The Rise of Regional AI Infrastructure
On September 11, 2024, in London, the CBRE Data Centre Forum brought together industry leaders to discuss the rapid transformation of data centre infrastructure driven by AI. Stelia CMO Paul Morrison joined Justin Nesbitt from nLighten on a panel moderated by Kevin Restivo to unpack the latest trends shaping AI deployment and compute capacity.
One resounding theme emerged: AI infrastructure is moving regional. While FLAP (Frankfurt, London, Amsterdam, Paris) markets have traditionally dominated, data suggests that ~100MW of GPU power has been deployed across Europe, largely outside these core hubs. Stelia’s research further indicates that ~75,000 NVIDIA H100 servers are now in operation across the UK and Europe.
Independent Operators Are Spearheading AI Compute
Non-hyperscale providers are playing a leading role in AI infrastructure expansion. Genesis Cloud, NexGen Cloud, and Northern Data Group are among the independent operators driving GPU deployments. This marks a shift from traditional cloud players, indicating that AI’s computational demands require specialized infrastructure solutions rather than generic hyperscale models. 
Emerging AI Availability Zones
New AI Availability Zones are forming across Europe in locations that offer lower-cost energy, proximity to emerging AI hubs, and improved data mobility. Key regions include:
- Oslo, Helsinki, and Iceland – Leveraging abundant renewable energy.
- Northern England and Portugal – Providing lower latency for enterprise AI workloads.
- Milan – Growing as a strategic AI compute center in Southern Europe.
This decentralisation is reshaping internet architecture, as AI compute shifts from CPU-based to GPU-driven infrastructure. Efficient data mobility and optimized GPU orchestration will be critical to supporting this distributed model.
The U.S. Still Dominates AI Compute
While Europe is making significant strides, the United States remains the leader in large-scale AI infrastructure.xAI’s 130MW facility in Tennessee, hosting 100,000 H100 servers, exemplifies the scale at which AI compute is being deployed. Globally, 2 to 2.5 million H100 servers are expected to be operational by the end of 2024, underscoring AI’s massive infrastructure footprint.
AI Execution Will Define the Next Decade
AI is no longer in the experimentation phase—it is now about execution at scale. With a $250 billion investment forecasted by 2025, AI infrastructure is set to become a central pillar of enterprise technology. The Middle East is emerging as a key player, investing heavily in AI-driven data centre expansion.
As AI models become more complex, latency bottlenecks and real-time inference will be the defining challenges for enterprises. Stelia believes that data mobility optimization and GPU orchestration will be essential to ensuring that AI-driven decisions are made instantly, not delayed by inefficient infrastructure.
Final Thoughts
The CBRE Data Centre Forum confirmed what industry experts have long anticipated: AI infrastructure is evolving rapidly, and enterprises must rethink their compute strategies to keep pace. The shift from hyperscale dominance to regional, decentralised AI infrastructure is underway, and success will depend on how well organisations optimise for AI execution, not just experimentation.
Want to learn more about how Stelia is enabling AI-first enterprises with purpose-built infrastructure? Reach out at connect@stelia.io or subscribe for insights on AI innovation and enterprise-scale execution.
