EU Leads Sovereign AI Cloud Deployments

The growth of AI continues unimpeded, creating new markets, accelerating new business ideas and making established business processes obsolete. At the same time, many nations have already created—or are in the process of creating—regulations for strict data control. A few examples are:

• The European Union (EU) has General Data Protection Regulation (GDPR) and the AI Act, which came into force in August 2024, with a gradual deployment across general and high-risk AI systems, aiming for completion by August 2026. It’s the first regional attempt to place regulations on AI.
• The United States put the Federal Risk and Authorization Management Program (FedRAMP) and an Executive Order (EO) on AI in 2023 and 2025, respectively. The first was limited to government, while the EO was designed to remove barriers for the development of AI. 
• Germany has BSI C5, focusing on compliance and data sovereignty for cloud providers.
• France has SecNumCloud, also focusing on certifying cloud providers on data sovereignty.
• China has multiple laws, including MLPS 2.0, Data Security Law, and Personal Information Protection Law, which regulate data localization and government access.

Other countries have similar laws, all of which differ in terms of application domain and complexity. The EU is regulating AI extensively, focusing on a compliance-heavy framework to ensure trust, transparency, and sovereignty in AI systems. On the contrary, the United States is relying on the market to regulate AI and lead with innovation. Furthermore, U.S. dominance in semiconductors (including AMD, Broadcom, Intel, NVIDIA, Marvell), coupled with the regulated export of processing platforms (AI Diffusion Rule), means the world will be split in different AI speeds of development.

The first wave of Generative Artificial Intelligence (Gen AI) and Large Language Model (LLM) training was, and still is, driven by massive data centers, huge training datasets, and a broad global focus. The success of current LLMs will lead to more targeted, vertical-specific applications that are applicable to single countries or even specific enterprises. When that happens, data localization, sovereignty, and security will become a much bigger concern, especially in healthcare, defense, automotive, and finance. These models will need to be trained in local infrastructure, in sovereign cloud environments (either public or private) with data collected, processed, and stored domestically. No foreign entity can be allowed to access this training data or model parameters, both of which need to be owned by a local company.  Open source controls are likely the solution in this case, where models in the future will be trained and open source, and enterprises will tune them with their own sovereign data. This still needs processing resources and for the tuning data to remain secure, but this market development is still evolving and is not widely deployed in the market.

Moreover, different companies across different verticals will have different motivations for implementing sovereign clouds, e.g., an industrial manufacturer may want to protect its data from foreign espionage, which may translate to data residency and not further organizational or technology structural changes. Other enterprises may have completely different priorities, meaning that hyperscalers, local cloud providers, and even Information Technology (IT) vendors for on-premise systems have an opportunity in this space. As of 2025, hyperscalers are showing signs of difficulty building AI infrastructure, meaning that the AI infrastructure deployment model will likely be hybrid and driven by neocloud providers, rather than hyperscalers themselves.

For the time being, most processing platforms, LLM models, supporting infrastructure, and software are developed and sold by U.S. companies. However, the above analysis indicates that domestic, or at least regional, operations and supply chains need to be established to fulfill these requirements. Except for China, no other market can come close to achieving this. ABI Research expects an acceleration of sovereign AI clouds in Europe and other large markets, including the Middle East, which will, in turn, help fuel the rollout of targeted LLMs and Agentic AI systems.

The EU is facing many challenges, including ecosystem acceleration, domestic technical capabilities (processing platforms and software), and cohesion across EU borders. There is also a talent gap in Europe, which follows the United States and China in terms of attracting AI professionals.  Nevertheless, companies like NVIDIA are building talent hubs around Europe to accelerate AI innovation on the continent.

But the continent is now well positioned to lead the sovereign AI cloud market and create a secure, sovereign, and regulated AI cloud for enterprise innovation. This is in line with the broader regulation landscape across Europe and can create global leadership for the continent. In the United States, the rate of AI model development is expanding like wildfire, often leaping much farther than enterprises can follow. This is why many Gen AI Proof of Concepts (PoCs) in many enterprise verticals fail, because enterprises are not ready to explore and take advantage of these new applications and processes. This is why Europe now has a chance to take a more measured approach, create cloud AI computing platforms, and help enterprises gradually create new AI applications, rather than rushing into Gen AI adoption. ABI Research expects more EU investment to funnel into homegrown cloud providers, including Aleph Alpha, Mistral AI, OVHcloud, and Scaleway, which will likely become continent-wide cloud providers in the long term. These companies have very different backgrounds, all of which have very different positioning and market development models:

• Telco Spin-Offs: Scaleway, Swisscom
• Neoclouds: Nebul, Nebius
• Model Developers: Mistral AI, LightOn, DeepL, iGenius

The regulatory environment is a concern for new entrants on the continent that may not be able to create scale and address a large market when the regulatory landscape is fragmented. However, potential future regulatory harmonization across Europe may provide a significant opportunity for them and will accelerate the market by making AI easier and safer to adopt and use.

European telcos have long struggled to monetize their networks and assets beyond consumer connectivity. Sovereign AI cloud is now a tangible opportunity for them. They know how to manage a distributed computing infrastructure, they have the trust of the public and private sector, and they are desperate for new business opportunities. EU telcos must invest time, effort, and capital in market initiatives like the AI ON consortium, while aligning with the EU on regulatory policy and aiming to attract funding from the commission. They should partner or even acquire neocloud companies and have an aggressive strategy to become the leading sovereign AI cloud owners in Europe. This will not only help them reach new markets, but will also position them well for 6G and future networks, where AI will likely play a crucial role.