Behind the growth of traditional and AI Data Centers in Europe: Sentiment vs Market Adoption

When it comes to technology, market sentiment has often diverged from actual adoption. The same holds true for AI today.

Although billions have been invested in the past few years (2022–2024) with major companies (tech giants, VCs, governments) injected huge sums into AI development; widespread adoption by enterprises is still limited.

While some companies have built AI into their operations, the vast majority are still only using AI via external tools (like ChatGPT, Microsoft Copilot, Salesforce Einstein, etc.) rather than developing their own bespoke AI systems. Full adoption across industries is still relatively slow.

This is an important consideration when investing in AI infrastructure and evaluating the commercial pipeline.

Additionally, the design of AI data centers varies depending on the type of
specific AI models, or applications (e.g., LLM training vs. inferencing vs. other ML tasks). AI infrastructure must be customized for specific use cases, GPU-heavy setups, cooling needs, high energy demands, and network architectures.
Moreover, dimensioning (capacity planning) for AI workload is still experimental—no fixed standards exist yet, and companies are still figuring out best practices.

These reflect hugely on AI data centers business models: 

colocation, AI dedicated cloud, AI infrastructure as a service IaaS, managed AI hosting, AI data center leasing/financial structuring, custom built AI data center enterprise owned.

AI data centers should be built with a strategy in mind. Business models developed in line with the topography of the data center location, land, power, and cooling. The latter is an aspect that is crucial, but still seems a bit underplayed. Everyone knows that the main difference between traditional data centers and AI data centers is the rack dimension, and how many servers can fit. We moved from 10-15 kW rack density in traditional data centers, to 125kW and even more expected in the next future up 600kW for AI data centers.

GPUs keep continue to become substantially more powerful with an evolving miniaturization that will create extremely powerful racks, generating so much heat with humongous needs of cooling and water. Data centers that build with sophisticated, and sustainable cooling systems, typically next to see weather, are data centers with a sustainable business model able to serve the AI workloads of the coming, soon future.

Building data centers takes time, also years, and the risk of  build something that become obsolete before going live, exist. Data centers - colocation ready - are a rarity, and the once that will go live in 2025/2026, have for sure a competitive advantage to potentially dominate specific market segments.

Paramount is business strategy and model, well pondered upfront; which usually is not technical teams forte, especially in startups.

These are some, of many aspects to take into account that are not present in traditional data centers.

In Europe the drives of traditional data centers growth has been, first waves, the migration from on premise to data center, then second waves from data centers to cloud, and now the booming of AI applications. 

These drivers are still going in parallel.

Digital transformation within enterprises are still ongoing although the could migration boost in 2020 from COVID, only 55% of SMEs reached at least a basic level of digital intensity (DESI 2023).

Larger companies have made more progress, but most smaller firms lag behind, especially in adopting advanced digital technologies like AI, cloud, and big data.

Advanced tech adoption (AI, IoT, cloud) remains limited outside of tech sectors.

Cloud Adoption Is Still Limited

• In 2023, only 42% of EU businesses reported using cloud computing services.

• And within that 42%, most only use basic services (like email hosting and file storage), not advanced cloud (like AI platforms, serverless computing, or cloud-native app development).

• Usage varies a lot by country:

• Finland, Sweden, and Denmark are leaders (around 70%+ cloud adoption).

• Germany, France, Italy, and especially Eastern/Southern Europe are lower than the EU average.

• Small businesses (under 50 employees) are particularly behind — often less than 30% cloud usage.

Sources:
• European Commission - DESI 2023 Report
• Eurostat - Cloud computing use by enterprises (2023)

Paradoxically some of the reasons of slowing adoption of cloud (and consequentially traditional data centers), like regulatory and data privacy, data sovereignty, GDPR, cybersecurity fear, are also the reasons why smaller AI data centers (not hyperscalers) will find a market in EU.

Invest in traditional data centers, and AI data centers, is different, although market adoption foundational behavior, in Europe, are the same.

Sustainability is paramount for AI Data Centers more than traditional one.

1- Land ownership and the ability to build and scale in the same land brings more advantages compare to, same capacity spread in different sites. 300MW in one site owned is more cost efficient, and less risky, and reduce lead time rather than 300MW spread between different sites and leasing contracts, which is typical of some specific geographies rather than others.

2- Sustainable power beside being carbon negative is less expensive for who runs the data centers and for their clients

3- Lead time to capacity. Greenfield vs brownfield. The right land should be also energized (brownfield), which usually takes time. Land without power doesn’t have value. A land with power can be valued 400k/MW, instead when a data center shell is ready to be online 750k/MW.

• Trends said that data centers AI ready in the next 1/2 years are the ones will conquer the market.

• Data centers colocation ready now are invaluable
  

• cooling systems next to the see weather are the most sustainable  environmentally, but also from a business stand point
  

Every data center will become an AI data centers, but not all data centers can, since the requirements to fulfill the two are strongly different, and this is where colocation come into place. Rather than competition, data centers are start collaborating between each other. A traditional data center bring experience, and anchor tenants (enterprise customers), the AI data center brings innovation, and needs also enterprise customers, along with startups, like neo clouds.

Data centers growing is expected between 15% to 27% CARGAR are due to AI booming, but on the adoption side, everything is still to be written and signed, like contract with multinational and SMEs.

CONLUSION: KEY MARKET TRENDS

• MASSIVE DEMAND FOR GPUs (especially Nvidia H100s and now Blackwell chips)

• ENERGY AND COOLING INNOVATIONS (liquid cooling, immersion cooling, nuclear-powered data centers)

• SHORTAGES in available AI data center capacity (2025–2026)

• PRIVATE CLOUDS FOR AI STARTUPS becoming very popular to avoid public cloud vendor lock-in

• ENHABXE SECURITY MEASURE: With the proliferation of AI technologies, securing data and applications in data centers has become a priority, prompting investments in advanced cybersecurity solutions.

  • Source: Cybersecurity Ventures, "Cybersecurity Market Analysis"