt the full sovereign cloud, there is a massive sustainability win, because what we've built is so much more efficient than where they're coming from," Colm says. "Those customers are motivated by our security and availability benefits too, but it's a massive win that we're all able to get this sustainability benefit as part of that journey."

AWS scales and innovates without compromising sovereignty standards

One concern that prospective customers frequently raise is whether operating within a sovereign framework means accepting a reduced or trailing service. AWS’ position is that the AWS European Sovereign Cloud runs the same software, services and systems as its global regions, including its leading AI capabilities, and that the consistency of approach underpinning the build is precisely what makes that possible.

"The AWS European Sovereign Cloud runs our same full hyperscale services, the same software and the same systems that we run elsewhere. We haven't had to make compromises or cut corners or reduce our offerings," Colm says. "We're able to offer our leading-edge AI services in the AWS European Sovereign Cloud. That's very much what we want to provide for customers – we don't want customers who have sovereignty needs to have to accept anything less than the best."

The operational model reflects the same logic. Although the AWS European Sovereign Cloud has dedicated operators, Colm describes the day-to-day running of the infrastructure as consistent with AWS’ standard approach to managing all of its systems. "Our consistency in approach and all of that automation means that for us this is business as usual," he says.

For Colm, the message to customers ultimately rests on two pillars: familiarity and rigour. "The European Sovereign Cloud is the AWS that you know and love. It's all of our APIs and services, a full hyperscale cloud, with the benefits that we have everywhere. 

“This includes, for example, the AWS Nitro System – purpose-built hardware and software that provides the underlying platform for all Amazon EC2 instances. By design, the Nitro System has no operator access, meaning there is no mechanism for any system or person at AWS to access customer data running on EC2 instances. 

“Only the customers have visibility and control over their own data. But then, on top of that, what we add in the AWS European Sovereign Cloud is those dedicated operators. We've also dotted every ‘i’ and crossed every ‘t’, and made sure that every dependency we have is within the EU. We've gone really, really deep on that."